Dr Matthew Blackledge
Group Leader: Computational Imaging
Biography and research overview
Dr Matthew Blackledge is conducting a three-year independent research project on medical image analysis of bone metastasis at The Institute of Cancer Research, London.
Dr Blackledge's research seeks to provide clinical imaging methodologies for evaluating and understanding response mechanisms associated with novel therapeutics for bone metastases. Dr Blackledge collaborates with Professor Martin Leach, David Collins and Dr Dow-Mu Koh within the Cancer Research UK Cancer Imaging Centre at the ICR, which provides a unique opportunity to investigate techniques that can be quickly translated for the clinical management of patients in The Royal Marsden.
Bone metastases arising from a range of primary tumour sites have severe consequences for the patient, and although many new treatments are becoming available, there is currently no standard technique for early stage detection of response. Dr Blackledge is hoping to uncover novel imaging methodologies that incorporate whole-body MRI in conjunction with PET/CT and innovative image analysis techniques to provide a biological map of areas of response throughout the body during treatment. These techniques provide a rich data set of quantitative information including tumour volume, tumour cellularity, tumour metabolism and bone status that can indicate regions of response and provide image-guided treatment for patients.
Dr Blackledge is also working on new statistical techniques to quantify and visualise regions of heterogeneous response to treatment. Recent studies have clearly identified differences in the evolution pattern between distinct metastases throughout the body. The genomic differences in these mutated lesions can lead to differential treatment resistance. By using medical imaging techniques such as MRI and PET/CT it is possible to non-invasively probe individual metastases, allowing quantitative evaluation and visualisation of heterogeneous response throughout the body.
Dr Blackledge is a member of the Cancer Research UK Convergence Science Centre, which brings together leading researchers in engineering, physical sciences, life sciences and medicine to develop innovative ways to address challenges in cancer.
Related pages
Types of Publications
Journal articles
OBJECTIVE: To compare geometric distortion, signal-to-noise ratio (SNR), apparent diffusion coefficient (ADC), efficacy of fat suppression and presence of artefact between monopolar (Stejskal and Tanner) and bipolar (twice-refocused, eddy-current-compensating) diffusion-weighted imaging (DWI) sequences in the abdomen and pelvis. MATERIALS AND METHODS: A semiquantitative distortion index (DI) was derived from the subtraction images with b = 0 and 1,000 s/mm(2) in a phantom and compared between the two sequences. Seven subjects were imaged with both sequences using four b values (0, 600, 900 and 1,050 s/mm(2)) and SNR, ADC for different organs and fat-to-muscle signal ratio (FMR) were compared. Image quality was evaluated by two radiologists on a 5-point scale. RESULTS: DI was improved in the bipolar sequence, indicating less geometric distortion. SNR was significantly lower for all tissues and b values in the bipolar images compared with the monopolar (p < 0.05), whereas FMR was not statistically different. ADC in liver, kidney and sacrum was higher in the bipolar scheme compared to the monopolar (p < 0.03), whereas in muscle it was lower (p = 0.018). Image quality scores were higher for the bipolar sequence (p ≤ 0.025). CONCLUSION: Artefact reduction makes the bipolar DWI sequence preferable in abdominopelvic applications, although the trade-off in SNR may compromise ADC measurements in muscle.
The purpose was to determine the reproducibility of apparent diffusion coefficient (ADC) measurements in a two-centre phase I clinical trial; and to track ADC changes in response to the sequential administration of the vascular disrupting agent, combretastatin A4 phosphate (CA4P), and the anti-angiogenic drug, bevacizumab. Sixteen patients with solid tumours received CA4P and bevacizumab treatment. Echo-planar diffusion-weighted MRI was performed using six b values (b = 0-750 s/mm(2)) before (x2), and at 3 and 72 h after a first dose of CA4P. Bevacizumab was given 4 h after a second dose of CA4P, and imaging performed 3 h post CA4P and 72 h after bevacizumab treatment. The coefficient of repeatability (r) of ADC total (all b values), ADC high (b = 100-750) and ADC low (b = 0-100) was calculated by Bland-Altman analysis. The ADC total and ADC high showed good measurement reproducibility (r% = 13.3, 14.1). There was poor reproducibility of the perfusion-sensitive ADC low (r% = 62.5). Significant increases in the median ADC total and ADC high occurred at 3 h after the second dose of CA4P (p < 0.05). ADC measurements were highly reproducible in a two-centre clinical trial setting and appear promising for evaluating the effects of drugs that target tumour vasculature.
PURPOSE: To describe computed diffusion weighted (DW) magnetic resonance (MR) imaging as a method for obtaining high-b-value images from DW MR imaging performed at lower b values and to investigate the feasibility of the technique to improve lesion detection in oncologic cases. MATERIALS AND METHODS: The study was approved by the institutional and research committee, and written informed consent was obtained from all patients. DW MR imaging was performed on a CuSO(4) phantom at 1.5 T with a range of b values and compared with computed DW MR imaging images synthesized from lower b values (0 and 600 sec/mm(2)). The signal-to-noise ratio (SNR) was compared, and agreement between the SNR of computed DW MR imaging and theoretical estimation assessed. Computed DW MR imaging was evaluated in 10 oncologic patients who underwent whole-body DW MR imaging with b values of 0 and 900 sec/mm(2). Computed DW MR images at computed b values of 1500 and 2000 sec/mm(2) were generated. The image quality and background suppression of acquired and computed images were rated by a radiologist using a four-point scale. The diagnostic performance for malignant lesion detection using these images was evaluated and compared by using the McNemar Test. RESULTS: The SNR of computed DW MR imaging of the phantom conformed closely to theoretical predictions. Computed DW MR imaging resulted in a higher SNR compared with acquired DW MR imaging, especially at b values greater than 840 sec/mm(2). In patients, images with a computed b value of 2000 sec/mm(2) produced good image quality and high background suppression (mean scores of 2.8 and 4.0, respectively). Evaluation of images with a computed b value of 2000 sec/mm(2) resulted in higher overall diagnostic sensitivity (96.0%) and specificity (96.6%) compared with images with an acquired b value of 900 sec/mm(2) (sensitivity, 89.4%; specificity, 87.5%; P < .01). CONCLUSION: Computed DW MR imaging in the body allows higher-b-value images to be obtained with a good SNR. Clinical computed DW MR imaging is feasible and may improve disease detection. SUPPLEMENTAL MATERIAL: http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.11101919/-/DC1.
OBJECTIVE: To determine the measurement reproducibility of perfusion fraction f, pseudodiffusion coefficient D and diffusion coefficient D in colorectal liver metastases and normal liver. METHODS: Fourteen patients with known colorectal liver metastases were examined twice using respiratory-triggered echo-planar DW-MRI with eight b values (0 to 900 s/mm(2)) 1 h apart. Regions of interests were drawn around target metastasis and normal liver in each patient to derive ADC (all b values), ADC(high) (b values ≥ 100 s/mm(2)) and intravoxel incoherent motion (IVIM) parameters f, D and D by least squares data fitting. Short-term measurement reproducibility of median ADC, ADC(high), f, D and D values were derived from Bland-Altman analysis. RESULTS: The measurement reproducibility for ADC, ADC(high) and D was worst in colorectal liver metastases (-21 % to +25 %) compared with liver parenchyma (-6 % to +8 %). Poor measurement reproducibility was observed for the perfusion-sensitive parameters of f (-75 % to +241 %) and D (-89 % to +2,120 %) in metastases, and to a lesser extent the f (-24 % to +25 %) and D (-31 % to +59 %) of liver. CONCLUSIONS: Estimates of f and D derived from the widely used least squares IVIM fitting showed poor measurement reproducibility. Efforts should be made to improve the measurement reproducibility of perfusion-sensitive IVIM parameters.
OBJECTIVE: We examine the clinical impetus for whole-body diffusion-weighted MRI and discuss how to implement the technique with clinical MRI systems. We include practical tips and tricks to optimize image quality and reduce artifacts. The interpretative pitfalls are enumerated, and potential challenges are highlighted. CONCLUSION: Whole-body diffusion-weighted MRI can be used for tumor staging and assessment of treatment response. Meticulous technique and knowledge of potential interpretive pitfalls will help to avoid mistakes and establish this modality in radiologic practice.
OBJECTIVES: Silicone breast prostheses prove technically challenging when performing diffusion-weighted MR imaging in the breasts. We describe a combined fat and chemical suppression scheme to achieve dual suppression of fat and silicone, thereby improving the quality of diffusion-weighted images in women with breast implants. METHODS: MR imaging was performed at 3.0 and 1.5 T in women with silicone breast implants using short-tau inversion recovery (STIR) fat-suppressed echo-planar (EPI) diffusion-weighted MR imaging (DWI) on its own and combined with the slice-select gradient-reversal (SSGR) technique. Imaging was performed using dedicated breast imaging coils. RESULTS: Complete suppression of the fat and silicone signal was possible at 3.0 T using EPI DWI with STIR and SSGR, evaluated with dedicated breast coils. However, a residual silicone signal was still perceptible at 1.5 T using this combined approach. Nevertheless, a further reduction in silicone signal at 1.5 T could be achieved by employing thinner slice partitions and the addition of the chemical-selective fat-suppression (CHESS) technique. CONCLUSIONS: DWI using combined STIR and SSGR chemical suppression techniques is feasible to eliminate or reduce silicone signal from prosthetic breast implants. KEY POINTS: Breast magnetic resonance imaging (MRI) is frequently needed following breast implants. Unsuppressed signal from silicone creates artefacts on diffusion-weighted MR sequences. Dual fat/chemical suppression can eliminate signal from fat and silicone. STIR with slice selective gradient reversal can suppress fat and silicone signal.
We describe our semi-automatic segmentation of whole-body diffusion-weighted MRI (WBDWI) using a Markov random field (MRF) model to derive tumor total diffusion volume (tDV) and associated global apparent diffusion coefficient (gADC); and demonstrate the feasibility of using these indices for assessing tumor burden and response to treatment in patients with bone metastases. WBDWI was performed on eleven patients diagnosed with bone metastases from breast and prostate cancers before and after anti-cancer therapies. Semi-automatic segmentation incorporating a MRF model was performed in all patients below the C4 vertebra by an experienced radiologist with over eight years of clinical experience in body DWI. Changes in tDV and gADC distributions were compared with overall response determined by all imaging, tumor markers and clinical findings at serial follow up. The segmentation technique was possible in all patients although erroneous volumes of interest were generated in one patient because of poor fat suppression in the pelvis, requiring manual correction. Responding patients showed a larger increase in gADC (median change = +0.18, range = -0.07 to +0.78 × 10(-3) mm2/s) after treatment compared to non-responding patients (median change = -0.02, range = -0.10 to +0.05 × 10(-3) mm2/s, p = 0.05, Mann-Whitney test), whereas non-responding patients showed a significantly larger increase in tDV (median change = +26%, range = +3 to +284%) compared to responding patients (median change = -50%, range = -85 to +27%, p = 0.02, Mann-Whitney test). Semi-automatic segmentation of WBDWI is feasible for metastatic bone disease in this pilot cohort of 11 patients, and could be used to quantify tumor total diffusion volume and median global ADC for assessing response to treatment.
Malignant pleural mesothelioma (MPM) typically demonstrates a non-spherical growth pattern, so it is often difficult to accurately categorize change in tumour burden using size-based tumour response criteria (e.g., WHO (World Health Organisation), RECIST (Response Evaluation Criteria in Solid Tumours) and modified RECIST). Functional imaging techniques are applied to derive quantitative measurements of tumours, which reflect particular aspects of the tumour pathophysiology. By quantifying how these measurements change with treatment, it is possible to observe treatment effects. In this review, we survey the existing roles of CT and MRI for the management of MPM, including the currently applied size measurement criteria for the assessment of treatment response. New functional imaging techniques, such as positron emission tomography (PET), diffusion-weighted MRI (DWI) and dynamic contrast-enhanced MRI (DCE-MRI) that may potentially improve the assessment of treatment response will be highlighted and discussed.
We present pyOsiriX, a plugin built for the already popular dicom viewer OsiriX that provides users the ability to extend the functionality of OsiriX through simple Python scripts. This approach allows users to integrate the many cutting-edge scientific/image-processing libraries created for Python into a powerful DICOM visualisation package that is intuitive to use and already familiar to many clinical researchers. Using pyOsiriX we hope to bridge the apparent gap between basic imaging scientists and clinical practice in a research setting and thus accelerate the development of advanced clinical image processing. We provide arguments for the use of Python as a robust scripting language for incorporation into larger software solutions, outline the structure of pyOsiriX and how it may be used to extend the functionality of OsiriX, and we provide three case studies that exemplify its utility. For our first case study we use pyOsiriX to provide a tool for smooth histogram display of voxel values within a user-defined region of interest (ROI) in OsiriX. We used a kernel density estimation (KDE) method available in Python using the scikit-learn library, where the total number of lines of Python code required to generate this tool was 22. Our second example presents a scheme for segmentation of the skeleton from CT datasets. We have demonstrated that good segmentation can be achieved for two example CT studies by using a combination of Python libraries including scikit-learn, scikit-image, SimpleITK and matplotlib. Furthermore, this segmentation method was incorporated into an automatic analysis of quantitative PET-CT in a patient with bone metastases from primary prostate cancer. This enabled repeatable statistical evaluation of PET uptake values for each lesion, before and after treatment, providing estaimes maximum and median standardised uptake values (SUVmax and SUVmed respectively). Following treatment we observed a reduction in lesion volume, SUVmax and SUVmed for all lesions, in agreement with a reduction in concurrent measures of serum prostate-specific antigen (PSA).
<h4>Purpose</h4>To evaluate the diagnostic sensitivity of computed diffusion-weighted (DW)-MR imaging for the detection of breast cancer.<h4>Materials and methods</h4>Local research ethics approval was obtained. A total of 61 women (median 48 years) underwent dynamic contrast enhanced (DCE)- and DW-MR between January 2011 and March 2012, including 27 with breast cancer on core biopsy and 34 normal cases. Standard ADC maps using all four b values (0, 350, 700, 1150) were used to generate computed DW-MR images at b = 1500 s/mm(2) and b = 2000 s/mm(2) . Four image sets were read sequentially by two readers: acquired b = 1150 s/mm(2) , computed b = 1500 s/mm(2) and b = 2000 s/mm(2) , and DCE-MR at an early time point. Cancer detection was rated using a five-point scale; image quality and background suppression were rated using a four-point scale. The diagnostic sensitivity for breast cancer detection was compared using the McNemar test and inter-reader agreement with a Kappa value.<h4>Results</h4>Computed DW-MR resulted in higher overall diagnostic sensitivity with b = 2000 s/mm(2) having a mean diagnostic sensitivity of 76% (range 49.8-93.7%) and b = 1500 s/mm(2) having a mean diagnostic sensitivity of 70.3% (range 32-97.7%) compared with 44.4% (range 25.5-64.7%) for acquired b = 1150 s/mm(2) (both p = 0.0001). Computed DW-MR images produced better image quality and background suppression (mean scores for both readers: 2.55 and 2.9 for b 1500 s/mm(2) ; 2.55 and 3.15 for b 2000 s/mm(2) , respectively) than the acquired b value 1150 s/mm(2) images (mean scores for both readers: 2.4 and 2.45, respectively).<h4>Conclusion</h4>Computed DW-MR imaging has the potential to improve the diagnostic sensitivity of breast cancer detection compared to acquired DW-MR. J. Magn. Reson. Imaging 2016;44:130-137.
Purpose To determine the correlation between the volume of bone metastasis as assessed with diffusion-weighted (DW) imaging and established prognostic factors in metastatic castration-resistant prostate cancer (mCRPC) and the association with overall survival (OS). Materials and Methods This retrospective study was approved by the institutional review board; informed consent was obtained from all patients. The authors analyzed whole-body DW images obtained between June 2010 and February 2013 in 53 patients with mCRPC at the time of starting a new line of anticancer therapy. Bone metastases were identified and delineated on whole-body DW images in 43 eligible patients. Total tumor diffusion volume (tDV) was correlated with the bone scan index (BSI) and other prognostic factors by using the Pearson correlation coefficient (r). Survival analysis was performed with Kaplan-Meier analysis and Cox regression. Results The median tDV was 503.1 mL (range, 5.6-2242 mL), and the median OS was 12.9 months (95% confidence interval [CI]: 8.7, 16.1 months). There was a significant correlation between tDV and established prognostic factors, including hemoglobin level (r = -0.521, P < .001), prostate-specific antigen level (r = 0.556, P < .001), lactate dehydrogenase level (r = 0.534, P < .001), alkaline phosphatase level (r = 0.572, P < .001), circulating tumor cell count (r = 0.613, P = .004), and BSI (r = 0.565, P = .001). A higher tDV also showed a significant association with poorer OS (hazard ratio, 1.74; 95% CI: 1.02, 2.96; P = .035). Conclusion Metastatic bone disease from mCRPC can be evaluated and quantified with whole-body DW imaging. Whole-body DW imaging-generated tDV showed correlation with established prognostic biomarkers and is associated with OS in mCRPC. (©) RSNA, 2016 Online supplemental material is available for this article.
Quantitative whole-body diffusion-weighted MRI (WB-DWI) is now possible using semi-automatic segmentation techniques. The method enables whole-body estimates of global Apparent Diffusion Coefficient (gADC) and total Diffusion Volume (tDV), both of which have demonstrated considerable utility for assessing treatment response in patients with bone metastases from primary prostate and breast cancers. Here we investigate the agreement (inter-observer repeatability) between two radiologists in their definition of Volumes Of Interest (VOIs) and subsequent assessment of tDV and gADC on an exploratory patient cohort of nine. Furthermore, each radiologist was asked to repeat his or her measurements on the same patient data sets one month later to identify the intra-observer repeatability of the technique. Using a Markov Chain Monte Carlo (MCMC) estimation method provided full posterior probabilities of repeatability measures along with maximum a-posteriori values and 95% confidence intervals. Our estimates of the inter-observer Intraclass Correlation Coefficient (ICCinter) for log-tDV and median gADC were 1.00 (0.97-1.00) and 0.99 (0.89-0.99) respectively, indicating excellent observer agreement for these metrics. Mean gADC values were found to have ICCinter = 0.97 (0.81-0.99) indicating a slight sensitivity to outliers in the derived distributions of gADC. Of the higher order gADC statistics, skewness was demonstrated to have good inter-user agreement with ICCinter = 0.99 (0.86-1.00), whereas gADC variance and kurtosis performed relatively poorly: 0.89 (0.39-0.97) and 0.96 (0.69-0.99) respectively. Estimates of intra-observer repeatability (ICCintra) demonstrated similar results: 0.99 (0.95-1.00) for log-tDV, 0.98 (0.89-0.99) and 0.97 (0.83-0.99) for median and mean gADC respectively, 0.64 (0.25-0.88) for gADC variance, 0.85 (0.57-0.95) for gADC skewness and 0.85 (0.57-0.95) for gADC kurtosis. Further investigation of two anomalous patient cases revealed that a very small proportion of voxels with outlying gADC values lead to instability in higher order gADC statistics. We therefore conclude that estimates of median/mean gADC and tumour volume demonstrate excellent inter- and intra-observer repeatability whilst higher order statistics of gADC should be used with caution when ascribing significance to clinical changes.
Whole-body diffusion-weighted (DW) imaging is a recent development. The image contrast is based on differences in mobility of water between tissues and reflects tissue cellularity and integrity of cell membranes. The tissue water diffusivity is quantified by the apparent diffusion coefficient. By performing imaging at multiple imaging stations, whole-body DW imaging has been applied to improve tumor staging, disease characterization, as well as for the assessment of treatment response. Information from DW imaging studies could be combined with those using PET imaging tracers to further refine and improve the assessment of patients with cancer.
Purpose To determine the usefulness of whole-body diffusion-weighted imaging (DWI) to assess the response of bone metastases to treatment in patients with metastatic castration-resistant prostate cancer (mCRPC). Materials and Methods A phase II prospective clinical trial of the poly-(adenosine diphosphate-ribose) polymerase inhibitor olaparib in mCRPC included a prospective magnetic resonance (MR) imaging substudy; the study was approved by the institutional research board, and written informed consent was obtained. Whole-body DWI was performed at baseline and after 12 weeks of olaparib administration by using 1.5-T MR imaging. Areas of abnormal signal intensity on DWI images in keeping with bone metastases were delineated to derive total diffusion volume (tDV); five target lesions were also evaluated. Associations of changes in volume of bone metastases and median apparent diffusion coefficient (ADC) with response to treatment were assessed by using the Mann-Whitney test and logistic regression; correlation with prostate-specific antigen level and circulating tumor cell count were assessed by using Spearman correlation (r). Results Twenty-one patients were included. All six responders to olaparib showed a decrease in tDV, while no decrease was observed in all nonresponders; this difference between responders and nonresponders was significant (P = .001). Increases in median ADC were associated with increased odds of response (odds ratio, 1.08; 95% confidence interval [CI]: 1.00, 1.15; P = .04). A positive association was detected between changes in tDV and best percentage change in prostate-specific antigen level and circulating tumor cell count (r = 0.63 [95% CI: 0.27, 0.83] and r = 0.77 [95% CI: 0.51, 0.90], respectively). When assessing five target lesions, decreases in volume were associated with response (odds ratio for volume increase, 0.89; 95% CI: 0.80, 0.99; P = .037). Conclusion This pilot study showed that decreases in volume and increases in median ADC of bone metastases assessed with whole-body DWI can potentially be used as indicators of response to olaparib in mCRPC. Online supplemental material is available for this article.
<h4>Purpose</h4>To introduce T<sub>2</sub>-adjusted computed DWI (T<sub>2</sub>-cDWI), a method that provides synthetic images at arbitrary b-values and echo times (TEs) that improve tissue contrast by removing or increasing T<sub>2</sub> contrast in diffusion-weighted images.<h4>Materials and methods</h4>In addition to the standard DWI acquisition protocol T<sub>2</sub>-weighted echo-planar images at multiple (≥2) echo times were acquired. This allows voxelwise estimation of apparent diffusion coefficient (ADC) and T<sub>2</sub> values, permitting synthetic images to be generated at any chosen b-value and echo time. An analytical model is derived for the noise properties in T<sub>2</sub>-cDWI, and validated using a diffusion test-object. Furthermore, we present T<sub>2</sub>-cDWI in two example clinical case studies: (i) a patient with mesothelioma demonstrating multiple disease tissue compartments and (ii) a patient with primary ovarian cancer demonstrating solid and cystic disease compartments.<h4>Results</h4>Measured image noise in T<sub>2</sub>-cDWI from phantom experiments conformed to the analytical model and demonstrated that T<sub>2</sub>-cDWI at high computed b-value/TE combinations achieves lower noise compared with conventional DWI. In patients, T<sub>2</sub>-cDWI with low b-value and long TE enhanced fluid signal while suppressing solid tumour components. Conversely, large b-values and short TEs overcome T<sub>2</sub> shine-through effects and increase the contrast between tumour and fluid compared with conventional high-b-value DW images.<h4>Conclusion</h4>T<sub>2</sub>-cDWI is a promising clinical tool for improving image signal-to-noise, image contrast, and tumour detection through suppression of T<sub>2</sub> shine-through effects.
Purpose To assess the repeatability of apparent diffusion coefficient (ADC) estimates in extracranial soft-tissue diffusion-weighted magnetic resonance imaging across a wide range of imaging protocols and patient populations. Materials and Methods Nine prospective patient studies and one prospective volunteer study, performed between 2006 and 2016 with research ethics committee approval and written informed consent from each subject, were included in this single-institution study. A total of 141 tumors and healthy organs were imaged twice (interval between repeated examinations, 45 minutes to 10 days, depending the on study) to assess the repeatability of median and mean ADC estimates. The Levene test was used to determine whether ADC repeatability differed between studies. The Pearson linear correlation coefficient was used to assess correlation between coefficient of variation (CoV) and the year the study started, study size, and volumes of tumors and healthy organs. The repeatability of ADC estimates from small, medium, and large tumors and healthy organs was assessed irrespective of study, and the Levene test was used to determine whether ADC repeatability differed between these groups. Results CoV aggregated across all studies was 4.1% (range for each study, 1.7%-6.5%). No correlation was observed between CoV and the year the study started or study size. CoV was weakly correlated with volume (r = -0.5, P = .1). Repeatability was significantly different between small, medium, and large tumors (P < .05), with the lowest CoV (2.6%) for large tumors. There was a significant difference in repeatability between studies-a difference that did not persist after the study with the largest tumors was excluded. Conclusion ADC is a robust imaging metric with excellent repeatability in extracranial soft tissues across a wide range of tumor sites, sizes, patient populations, and imaging protocol variations. Online supplemental material is available for this article.
<h4>Purpose</h4>To determine the test-retest repeatability of Apparent Diffusion Coefficient (ADC) measurements across institutions and MRI vendors, plus investigate the effect of post-processing methodology on measurement precision.<h4>Methods</h4>Thirty malignant lung lesions >2 cm in size (23 patients) were scanned on two occasions, using echo-planar-Diffusion-Weighted (DW)-MRI to derive whole-tumour ADC (b = 100, 500 and 800smm<sup>-2</sup>). Scanning was performed at 4 institutions (3 MRI vendors). Whole-tumour volumes-of-interest were copied from first visit onto second visit images and from one post-processing platform to an open-source platform, to assess ADC repeatability and cross-platform reproducibility.<h4>Results</h4>Whole-tumour ADC values ranged from 0.66-1.94x10<sup>-3</sup>mm<sup>2</sup>s<sup>-1</sup> (mean = 1.14). Within-patient coefficient-of-variation (wCV) was 7.1% (95% CI 5.7-9.6%), limits-of-agreement (LoA) -18.0 to 21.9%. Lesions >3 cm had improved repeatability: wCV 3.9% (95% CI 2.9-5.9%); and LoA -10.2 to 11.4%. Variability for lesions <3 cm was 2.46 times higher. ADC reproducibility across different post-processing platforms was excellent: Pearson's R<sup>2</sup> = 0.99; CoV 2.8% (95% CI 2.3-3.4%); and LoA -7.4 to 8.0%.<h4>Conclusion</h4>A free-breathing DW-MRI protocol for imaging malignant lung tumours achieved satisfactory within-patient repeatability and was robust to changes in post-processing software, justifying its use in multi-centre trials. For response evaluation in individual patients, a change in ADC >21.9% will reflect treatment-related change.<h4>Key points</h4>• In lung cancer, free-breathing DWI-MRI produces acceptable images with evaluable ADC measurement. • ADC repeatability coefficient-of-variation is 7.1% for lung tumours >2 cm. • ADC repeatability coefficient-of-variation is 3.9% for lung tumours >3 cm. • ADC measurement precision is unaffected by the post-processing software used. • In multicentre trials, 22% increase in ADC indicates positive treatment response.
<h4>Objective</h4>The purpose of this article is to review current image acquisition and interpretation for whole-body MRI, clinical applications, and the emerging roles in oncologic imaging, especially in the assessment of bone marrow diseases.<h4>Conclusion</h4>Whole-body MRI is an emerging technique used for early diagnosis, staging, and assessment of therapeutic response in oncology. The improved accessibility and advances in technology, including widely available sequences (Dixon and DWI), have accelerated its deployment and acceptance in clinical practice.
<h4>Objective</h4>Application of whole body diffusion-weighted MRI (WB-DWI) for oncology are rapidly increasing within both research and routine clinical domains. However, WB-DWI as a quantitative imaging biomarker (QIB) has significantly slower adoption. To date, challenges relating to accuracy and reproducibility, essential criteria for a good QIB, have limited widespread clinical translation. In recognition, a UK workgroup was established in 2016 to provide technical consensus guidelines (to maximise accuracy and reproducibility of WB-MRI QIBs) and accelerate the clinical translation of quantitative WB-DWI applications for oncology.<h4>Methods</h4>A panel of experts convened from cancer centres around the UK with subspecialty expertise in quantitative imaging and/or the use of WB-MRI with DWI. A formal consensus method was used to obtain consensus agreement regarding best practice. Questions were asked about the appropriateness or otherwise on scanner hardware and software, sequence optimisation, acquisition protocols, reporting, and ongoing quality control programs to monitor precision and accuracy and agreement on quality control.<h4>Results</h4>The consensus panel was able to reach consensus on 73% (255/351) items and based on consensus areas made recommendations to maximise accuracy and reproducibly of quantitative WB-DWI studies performed at 1.5T. The panel were unable to reach consensus on the majority of items related to quantitative WB-DWI performed at 3T.<h4>Conclusion</h4>This UK Quantitative WB-DWI Technical Workgroup consensus provides guidance on maximising accuracy and reproducibly of quantitative WB-DWI for oncology. The consensus guidance can be used by researchers and clinicians to harmonise WB-DWI protocols which will accelerate clinical translation of WB-DWI-derived QIBs.
<h4>Objectives</h4>The aim of this study was to identify apparent diffusion coefficient (ADC) values for typical haemangiomas in the spine and to compare them with active malignant focal deposits.<h4>Methods</h4>This was a retrospective single-institution study. Whole-body magnetic resonance imaging (MRI) scans of 106 successive patients with active multiple myeloma, metastatic prostate or breast cancer were analysed. ADC values of typical vertebral haemangiomas and malignant focal deposits were recorded.<h4>Results</h4>The ADC of haemangiomas (72 ROIs, median ADC 1,085×10<sup>-6</sup>mm<sup>2</sup>s<sup>-1</sup>, interquartile range 927-1,295×10<sup>-6</sup>mm<sup>2</sup>s<sup>-1</sup>) was significantly higher than the ADC of malignant focal deposits (97 ROIs, median ADC 682×10<sup>-6</sup>mm<sup>2</sup>s<sup>-1</sup>, interquartile range 583-781×10<sup>-6</sup>mm<sup>2</sup>s<sup>-1</sup>) with a p-value < 10<sup>-6</sup>. Receiver operating characteristic (ROC) analysis produced an area under the curve of 0.93. An ADC threshold of 872×10<sup>-6</sup>mm<sup>2</sup>s<sup>-1</sup> separated haemangiomas from malignant focal deposits with a sensitivity of 84.7 % and specificity of 91.8 %.<h4>Conclusions</h4>ADC values of classical vertebral haemangiomas are significantly higher than malignant focal deposits. The high ADC of vertebral haemangiomas allows them to be distinguished visually and quantitatively from active sites of disease, which show restricted diffusion.<h4>Key points</h4>• Whole-body diffusion-weighted MRI is becoming widely used in myeloma and bone metastases. • ADC values of vertebral haemangiomas are significantly higher than malignant focal deposits. • High ADCs of haemangiomas allows them to be distinguished from active disease.
Diffusion-weighted magnetic resonance imaging (DWI) enables non-invasive, quantitative staging of prostate cancer via measurement of the apparent diffusion coefficient (ADC) of water within tissues. In cancer, more advanced disease is often characterized by higher cellular density (cellularity), which is generally accepted to correspond to a lower measured ADC. A quantitative relationship between tissue structure and <i>in vivo</i> measurements of ADC has yet to be determined for prostate cancer. In this study, we establish a theoretical framework for relating ADC measurements with tissue cellularity and the proportion of space occupied by prostate lumina, both of which are estimated through automatic image processing of whole-slide digital histology samples taken from a cohort of six healthy mice and nine transgenic adenocarcinoma of the mouse prostate (TRAMP) mice. We demonstrate that a significant inverse relationship exists between ADC and tissue cellularity that is well characterized by our model, and that a decrease of the luminal space within the prostate is associated with a decrease in ADC and more aggressive tumor subtype. The parameters estimated from our model in this mouse cohort predict the diffusion coefficient of water within the prostate-tissue to be 2.18 × 10<sup>-3</sup> mm<sup>2</sup>/s (95% CI: 1.90, 2.55). This value is significantly lower than the diffusion coefficient of free water at body temperature suggesting that the presence of organelles and macromolecules within tissues can drastically hinder the random motion of water molecules within prostate tissue. We validate the assumptions made by our model using novel <i>in silico</i> analysis of whole-slide histology to provide the simulated ADC (sADC); this is demonstrated to have a significant positive correlation with <i>in vivo</i> measured ADC (r<sup>2</sup> = 0.55) in our mouse population. The estimation of the structural properties of prostate tissue is vital for predicting and staging cancer aggressiveness, but prostate tissue biopsies are painful, invasive, and are prone to complications such as sepsis. The developments made in this study provide the possibility of estimating the structural properties of prostate tissue via non-invasive virtual biopsies from MRI, minimizing the need for multiple tissue biopsies and allowing sequential measurements to be made for prostate cancer monitoring.
Whole-body diffusion-weighted MRI has emerged as a powerful diagnostic tool for disease detection and staging mainly used in systemic bone disease. The large field-of-view functional imaging technique highlights cellular tumor and suppresses normal tissue signal, allowing quantification of an estimate of total disease burden, summarized as the total diffusion volume (tDV), as well as global apparent diffusion coefficient (gADC) measurements. Both tDV and gADC have been shown to be repeatable quantitative parameters that indicate tumor heterogeneity and treatment effects, thus potential, noninvasive, imaging biomarkers informing on disease prognosis and therapy response.
Acknowledging the increasingly important role of whole-body MRI for directing patient care in myeloma, a multidisciplinary, international, and expert panel of radiologists, medical physicists, and hematologists with specific expertise in whole-body MRI in myeloma convened to discuss the technical performance standards, merits, and limitations of currently available imaging methods. Following guidance from the International Myeloma Working Group and the National Institute for Clinical Excellence in the United Kingdom, the Myeloma Response Assessment and Diagnosis System (or MY-RADS) imaging recommendations are designed to promote standardization and diminish variations in the acquisition, interpretation, and reporting of whole-body MRI in myeloma and allow response assessment. This consensus proposes a core clinical protocol for whole-body MRI and an extended protocol for advanced assessments. Published under a CC BY 4.0 license. Online supplemental material is available for this article.
Childhood neuroblastoma is a hypervascular tumor of neural origin, for which antiangiogenic drugs are currently being evaluated; however, predictive biomarkers of treatment response, crucial for successful delivery of precision therapeutics, are lacking. We describe an MRI-pathologic cross-correlative approach using intrinsic susceptibility (IS) and susceptibility contrast (SC) MRI to noninvasively map the vascular phenotype in neuroblastoma Th-MYCN transgenic mice treated with the vascular endothelial growth factor receptor inhibitor cediranib. We showed that the transverse MRI relaxation rate <i>R</i> <sub>2</sub>* (second<sup>-1</sup>) and fractional blood volume (<i>f</i>BV, %) were sensitive imaging biomarkers of hemorrhage and vascular density, respectively, and were also predictive biomarkers of response to cediranib. Comparison with MRI and pathology from patients with MYCN-amplified neuroblastoma confirmed the high degree to which the Th-MYCN model vascular phenotype recapitulated that of the clinical phenotype, thereby supporting further evaluation of IS- and SC-MRI in the clinic. This study reinforces the potential role of functional MRI in delivering precision medicine to children with neuroblastoma. SIGNIFICANCE: This study shows that functional MRI predicts response to vascular-targeted therapy in a genetically engineered murine model of neuroblastoma.
<b>Background:</b> Multi-parametric MRI provides non-invasive methods for response assessment of soft-tissue sarcoma (STS) from non-surgical treatments. However, evaluation of MRI parameters over the whole tumor volume may not reveal the full extent of post-treatment changes as STS tumors are often highly heterogeneous, including cellular tumor, fat, necrosis, and cystic tissue compartments. In this pilot study, we investigate the use of machine-learning approaches to automatically delineate tissue compartments in STS, and use this approach to monitor post-radiotherapy changes. <b>Methods:</b> Eighteen patients with retroperitoneal sarcoma were imaged using multi-parametric MRI; 8/18 received a follow-up imaging study 2-4 weeks after pre-operative radiotherapy. Eight commonly-used supervised machine-learning techniques were optimized for classifying pixels into one of five tissue sub-types using an exhaustive cross-validation approach and expert-defined regions of interest as a gold standard. Final pixel classification was smoothed using a Markov Random Field (MRF) prior distribution on the final machine-learning models. <b>Findings:</b> 5/8 machine-learning techniques demonstrated high median cross-validation accuracies (82.2%, range 80.5-82.5%) with no significant difference between these five methods. One technique was selected (Naïve-Bayes) due to its relatively short training and class-prediction times (median 0.73 and 0.69 ms, respectively on a 3.5 GHz personal machine). When combined with the MRF-prior, this approach was successfully applied in all eight post-radiotherapy imaging studies and provided visualization and quantification of changes to independent STS sub-regions following radiotherapy for heterogeneous response assessment. <b>Interpretation:</b> Supervised machine-learning approaches to tissue classification in multi-parametric MRI of soft-tissue sarcomas provide quantitative evaluation of heterogeneous tissue changes following radiotherapy.
<b>Purpose:</b> To characterize the voxel-wise uncertainties of Apparent Diffusion Coefficient (ADC) estimation from whole-body diffusion-weighted imaging (WBDWI). This enables the calculation of a new parametric map based on estimates of ADC and ADC uncertainty to improve WBDWI imaging standardization and interpretation: NoIse-Corrected Exponentially-weighted diffusion-weighted MRI (niceDWI). <b>Methods:</b> Three approaches to the joint modeling of voxel-wise ADC and ADC uncertainty (σ<sub>ADC</sub>) are evaluated: (i) direct weighted least squares (DWLS), (ii) iterative linear-weighted least-squares (IWLS), and (iii) smoothed IWLS (SIWLS). The statistical properties of these approaches in terms of ADC/σ<sub>ADC</sub> accuracy and precision is compared using Monte Carlo simulations. Our proposed post-processing methodology (niceDWI) is evaluated using an ice-water phantom, by comparing the contrast-to-noise ratio (CNR) with conventional exponentially-weighted DWI. We present the clinical feasibility of niceDWI in a pilot cohort of 16 patients with metastatic prostate cancer. <b>Results:</b> The statistical properties of ADC and σ<sub>ADC</sub> conformed closely to the theoretical predictions for DWLS, IWLS, and SIWLS fitting routines (a minor bias in parameter estimation is observed with DWLS). Ice-water phantom experiments demonstrated that a range of CNR could be generated using the niceDWI approach, and could improve CNR compared to conventional methods. We successfully implemented the niceDWI technique in our patient cohort, which visually improved the in-plane bias field compared with conventional WBDWI. <b>Conclusions:</b> Measurement of the statistical uncertainty in ADC estimation provides a practical way to standardize WBDWI across different scanners, by providing quantitative image signals that improve its reliability. Our proposed method can overcome inter-scanner and intra-scanner WBDWI signal variations that can confound image interpretation.
Advances in hardware and software enable high-quality body diffusion-weighted images to be acquired for oncologic assessment. 3.0 T affords improved signal/noise for higher spatial resolution and smaller field-of-view diffusion-weighted imaging (DWI). DWI at 3.0 T can be applied as at 1.5 T to improve tumor detection, disease characterization, and the assessment of treatment response. DWI at 3.0 T can be acquired on a hybrid PET-MR imaging system, to allow functional MR information to be combined with molecular imaging.
Whole-body MRI (WB-MRI) has evolved since its first introduction in the 1970s as an imaging technique to detect and survey disease across multiple sites and organ systems in the body. The development of diffusion-weighted MRI (DWI) has added a new dimension to the implementation of WB-MRI on modern scanners, offering excellent lesion-to-background contrast, while achieving acceptable spatial resolution to detect focal lesions 5 to 10 mm in size. MRI hardware and software advances have reduced acquisition times, with studies taking 40-50 min to complete.The rising awareness of medical radiation exposure coupled with the advantages of MRI has resulted in increased utilization of WB-MRI in oncology, paediatrics, rheumatological and musculoskeletal conditions and more recently in population screening. There is recognition that WB-MRI can be used to track disease evolution and monitor response heterogeneity in patients with cancer. There are also opportunities to combine WB-MRI with molecular imaging on PET-MRI systems to harness the strengths of hybrid imaging. The advent of artificial intelligence and machine learning will shorten image acquisition times and image analyses, making the technique more competitive against other imaging technologies.
Noninvasive early indicators of treatment response are crucial to the successful delivery of precision medicine in children with cancer. Neuroblastoma is a common solid tumor of young children that arises from anomalies in neural crest development. Therapeutic approaches aiming to destabilize <i>MYCN</i> protein, such as small-molecule inhibitors of Aurora A and mTOR, are currently being evaluated in early phase clinical trials in children with high-risk <i>MYCN</i>-driven disease, with limited ability to evaluate conventional pharmacodynamic biomarkers of response. T<sub>1</sub> mapping is an MRI scan that measures the proton spin-lattice relaxation time T<sub>1</sub>. Using a multiparametric MRI-pathologic cross-correlative approach and computational pathology methodologies including a machine learning-based algorithm for the automatic detection and classification of neuroblasts, we show here that T<sub>1</sub> mapping is sensitive to the rich histopathologic heterogeneity of neuroblastoma in the Th-<i>MYCN</i> transgenic model. Regions with high native T<sub>1</sub> corresponded to regions dense in proliferative undifferentiated neuroblasts, whereas regions characterized by low T<sub>1</sub> were rich in apoptotic or differentiating neuroblasts. Reductions in tumor-native T<sub>1</sub> represented a sensitive biomarker of response to treatment-induced apoptosis with two <i>MYCN</i>-targeted small-molecule inhibitors, Aurora A kinase inhibitor alisertib (MLN8237) and mTOR inhibitor vistusertib (AZD2014). Overall, we demonstrate the potential of T<sub>1</sub> mapping, a scan readily available on most clinical MRI scanners, to assess response to therapy and guide clinical trials for children with neuroblastoma. The study reinforces the potential role of MRI-based functional imaging in delivering precision medicine to children with neuroblastoma. SIGNIFICANCE: This study shows that MRI-based functional imaging can detect apoptotic responses to <i>MYCN</i>-targeted small-molecule inhibitors in a genetically engineered murine model of <i>MYCN</i>-driven neuroblastoma.
<h4>Background</h4>Whole body magnetic resonance imaging (MRI) is now incorporated into international guidance for imaging patients with multiple myeloma. The aim of this study was to investigate inter-observer agreement of triple reported baseline whole-body MRI in myeloma and highlight potential pitfalls.<h4>Methods</h4>Fifty-seven patients with symptomatic myeloma at first presentation or relapse and planned for autologous stem cell transplant were included. All patients completed baseline whole body MRI within 2 weeks prior to starting treatment. Each scan was reported independently by 3 radiologists using a defined scoring system. Differences in observer scores were compared using analysis of variance (ANOVA) and inter-observer agreement assessed using intra class correlation coefficient (ICC).<h4>Results</h4>There was no significant difference in mean observer scores for whole skeleton and ICC demonstrated excellent inter-observer agreement at 0.91. ICC varied between skeletal regions with spine, pelvis and ribs showing good inter-observer agreement, whereas skull and long bones were moderate. Scans with variation in observer scores were re-examined and cause of discrepancies identified. This information was used to describe potential anatomical pitfalls in reporting .<h4>Conclusion</h4>Whole-body MRI has excellent inter-observer agreement in reporting symptomatic myeloma at baseline. Inter-observer agreement varied between skeletal regions highlighting specific areas of difficulty.
<h4>Purpose</h4>To evaluate robustness and repeatability of magnetic resonance imaging (MRI) texture features in water and tissue phantom test-retest study.<h4>Materials and methods</h4>Separate water and tissue phantoms were imaged twice with the same protocol in a test-retest experiment using a 1.5-T scanner. Protocols were acquired to favour signal-to-noise ratio and resolution. Forty-six features including first order statistics and second-order texture features were extracted, and repeatability was assessed by calculating the concordance correlation coefficient. Separately, base image noise and resolution were manipulated in an in silico experiment, and robustness of features was calculated by assessing percentage coefficient of variation and linear correlation of features with noise and resolution. These simulation data were compared with the acquired data. Features were classified by their degree (high, intermediate, or low) of robustness and repeatability.<h4>Results</h4>Eighty percent of the MRI features were repeatable (concordance correlation coefficient > 0.9) in the phantom test-retest experiment. The majority (approximately 90%) demonstrated a strong or intermediate correlation with image acquisition parameter, and 19/46 (41%) and 13/46 (28%) of features were highly robust to noise and resolution, respectively (coefficient of variation < 5%). Agreement between the acquired and simulation data varied, with the range of agreement within feature classes between 11 and 92%.<h4>Conclusion</h4>Most MRI features were repeatable in a phantom test-retest study. This phantom data may serve as a lower limit of feature MRI repeatability. Robustness of features varies with acquisition parameter, and appropriate features can be selected for clinical validation studies.
<h4>Objective</h4>To assess intra- and inter-reader variability of apparent diffusion coefficient (ADC) and fat fraction (FF) measurement in focal myeloma bone lesions and the influence of lesion size.<h4>Methods</h4>22 myeloma patients with focal active disease on whole body MRI were included. Two readers outlined a small (5-10 mm) and large lesion (>10 mm) in each subject on derived ADC and FF maps; one reader performed this twice. Intra- and inter-reader agreement for small and large lesion groups were calculated for derived statistics from each map using within-subject standard deviation, coefficient of variation, interclass correlation coefficient measures, and visualized with Bland-Altman plots.<h4>Results</h4>For mean ADC, intra- and inter-reader repeatability demonstrated equivalently low coefficient of variation (3.0-3.6%) and excellent interclass correlation coefficient (0.975-0.982) for both small and large lesions. For mean FF, intra- and inter-reader repeatability was significantly poorer for small lesions compared to large lesions (intra-reader within-subject standard variation estimate is 2.7 times higher for small lesions than large lesions (<i>p</i> = 0.0071), and for inter-reader variations is 3.8 times higher (<i>p</i> = 0.0070)).<h4>Conclusion</h4>There is excellent intra- and inter-reader agreement for mean ADC estimates, even for lesions as small as 5 mm. For FF measurements, there is a significant increase in coefficient of variation for smaller lesions, suggesting lesions >10 mm should be selected for lesion FF measurement.<h4>Advances in knowledge</h4>ADC measurements of focal myeloma have excellent intra- and inter-reader agreement. FF measurements are more susceptible to lesion size as intra- and inter-reader agreement is significantly impaired in lesions less than 10 mm.
<h4>Purpose</h4>To use deep learning to improve the image quality of subsampled images (number of acquisitions = 1 [NOA<sub>1</sub>]) to reduce whole-body diffusion-weighted MRI (WBDWI) acquisition times.<h4>Materials and methods</h4>Both retrospective and prospective patient groups were used to develop a deep learning-based denoising image filter (DNIF) model. For initial model training and validation, 17 patients with metastatic prostate cancer with acquired WBDWI NOA<sub>1</sub> and NOA<sub>9</sub> images (acquisition period, 2015-2017) were retrospectively included. An additional 22 prospective patients with advanced prostate cancer, myeloma, and advanced breast cancer were used for model testing (2019), and the radiologic quality of DNIF-processed NOA<sub>1</sub> (NOA<sub>1-DNIF</sub>) images were compared with NOA<sub>1</sub> images and clinical NOA<sub>16</sub> images by using a three-point Likert scale (good, average, or poor; statistical significance was calculated by using a Wilcoxon signed ranked test). The model was also retrained and tested in 28 patients with malignant pleural mesothelioma (MPM) who underwent lung MRI (2015-2017) to demonstrate feasibility in other body regions.<h4>Results</h4>The model visually improved the quality of NOA<sub>1</sub> images in all test patients, with the majority of NOA<sub>1-DNIF</sub> and NOA<sub>16</sub> images being graded as either "average" or "good" across all image-quality criteria. From validation data, the mean apparent diffusion coefficient (ADC) values within NOA<sub>1-DNIF</sub> images of bone disease deviated from those within NOA<sub>9</sub> images by an average of 1.9% (range, 1.1%-2.6%). The model was also successfully applied in the context of MPM; the mean ADCs from NOA<sub>1-DNIF</sub> images of MPM deviated from those measured by using clinical-standard images (NOA<sub>12</sub>) by 3.7% (range, 0.2%-10.6%).<h4>Conclusion</h4>Clinical-standard images were generated from subsampled images by using a DNIF.<b>Keywords:</b> Image Postprocessing, MR-Diffusion-weighted Imaging, Neural Networks, Oncology, Whole-Body Imaging, Supervised Learning, MR-Functional Imaging, Metastases, Prostate, Lung <i>Supplemental material is available for this article.</i> Published under a CC BY 4.0 license.
The recent rise of deep learning (DL) and its promising capabilities in capturing non-explicit detail from large datasets have attracted substantial research attention in the field of medical image processing. DL provides grounds for technological development of computer-aided diagnosis and segmentation in radiology and radiation oncology. Amongst the anatomical locations where recent auto-segmentation algorithms have been employed, the pelvis remains one of the most challenging due to large intra- and inter-patient soft-tissue variabilities. This review provides a comprehensive, non-systematic and clinically-oriented overview of 74 DL-based segmentation studies, published between January 2016 and December 2020, for bladder, prostate, cervical and rectal cancers on computed tomography (CT) and magnetic resonance imaging (MRI), highlighting the key findings, challenges and limitations.
Purpose To compare disease detection of myeloma using contemporary whole-body (WB) MRI and fluorine 18 (<sup>18</sup>F) fluorodeoxyglucose (FDG) PET/CT protocols and to correlate imaging with laboratory estimates of disease burden, including molecular characteristics. Materials and Methods In this observational, prospective study, participants were recruited from November 2015 to March 2018 who had a diagnosis of myeloma, who were planned to undergo chemotherapy and autologous stem cell transplantation, and who underwent baseline WB-MRI and FDG PET/CT (ClinicalTrials.gov identifier NCT02403102). Baseline clinical data, including genetics, were collected. Paired methods were used to compare burden and patterns of disease. Results Sixty participants (mean age, 60 years ± 9 [standard deviation]; 35 men) underwent baseline WB-MRI and FDG PET/CT. WB-MRI showed significantly higher detection for focal lesions at all anatomic sites (except ribs, scapulae, and clavicles) and for diffuse disease at all sites. Two participants presented with two or more focal lesions smaller than 5 mm only at WB-MRI but not FDG PET/CT. Participants with diffuse disease at MRI had higher plasma cell infiltration (percentage of nucleated cells: median, 60% [interquartile range {IQR}, 50%-61%] vs 15% [IQR, 4%-50%]; <i>P</i> = .03) and paraprotein levels (median, 32.0 g/L [IQR, 24.0-48.0 g/L] vs 20.0 g/L [IQR, 12.0-22.6 g/L]; <i>P</i> = .02) compared with those without diffuse disease. All genetically high-risk tumors showed diffuse infiltration at WB-MRI. Conclusion WB-MRI helped detect a higher number of myeloma lesions than FDG PET/CT, and diffuse disease detected at WB-MRI correlated with laboratory measures of disease burden and molecular markers of risk. <b>Keywords:</b> MR-Imaging, Skeletal-Appendicular, Skeletal-Axial, Bone Marrow, Hematologic Diseases, Oncology Clinical trial registration no. NCT02403102. <i>Supplemental material is available for this article.</i> © RSNA, 2021.
<h4>Background</h4>Whole-body (WB) MRI, which includes diffusion-weighted imaging (DWI) and T<sub>1</sub>-w Dixon, permits sensitive detection of marrow disease in addition to qualitative and quantitative measurements of disease and response to treatment of bone marrow. We report on the first study to embed standardised WB-MRI within a prospective, multi-centre myeloma clinical trial (IMAGIMM trial, sub-study of OPTIMUM/MUKnine) to explore the use of WB-MRI to detect minimal residual disease after treatment.<h4>Methods</h4>The standardised MY-RADS WB-MRI protocol was set up on a local 1.5 T scanner. An imaging manual describing the MR protocol, quality assurance/control procedures and data transfer was produced and provided to sites. For non-identical scanners (different vendor or magnet strength), site visits from our physics team were organised to support protocol optimisation. The site qualification process included review of phantom and volunteer data acquired at each site and a teleconference to brief the multidisciplinary team. Image quality of initial patients at each site was assessed.<h4>Results</h4>WB-MRI was successfully set up at 12 UK sites involving 3 vendor systems and two field strengths. Four main protocols (1.5 T Siemens, 3 T Siemens, 1.5 T Philips and 3 T GE scanners) were generated. Scanner limitations (hardware and software) and scanning time constraint required protocol modifications for 4 sites. Nevertheless, shared methodology and imaging protocols enabled other centres to obtain images suitable for qualitative and quantitative analysis.<h4>Conclusions</h4>Standardised WB-MRI protocols can be implemented and supported in prospective multi-centre clinical trials. Trial registration NCT03188172 clinicaltrials.gov; registration date 15th June 2017 https://clinicaltrials.gov/ct2/show/study/NCT03188172.
<h4>Objectives</h4>Investigate the laboratory, imaging and procedural factors that are associated with a tumour-positive and/or NGS-feasible CT-guided sclerotic bone lesion biopsy result in cancer patients.<h4>Methods</h4>In total, 113 CT-guided bone biopsies performed in cancer patients by an interventional radiologist in one institution were retrospectively reviewed. Sixty-five sclerotic bone biopsies were eventually included and routine blood parameters and tumour marker levels were recorded. Non-contrast (NC) biopsy CTs (65), contrast-enhanced CTs (24), and PET/CTs (22) performed within four weeks of biopsy were reviewed; lesion location, diameter, lesion-to-cortex distance, and NC-CT appearance (dense-sclerosis versus mild-sclerosis) were noted. Mean NC-CT, CE-CT HU, and PET SUVmax were derived from biopsy tract and lesion segmentations. Needle diameter, tract length, and number of samples were noted. Comparisons between tumour-positive/negative and next-generation sequencing (NGS)-feasible/non-feasible biopsies determined significant (p < 0.05) laboratory, imaging, and procedural parameter differences.<h4>Results</h4>Seventy-four percent of biopsies were tumour-positive. NGS was feasible in 22/30 prostate cancer patients (73%). Neither laboratory blood parameters, PET/CT availability, size, nor lesion-to-cortex distance affected diagnostic yield or NGS feasibility (p > 0.298). Eighty-seven percent of mildly sclerotic bone (mean 244 HU) biopsies were positive compared with 56% in dense sclerosis (622 HU, p = 0.005) and NC-CT lesion HU was significantly lower in positive biopsies (p = 0.003). A 610 HU threshold yielded 89% PPV for tumour-positive biopsies and a 370 HU threshold 94% PPV for NGS-feasible biopsies. FDG-PET and procedural parameters were non-significant factors (each p > 0.055).<h4>Conclusion</h4>In cancer patients with sclerotic bone disease, targeting areas of predominantly mild sclerosis in lower CT-attenuation lesions can improve tumour tissue yield and NGS feasibility.<h4>Key points</h4>• Areas of predominantly mild sclerosis should be preferred to areas of predominantly dense sclerosis for CT-guided bone biopsies in cancer patients. • Among sclerotic bone lesions in prostate cancer patients, lesions with a mean HU below 370 should be preferred as biopsy targets to improve NGS feasibility. • Laboratory parameters and procedure related factors may have little implications for CT-guided sclerotic bone biopsy success.
Whole-body magnetic resonance imaging (MRI) is now a crucial tool for the assessment of the extent of systemic malignant bone disease and response to treatment, and forms part of national and international recommendations for imaging patients with myeloma or metastatic prostate cancer. Recent developments in scanners have enabled acquisition of good-quality whole-body MRI data within 45 minutes on modern MRI systems from all main manufacturers. This provides complimentary morphological and functional whole-body imaging; however, lack of prior experience and acquisition times required can act as a barrier to adoption in busy radiology departments. This article aims to tackle the former by reviewing the indications and providing guidance for technical delivery and clinical interpretation of whole-body MRI for patients with malignant bone disease.
<h4>Background</h4>Morphologic features yield low diagnostic accuracy to distinguish between diseased and normal lymph nodes. The purpose of this study was to compare diseased lymphomatous and normal lymph nodes using global apparent diffusion coefficient (gADC) histogram parameters derived from whole-body diffusion-weighted MRI (WB-DWI).<h4>Methods</h4>1.5 Tesla WB-DWI of 23 lymphoma patients and 20 healthy volunteers performed between 09/2010 and 07/2015 were retrospectively reviewed. All diseased nodal groups in the lymphoma cohort and all nodes visible on b900 images in healthy volunteers were segmented from neck to groin to generate a total diffusion volume (tDV). A connected component-labelling algorithm separated spatially distinct nodes. Mean, median, skewness, kurtosis, minimum, maximum, interquartile range (IQR), standard deviation (SD), 10<sup>th</sup> and 90<sup>th</sup> centile of the gADC distribution were derived from the tDV of each patient/volunteer and from spatially distinct nodes. gADC and regional nodal ADC parameters were compared between malignant and normal nodes using <i>t</i>-tests and ROC curve analyses. A P value ≤0.05 was deemed statistically significant.<h4>Results</h4>Mean, median, IQR, 10th and 90th centiles of gADC and regional nodal ADC values were significantly lower in diseased compared with normal lymph nodes. Skewness, kurtosis and tDV were significantly higher in lymphoma. The SD, min and max gADC showed no significant difference between the two groups (P>0.128). The diagnostic accuracies of gADC parameters by AUC from highest to lowest were: 10th centile, mean, median, 90th centile, skewness, kurtosis and IQR. A 10th centile gADC threshold of 0.68×10<sup>-3</sup> mm<sup>2</sup>/s identified diseased lymphomatous nodes with 91% sensitivity and 95% specificity.<h4>Conclusions</h4>WB-DWI derived gADC histogram parameters can distinguish between malignant lymph nodes of lymphoma patients and normal lymph nodes of healthy individuals.
<h4>Background</h4>Radium-223 is a bone-seeking, alpha-emitting radionuclide used in metastatic castration-resistant prostate cancer (mCRPC). Radium-223 increases the risk of fracture when used in combination with abiraterone and prednisolone. The risk of fracture in men receiving radium-223 monotherapy is unclear.<h4>Patients and methods</h4>This was a prospective, multicenter phase II study of radium-223 in 36 men with mCRPC and a reference cohort (n = 36) matched for fracture risk and not treated with radium-223. Bone fractures were assessed using whole-body magnetic resonance imaging. The primary outcome was risk of new fractures.<h4>Results</h4>Thirty-six patients were treated with up to six 4-week cycles of radium-223. With a median follow-up of 16.3 months, 74 new fractures were identified in 20 patients. Freedom from fracture was 56% (95% confidence interval, 35.3-71.6) at 12 months. On multivariate analysis, prior corticosteroid use was associated with risk of fracture. In the reference cohort (n = 36), 16 new fractures were identified in 12 patients over a median follow-up of 24 months. Across both cohorts, 67% of all fractures occurred at uninvolved bone.<h4>Conclusions</h4>Men with mCRPC, and particularly those treated with radium-223, are at risk of fracture. They should receive a bone health agent to reduce the risk of fragility fractures.
<h4>Objectives</h4>To evaluate whether multiparametric bone MRI (mpBMRI) utilising a combination of DWI signal, ADC and relative fat-fraction (rFF) can identify bone metastases, which provide high diagnostic biopsy yield and next-generation genomic sequencing (NGS) feasibility.<h4>Methods</h4>A total of 150 CT-guided bone biopsies performed by interventional radiologists (3/2013 to 2/2021) at our centre were reviewed. In 43 patients, contemporaneous DWI and rFF images, calculated from 2-point T1w Dixon MRI, were available. For each biopsied lesion, a region of interest (ROI) was delineated on ADC and rFF images and the following MRI parameters were recorded: visual classification of DWI signal intensity (SI), mean, median, 10th and 90th centile ADC and rFF values. Non-parametric tests were used to compare values between tumour positive/negative biopsies and feasible/non-feasible NGS, with p-values < 0.05 deemed significant.<h4>Results</h4>The mpBMRI combination high DWI signal, mean ADC < 1100 µm<sup>2</sup>/s and mean rFF < 20% identified tumour-positive biopsies with 82% sensitivity, 80% specificity, a positive predictive value (PPV) of 93% (p = 0.001) and NGS feasibility with 91% sensitivity, 78% specificity and 91% PPV (p < 0.001). The single MRI parameters DWI signal, ADC and rFF failed to distinguish between tumour-positive and tumour-negative biopsies (each p > 0.082). In NGS feasible biopsies, mean and 90th centile rFF were significantly smaller (each p < 0.041). Single ADC parameters did not show significant difference regarding NGS feasibility (each p > 0.292).<h4>Conclusions</h4>MpBMRI utilising the combination of DWI signal, ADC and rFF can identify active bone metastases, which provide biopsy tissue with high diagnostic yield and NGS feasibility.<h4>Key points</h4>• Multiparametric bone MRI with diffusion-weighted and relative fat-fraction images helps to identify active bone metastases suitable for CT-guided biopsy. • Target lesions for CT-guided bone biopsies in cancer patients can be chosen with greater confidence. • CT-guided bone biopsy success rates, especially yielding sufficient viable tissue for advanced molecular tissue analyses, can be improved.
<h4>Background</h4>Whole body DWI (WB-DWI) enables the identification of lymph nodes for disease evaluation. However, quantitative data of benign lymph nodes across the body are lacking to allow meaningful comparison of diseased states. We evaluated apparent diffusion coefficient (ADC) histogram parameters of all visible lymph nodes in healthy volunteers on WB-DWI and compared differences in nodal ADC values between anatomical regions.<h4>Methods</h4>WB-DWI was performed on a 1.5 T MR system in 20 healthy volunteers (7 female, 13 male, mean age 35 years). The b900 images were evaluated by two radiologists and all visible nodes from the neck to groin areas were segmented and individual nodal median ADC recorded. All segmented nodes in a patient were summated to generate the total nodal volume. Descriptors of the global ADC histogram, derived from individual node median ADCs, including mean, median, skewness and kurtosis were obtained for the global volume and each nodal region per patient. ADC values between nodal regions were compared using one-way ANOVA with Bonferroni post hoc tests and a p-value ≤0.05 was deemed statistically significant.<h4>Results</h4>One thousand sixty-seven lymph nodes were analyzed. The global mean and median ADC of all lymph nodes were 1.12 ± 0.27 (10<sup>- 3</sup> mm<sup>2</sup>/s) and 1.09 (10<sup>- 3</sup> mm<sup>2</sup>/s). The average median ADC skewness was 0.25 ± 0.02 and average median ADC kurtosis was 0.34 ± 0.04. The ADC values of intrathoracic, portal and retroperitoneal nodes were significantly higher (1.53 × 10<sup>- 3</sup>, 1.75 × 10<sup>- 3</sup> and 1.58 × 10<sup>- 3</sup> mm<sup>2</sup>/s respectively) than in other regions. Intrathoracic, portal and mesenteric nodes were relatively uncommon, accounting for only 3% of the total nodes segmented.<h4>Conclusions</h4>The global mean and median ADC of all lymph nodes were 1.12 ± 0.27 (10<sup>- 3</sup> mm<sup>2</sup>/s) and 1.09 (10<sup>- 3</sup> mm<sup>2</sup>/s). Intrathoracic, portal and retroperitoneal nodes display significantly higher ADCs. Normal intrathoracic, portal and mesenteric nodes are infrequently visualized on WB-DWI of healthy individuals.<h4>Trial registration</h4>Royal Marsden Hospital committee for clinical research registration number 09/H0801/86, 19.10.2009.
<h4>Background</h4>Surveillance is universally recommended for non-small cell lung cancer (NSCLC) patients treated with curative-intent radiotherapy. High-quality evidence to inform optimal surveillance strategies is lacking. Machine learning demonstrates promise in accurate outcome prediction for a variety of health conditions. The purpose of this study was to utilise readily available patient, tumour, and treatment data to develop, validate and externally test machine learning models for predicting recurrence, recurrence-free survival (RFS) and overall survival (OS) at 2 years from treatment.<h4>Methods</h4>A retrospective, multicentre study of patients receiving curative-intent radiotherapy for NSCLC was undertaken. A total of 657 patients from 5 hospitals were eligible for inclusion. Data pre-processing derived 34 features for predictive modelling. Combinations of 8 feature reduction methods and 10 machine learning classification algorithms were compared, producing risk-stratification models for predicting recurrence, RFS and OS. Models were compared with 10-fold cross validation and an external test set and benchmarked against TNM-stage and performance status. Youden Index was derived from validation set ROC curves to distinguish high and low risk groups and Kaplan-Meier analyses performed.<h4>Findings</h4>Median follow-up time was 852 days. Parameters were well matched across training-validation and external test sets: Mean age was 73 and 71 respectively, and recurrence, RFS and OS rates at 2 years were 43% vs 34%, 54% vs 47% and 54% vs 47% respectively. The respective validation and test set AUCs were as follows: 1) RFS: 0·682 (0·575-0·788) and 0·681 (0·597-0·766), 2) Recurrence: 0·687 (0·582-0·793) and 0·722 (0·635-0·81), and 3) OS: 0·759 (0·663-0·855) and 0·717 (0·634-0·8). Our models were superior to TNM stage and performance status in predicting recurrence and OS.<h4>Interpretation</h4>This robust and ready to use machine learning method, validated and externally tested, sets the stage for future clinical trials entailing quantitative personalised risk-stratification and surveillance following curative-intent radiotherapy for NSCLC.<h4>Funding</h4>A full list of funding bodies that contributed to this study can be found in the Acknowledgements section.
<h4>Background</h4>Computed tomography (CT) and magnetic resonance imaging (MRI) are the mainstay imaging modalities in radiotherapy planning. In MR-Linac treatment, manual annotation of organs-at-risk (OARs) and clinical volumes requires a significant clinician interaction and is a major challenge. Currently, there is a lack of available pre-annotated MRI data for training supervised segmentation algorithms. This study aimed to develop a deep learning (DL)-based framework to synthesize pelvic T<sub>1</sub>-weighted MRI from a pre-existing repository of clinical planning CTs.<h4>Methods</h4>MRI synthesis was performed using UNet++ and cycle-consistent generative adversarial network (Cycle-GAN), and the predictions were compared qualitatively and quantitatively against a baseline UNet model using pixel-wise and perceptual loss functions. Additionally, the Cycle-GAN predictions were evaluated through qualitative expert testing (4 radiologists), and a pelvic bone segmentation routine based on a UNet architecture was trained on synthetic MRI using CT-propagated contours and subsequently tested on real pelvic T<sub>1</sub> weighted MRI scans.<h4>Results</h4>In our experiments, Cycle-GAN generated sharp images for all pelvic slices whilst UNet and UNet++ predictions suffered from poorer spatial resolution within deformable soft-tissues (e.g. bladder, bowel). Qualitative radiologist assessment showed inter-expert variabilities in the test scores; each of the four radiologists correctly identified images as acquired/synthetic with 67%, 100%, 86% and 94% accuracy. Unsupervised segmentation of pelvic bone on T1-weighted images was successful in a number of test cases.<h4>Conclusion</h4>Pelvic MRI synthesis is a challenging task due to the absence of soft-tissue contrast on CT. Our study showed the potential of deep learning models for synthesizing realistic MR images from CT, and transferring cross-domain knowledge which may help to expand training datasets for 21 development of MR-only segmentation models.
A shift in radiology to a data-driven specialty has been unlocked by synergistic developments in imaging biomarkers (IB) and computational science. This is advancing the capability to deliver "virtual biopsies" within oncology. The ability to non-invasively probe tumour biology both spatially and temporally would fulfil the potential of imaging to inform management of complex tumours; improving diagnostic accuracy, providing new insights into inter- and intra-tumoral heterogeneity and individualised treatment planning and monitoring. Soft tissue sarcomas (STS) are rare tumours of mesenchymal origin with over 150 histological subtypes and notorious heterogeneity. The combination of inter- and intra-tumoural heterogeneity and the rarity of the disease remain major barriers to effective treatments. We provide an overview of the process of successful IB development, the key imaging and computational advancements in STS including quantitative magnetic resonance imaging, radiomics and artificial intelligence, and the studies to date that have explored the potential biological surrogates to imaging metrics. We discuss the promising future directions of IBs in STS and illustrate how the routine clinical implementation of a virtual biopsy has the potential to revolutionise the management of this group of complex cancers and improve clinical outcomes.
<h4>Background and purpose</h4>Magnetic resonance imaging integrated linear accelerator (MR-Linac) platforms enable acquisition of diffusion weighted imaging (DWI) during treatment providing potential information about treatment response. Obtaining DWI on these platforms is technically different from diagnostic magnetic resonance imaging (MRI) scanners. The aim of this project was to determine feasibility of obtaining DWI and calculating apparent diffusion coefficient (ADC) parameters longitudinally in rectal cancer patients on the MR-Linac.<h4>Materials and methods</h4>Nine patients undergoing treatment on MR-Linac had DWI acquired using b-values 0, 30, 150, 500 s/mm<sup>2</sup>. Gross tumour volume (GTV) and normal tissue was delineated on DWI throughout treatment and median ADC was calculated using an in-house tool (pyOsirix ®).<h4>Results</h4>Seven out of nine patients were included in the analysis; all demonstrated downstaging at follow-up. A total of 63 out of 70 DWI were analysed (7 excluded due to poor image quality). An increasing trend of ADC median for GTV (1.15 × 10<sup>-3</sup> mm<sup>2</sup>/s interquartile range (IQ): 1.05-1.17 vs 1.59 × 10<sup>-3</sup> mm<sup>2</sup>/s IQ: 1.37 - 1.64; <i>p = 0.0156</i>), correlating to treatment response. In comparison ADC median for normal tissue remained the same between first and last fraction (1.61 × 10<sup>-3</sup> mm<sup>2</sup>/s IQ: 1.56-1.71 vs 1.67 × 10<sup>-3</sup> mm<sup>2</sup>/s IQ: 1.37-2.00; p = 0.9375).<h4>Conclusions</h4>DWI assessment in rectal cancer patients on MR-Linac is feasible. Initial results provide foundations for further studies to determine DWI use for treatment adaptation in rectal cancer.
<h4>Background</h4>Size-based assessments are inaccurate indicators of tumor response in soft-tissue sarcoma (STS), motivating the requirement for new response imaging biomarkers for this rare and heterogeneous disease. In this study, we assess the test-retest repeatability of radiomic features from MR diffusion-weighted imaging (DWI) and derived maps of apparent diffusion coefficient (ADC) in retroperitoneal STS and compare baseline repeatability with changes in radiomic features following radiotherapy (RT).<h4>Materials and methods</h4>Thirty patients with retroperitoneal STS received an MR examination prior to treatment, of whom 23/30 were investigated in our repeatability analysis having received repeat baseline examinations and 14/30 patients were investigated in our post-treatment analysis having received an MR examination after completing pre-operative RT. One hundred and seven radiomic features were extracted from the full manually delineated tumor region using PyRadiomics. Test-retest repeatability was assessed using an intraclass correlation coefficient (baseline ICC), and post-radiotherapy variance analysis (post-RT-IMS) was used to compare the change in radiomic feature value to baseline repeatability.<h4>Results</h4>For the ADC maps and DWI images, 101 and 102 features demonstrated good baseline repeatability (baseline ICC > 0.85), respectively. Forty-three and 2 features demonstrated both good baseline repeatability and a high post-RT-IMS (>0.85), respectively. Pearson correlation between the baseline ICC and post-RT-IMS was weak (0.432 and 0.133, respectively).<h4>Conclusions</h4>The ADC-based radiomic analysis shows better test-retest repeatability compared with features derived from DWI images in STS, and some of these features are sensitive to post-treatment change. However, good repeatability at baseline does not imply sensitivity to post-treatment change.
<h4>Purpose</h4>To use deep learning to calculate the uncertainty in apparent diffusion coefficient (σADC) voxel-wise measurements to clinically impact the monitoring of treatment response and improve the quality of ADC maps.<h4>Materials and methods</h4>We use a uniquely designed diffusion-weighted imaging (DWI) acquisition protocol that provides gold-standard measurements of σADC to train a deep learning model on two separate cohorts: 16 patients with prostate cancer and 28 patients with mesothelioma. Our network was trained with a novel cost function, which incorporates a perception metric and a b-value regularisation term, on ADC maps calculated by combinations of 2 or 3 b-values (e.g. 50/600/900, 50/900, 50/600, 600/900 s/mm<sup>2</sup>). We compare the accuracy of the deep-learning based approach for estimation of σADC with gold-standard measurements.<h4>Results</h4>The model accurately predicted the σADC for every b-value combination in both cohorts. Mean values of σADC within areas of active disease deviated from those measured by the gold-standard by 4.3% (range, 2.87-6.13%) for the prostate and 3.7% (range, 3.06-4.54%) for the mesothelioma cohort. We also showed that the model can easily be adapted for a different DWI protocol and field-of-view with only a few images (as little as a single patient) using transfer learning.<h4>Conclusion</h4>Deep learning produces maps of σADC from standard clinical diffusion-weighted images (DWI) when 2 or more b-values are available.
<h4>Objective</h4>To establish optimised diffusion weightings ('b-values') for acquisition of whole-body diffusion-weighted MRI (WB-DWI) for estimation of the apparent diffusion coefficient (ADC) in patients with metastatic melanoma (MM). Existing recommendations for WB-DWI have not been optimised for the tumour properties in MM; therefore, evaluation of acquisition parameters is essential before embarking on larger studies.<h4>Methods</h4>Retrospective clinical data and phantom experiments were used. Clinical data comprised 125 lesions from 14 examinations in 11 patients with multifocal MM, imaged before and/or after treatment with immunotherapy at a single institution. ADC estimates from these data were applied to a model to estimate the optimum b-value. A large non-diffusing phantom was used to assess eddy current-induced geometric distortion.<h4>Results</h4>Considering all tumour sites from pre- and post-treatment examinations together, metastases exhibited a large range of mean ADC values, [0.67-1.49] × 10<sup>-3</sup> mm<sup>2</sup>/s, and the optimum high b-value (b<sub>high</sub>) for ADC estimation was 1100 (10th-90th percentile: 740-1790) s/mm<sup>2</sup>. At higher b-values, geometric distortion increased, and longer echo times were required, leading to reduced signal.<h4>Conclusions</h4>Theoretical optimisation gave an optimum b<sub>high</sub> of 1100 (10th-90th percentile: 740-1790) s/mm<sup>2</sup> for ADC estimation in MM, with the large range of optimum b-values reflecting the wide range of ADC values in these tumours. Geometric distortion and minimum echo time increase at higher b-values and are not included in the theoretical optimisation; b<sub>high</sub> in the range 750-1100 s/mm<sup>2</sup> should be adopted to maintain acceptable image quality but performance should be evaluated for a specific scanner.<h4>Key points</h4>• Theoretical optimisation gave an optimum high b-value of 1100 (10th-90th percentile: 740-1790) s/mm<sup>2</sup> for ADC estimation in metastatic melanoma. • Considering geometric distortion and minimum echo time (TE), a b-value in the range 750-1100 s/mm<sup>2</sup> is recommended. • Sites should evaluate the performance of specific scanners to assess the effect of geometric distortion and minimum TE.
<h4>Background</h4>Large lung nodules (≥15 mm) have the highest risk of malignancy, and may exhibit important differences in phenotypic or clinical characteristics to their smaller counterparts. Existing risk models do not stratify large nodules well. We aimed to develop and validate an integrated segmentation and classification pipeline, incorporating deep-learning and traditional radiomics, to classify large lung nodules according to cancer risk.<h4>Methods</h4>502 patients from five U.K. centres were recruited to the large-nodule arm of the retrospective LIBRA study between July 2020 and April 2022. 838 CT scans were used for model development, split into training and test sets (70% and 30% respectively). An nnUNet model was trained to automate lung nodule segmentation. A radiomics signature was developed to classify nodules according to malignancy risk. Performance of the radiomics model, termed the large-nodule radiomics predictive vector (LN-RPV), was compared to three radiologists and the Brock and Herder scores.<h4>Findings</h4>499 patients had technically evaluable scans (mean age 69 ± 11, 257 men, 242 women). In the test set of 252 scans, the nnUNet achieved a DICE score of 0.86, and the LN-RPV achieved an AUC of 0.83 (95% CI 0.77-0.88) for malignancy classification. Performance was higher than the median radiologist (AUC 0.75 [95% CI 0.70-0.81], DeLong p = 0.03). LN-RPV was robust to auto-segmentation (ICC 0.94). For baseline solid nodules in the test set (117 patients), LN-RPV had an AUC of 0.87 (95% CI 0.80-0.93) compared to 0.67 (95% CI 0.55-0.76, DeLong p = 0.002) for the Brock score and 0.83 (95% CI 0.75-0.90, DeLong p = 0.4) for the Herder score. In the international external test set (n = 151), LN-RPV maintained an AUC of 0.75 (95% CI 0.63-0.85). 18 out of 22 (82%) malignant nodules in the Herder 10-70% category in the test set were identified as high risk by the decision-support tool, and may have been referred for earlier intervention.<h4>Interpretation</h4>The model accurately segments and classifies large lung nodules, and may improve upon existing clinical models.<h4>Funding</h4>This project represents independent research funded by: 1) Royal Marsden Partners Cancer Alliance, 2) the Royal Marsden Cancer Charity, 3) the National Institute for Health Research (NIHR) Biomedical Research Centre at the Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, London, 4) the National Institute for Health Research (NIHR) Biomedical Research Centre at Imperial College London, 5) Cancer Research UK (C309/A31316).
Handcrafted and deep learning (DL) radiomics are popular techniques used to develop computed tomography (CT) imaging-based artificial intelligence models for COVID-19 research. However, contrast heterogeneity from real-world datasets may impair model performance. Contrast-homogenous datasets present a potential solution. We developed a 3D patch-based cycle-consistent generative adversarial network (cycle-GAN) to synthesize non-contrast images from contrast CTs, as a data homogenization tool. We used a multi-centre dataset of 2078 scans from 1,650 patients with COVID-19. Few studies have previously evaluated GAN-generated images with handcrafted radiomics, DL and human assessment tasks. We evaluated the performance of our cycle-GAN with these three approaches. In a modified Turing-test, human experts identified synthetic vs acquired images, with a false positive rate of 67% and Fleiss' Kappa 0.06, attesting to the photorealism of the synthetic images. However, on testing performance of machine learning classifiers with radiomic features, performance decreased with use of synthetic images. Marked percentage difference was noted in feature values between pre- and post-GAN non-contrast images. With DL classification, deterioration in performance was observed with synthetic images. Our results show that whilst GANs can produce images sufficient to pass human assessment, caution is advised before GAN-synthesized images are used in medical imaging applications.
Recurrence occurs in up to 36% of patients treated with curative-intent radiotherapy for NSCLC. Identifying patients at higher risk of recurrence for more intensive surveillance may facilitate the earlier introduction of the next line of treatment. We aimed to use radiotherapy planning CT scans to develop radiomic classification models that predict overall survival (OS), recurrence-free survival (RFS) and recurrence two years post-treatment for risk-stratification. A retrospective multi-centre study of >900 patients receiving curative-intent radiotherapy for stage I-III NSCLC was undertaken. Models using radiomic and/or clinical features were developed, compared with 10-fold cross-validation and an external test set, and benchmarked against TNM-stage. Respective validation and test set AUCs (with 95% confidence intervals) for the radiomic-only models were: (1) OS: 0.712 (0.592-0.832) and 0.685 (0.585-0.784), (2) RFS: 0.825 (0.733-0.916) and 0.750 (0.665-0.835), (3) Recurrence: 0.678 (0.554-0.801) and 0.673 (0.577-0.77). For the combined models: (1) OS: 0.702 (0.583-0.822) and 0.683 (0.586-0.78), (2) RFS: 0.805 (0.707-0.903) and 0·755 (0.672-0.838), (3) Recurrence: 0·637 (0.51-0.·765) and 0·738 (0.649-0.826). Kaplan-Meier analyses demonstrate OS and RFS difference of >300 and >400 days respectively between low and high-risk groups. We have developed validated and externally tested radiomic-based prediction models. Such models could be integrated into the routine radiotherapy workflow, thus informing a personalised surveillance strategy at the point of treatment. Our work lays the foundations for future prospective clinical trials for quantitative personalised risk-stratification for surveillance following curative-intent radiotherapy for NSCLC.
<h4>Objectives</h4>To assess the repeatability of quantitative multiparametric whole-body MRI (mpWB-MRI) parameters in advanced prostate cancer (APC) bone metastases.<h4>Methods</h4>1.5T MRI was performed twice on the same day in 10 APC patients. MpWB-MRI-included diffusion weighted imaging (DWI) and <i>T</i><sub>1</sub>-weighted gradient-echo 2-point Dixon sequences. ADC and relative fat-fraction percentage (rFF%) maps were calculated, respectively. A radiologist delineated up to 10 target bone metastases per study. Means of ADC, b900 signal intensity(SI), normalised b900 SI, rFF% and maximum diameter (MD) for each target lesion and overall parameter averages across all targets per patient were recorded. The total disease volume (tDV in ml) was manually delineated on b900 images and mean global (g)ADC was derived. Bland-Altman analyses were performed with calculation of 95% repeatability coefficients (RC).<h4>Results</h4>Seventy-three individual targets (median MD 26 mm) were included. Lesion mean ADC RC was 12.5%, mean b900 SI RC 137%, normalised mean b900 SI RC 110%, rFF% RC 3.2 and target MD RC 5.5 mm (16.3%). Patient target lesion average mean ADC RC was 6.4%, b900 SI RC 104% and normalised mean b900 SI RC 39.6%. Target average rFF% RC was 1.8, average MD RC 1.3 mm (4.8%). tDV segmentation RC was 6.4% and mean gADC RC 5.3%.<h4>Conclusions</h4>APC bone metastases' ADC, rFF% and maximum diameter, tDV and gADC show good repeatability.<h4>Advances in knowledge</h4>APC bone metastases' mean ADC and rFF% measurements of single lesions and global disease volumes are repeatable, supporting their potential role as quantitative biomarkers in metastatic bone disease.
T<sub>2</sub>-weighted magnetic resonance imaging (MRI) and diffusion-weighted imaging (DWI) are essential components of cervical cancer diagnosis. However, combining these channels for the training of deep learning models is challenging due to image misalignment. Here, we propose a novel multi-head framework that uses dilated convolutions and shared residual connections for the separate encoding of multiparametric MRI images. We employ a residual U-Net model as a baseline, and perform a series of architectural experiments to evaluate the tumor segmentation performance based on multiparametric input channels and different feature encoding configurations. All experiments were performed on a cohort of 207 patients with locally advanced cervical cancer. Our proposed multi-head model using separate dilated encoding for T<sub>2</sub>W MRI and combined b1000 DWI and apparent diffusion coefficient (ADC) maps achieved the best median Dice similarity coefficient (DSC) score, 0.823 (confidence interval (CI), 0.595-0.797), outperforming the conventional multi-channel model, DSC 0.788 (95% CI, 0.568-0.776), although the difference was not statistically significant (<i>p</i> > 0.05). We investigated channel sensitivity using 3D GRAD-CAM and channel dropout, and highlighted the critical importance of T<sub>2</sub>W and ADC channels for accurate tumor segmentation. However, our results showed that b1000 DWI had a minor impact on the overall segmentation performance. We demonstrated that the use of separate dilated feature extractors and independent contextual learning improved the model's ability to reduce the boundary effects and distortion of DWI, leading to improved segmentation performance. Our findings could have significant implications for the development of robust and generalizable models that can extend to other multi-modal segmentation applications.
<h4>Background</h4>Whole-Body Diffusion-Weighted Imaging (WBDWI) is an established technique for staging and evaluating treatment response in patients with multiple myeloma (MM) and advanced prostate cancer (APC). However, WBDWI scans show inter- and intra-patient intensity signal variability. This variability poses challenges in accurately quantifying bone disease, tracking changes over follow-up scans, and developing automated tools for bone lesion delineation. Here, we propose a novel automated pipeline for inter-station, inter-scan image signal standardisation on WBDWI that utilizes robust segmentation of the spinal canal through deep learning.<h4>Methods</h4>We trained and validated a supervised 2D U-Net model to automatically delineate the spinal canal (both the spinal cord and surrounding cerebrospinal fluid, CSF) in an initial cohort of 40 patients who underwent WBDWI for treatment response evaluation (80 scans in total). Expert-validated contours were used as the target standard. The algorithm was further semi-quantitatively validated on four additional datasets (three internal, one external, 207 scans total) by comparing the distributions of average apparent diffusion coefficient (ADC) and volume of the spinal cord derived from a two-component Gaussian mixture model of segmented regions. Our pipeline subsequently standardises WBDWI signal intensity through two stages: (i) normalisation of signal between imaging stations within each patient through histogram equalisation of slices acquired on either side of the station gap, and (ii) inter-scan normalisation through histogram equalisation of the signal derived within segmented spinal canal regions. This approach was semi-quantitatively validated in all scans available to the study (N = 287).<h4>Results</h4>The test dice score, precision, and recall of the spinal canal segmentation model were all above 0.87 when compared to manual delineation. The average ADC for the spinal cord (1.7 × 10<sup>-3</sup> mm<sup>2</sup>/s) showed no significant difference from the manual contours. Furthermore, no significant differences were found between the average ADC values of the spinal cord across the additional four datasets. The signal-normalised, high-b-value images were visualised using a fixed contrast window level and demonstrated qualitatively better signal homogeneity across scans than scans that were not signal-normalised.<h4>Conclusion</h4>Our proposed intensity signal WBDWI normalisation pipeline successfully harmonises intensity values across multi-centre cohorts. The computational time required is less than 10 s, preserving contrast-to-noise and signal-to-noise ratios in axial diffusion-weighted images. Importantly, no changes to the clinical MRI protocol are expected, and there is no need for additional reference MRI data or follow-up scans.
<h4>Objectives</h4>To investigate whether baseline 18F-sodium fluoride (NaF) and 18F-choline PET activity is associated with metastatic castration-resistant prostate cancer (mCRPC) global and individual bone metastases' DWI MR imaging response to radium-223 treatment.<h4>Methods</h4>Thirty-six bone-only mCRPC patients were prospectively recruited from three centers. Whole-body (WB)-MRI with DWI and 18F-NaF and 18F-choline PET/CT were performed at therapy baseline and 8-week intervals. In each patient, bone disease median global (g)ADC change between baseline and follow-up was calculated. Additionally, up to five bone target lesions per patient were delineated and individual median ADC change recorded. An ADC increase > 30% defined response per-patient and per-lesion. For the same targets, baseline 18F-NaF and 18F-choline PET SUVmax were recorded. Mean SUVmax across patient targets was correlated with gADC change and lesion SUVmax with per-lesion ADC change.<h4>Results</h4>A total of 133 lesions in 36 patients (14 responders) were analyzed. 18F-NaF PET per-patient mean SUVmax was significantly higher in responders (median = 56.0 versus 38.7 in non-responders; p = 0.008), with positive correlation between SUVmax and gADC increase (rho = 0.42; p = 0.015). A 48.7 SUVmax threshold identified responders with 77% sensitivity and 75% specificity. Baseline 18F-NaF PET per-lesion SUVmax was higher in responding metastases (median = 51.6 versus 31.8 in non-responding metastases; p = 0.001), with positive correlation between baseline lesion SUVmax and ADC increase (rho = 0.39; p < 0.001). A 36.8 SUVmax threshold yielded 72% sensitivity and 63% specificity. No significant association was found between baseline 18F-choline PET SUVmax and ADC response on a per-patient (p = 0.164) or per-lesion basis (p = 0.921).<h4>Conclusion</h4>18F-NaF PET baseline SUVmax of target mCRPC bone disease showed significant association with response to radium-223 defined by ADC change.<h4>Clinical relevance statement</h4>18F-sodium fluoride PET/CT baseline maximum SUV of castration-resistant prostate cancer bone metastases could be used as a predictive biomarker for response to radium-223 therapy.<h4>Key points</h4>• 18F-sodium fluoride PET baseline SUVmax of castration-resistant prostate cancer bone metastases showed significant association with response to radium-223. • Baseline 18F-sodium fluoride PET can improve patient selection for radium-223 therapy. • Change in whole-body DWI parameters can be used for response correlation with baseline 18F-sodium fluoride PET SUVmax in castration-resistant prostate cancer bone metastases.
<h4>Background</h4>Radium-223 is a bone-seeking, ɑ-emitting radionuclide used to treat men with bone metastases from castration-resistant prostate cancer. Sclerotic bone lesions cannot be evaluated using Response Evaluation Criteria in Solid Tumors. Therefore, imaging response biomarkers are needed.<h4>Methods</h4>We conducted a phase 2 randomized trial to assess disease response to radium-223. Men with metastatic castration-resistant prostate cancer and bone metastases were randomly allocated to 55 or 88 kBq/kg radium-223 every 4 weeks for 6 cycles. Whole-body diffusion-weighted magnetic resonance imaging (DWI) was performed at baseline, at cycles 2 and 4, and after treatment. The primary endpoint was defined as a 30% increase in global median apparent diffusion coefficient.<h4>Results</h4>Disease response on DWI was seen in 14 of 36 evaluable patients (39%; 95% confidence interval = 23% to 56%), with marked interpatient and intrapatient heterogeneity of response. There was an association between prostate-specific antigen response and MRI response (odds ratio = 18.5, 95% confidence interval = 1.32 to 258, P = .013). Mean administered activity of radium-223 per cycle was not associated with global MRI response (P = .216) but was associated with DWI response using a 5-target-lesion evaluation (P = .007). In 26 of 36 (72%) patients, new bone metastases, not present at baseline, were seen on DWI scans during radium-223 treatment.<h4>Conclusions</h4>DWI is useful for assessment of disease response in bone. Response to radium-223 is heterogeneous, both between patients and between different metastases in the same patient. New bone metastases appear during radium-223 treatment.The REASURE trial is registered under ISRCTN17805587.
(1) Background: We assessed the test-re-test repeatability of radiomics in metastatic castration-resistant prostate cancer (mCPRC) bone disease on whole-body diffusion-weighted (DWI) and T1-weighted Dixon MRI. (2) Methods: In 10 mCRPC patients, 1.5 T MRI, including DWI and T1-weighted gradient-echo Dixon sequences, was performed twice on the same day. Apparent diffusion coefficient (ADC) and relative fat-fraction-percentage (rFF%) maps were calculated. Per study, up to 10 target bone metastases were manually delineated on DWI and Dixon images. All 106 radiomic features included in the Pyradiomics toolbox were derived for each target volume from the ADC and rFF% maps. To account for inter- and intra-patient measurement repeatability, the log-transformed individual target measurements were fitted to a hierarchical model, represented as a Bayesian network. Repeatability measurements, including the intraclass correlation coefficient (ICC), were derived. Feature ICCs were compared with mean ADC and rFF ICCs. (3) Results: A total of 65 DWI and 47 rFF% targets were analysed. There was no significant bias for any features. Pairwise correlation revealed fifteen ADC and fourteen rFF% feature sub-groups, without specific patterns between feature classes. The median intra-patient ICC was generally higher than the inter-patient ICC. Features that describe extremes in voxel values (minimum, maximum, range, skewness, and kurtosis) showed generally lower ICCs. Several mostly shape-based texture features were identified, which showed high inter- and intra-patient ICCs when compared with the mean ADC or mean rFF%, respectively. (4) Conclusions: Pyradiomics texture features of mCRPC bone metastases varied greatly in inter- and intra-patient repeatability. Several features demonstrated good repeatability, allowing for further exploration as diagnostic parameters in mCRPC bone disease.