A CT-based radiomics classification model for the prediction of histological type and tumour grade in retroperitoneal sarcoma (RADSARC-R): a retrospective multicohort analysis.

BACKGROUND: Retroperitoneal sarcomas are tumours with a poor prognosis. Upfront characterisation of the tumour is difficult, and under-grading is common. Radiomics has the potential to non-invasively characterise the so-called radiological phenotype of tumours. We aimed to develop and independently validate a CT-based radiomics classification model for the prediction of histological type and grade in retroperitoneal leiomyosarcoma and liposarcoma. METHODS: A retrospective discovery cohort was collated at our centre (Royal Marsden Hospital, London, UK) and an independent validation cohort comprising patients recruited in the phase 3 STRASS study of neoadjuvant radiotherapy in retroperitoneal sarcoma. Patients aged older than 18 years with confirmed primary leiomyosarcoma or liposarcoma proceeding to surgical resection with available contrast-enhanced CT scans were included. Using the discovery dataset, a CT-based radiomics workflow was developed, including manual delineation, sub-segmentation, feature extraction, and predictive model building. Separate probabilistic classifiers for the prediction of histological type and low versus intermediate or high grade tumour types were built and tested. Independent validation was then performed. The primary objective of the study was to develop radiomic classification models for the prediction of retroperitoneal leiomyosarcoma and liposarcoma type and histological grade. FINDINGS: 170 patients recruited between Oct 30, 2016, and Dec 23, 2020, were eligible in the discovery cohort and 89 patients recruited between Jan 18, 2012, and April 10, 2017, were eligible in the validation cohort. In the discovery cohort, the median age was 63 years (range 27-89), with 83 (49%) female and 87 (51%) male patients. In the validation cohort, median age was 59 years (range 33-77), with 46 (52%) female and 43 (48%) male patients. The highest performing model for the prediction of histological type had an area under the receiver operator curve (AUROC) of 0·928 on validation, based on a feature set of radiomics and approximate radiomic volume fraction. The highest performing model for the prediction of histological grade had an AUROC of 0·882 on validation, based on a radiomics feature set. INTERPRETATION: Our validated radiomics model can predict the histological type and grade of retroperitoneal sarcomas with excellent performance. This could have important implications for improving diagnosis and risk stratification in retroperitoneal sarcomas. FUNDING: Wellcome Trust, European Organisation for Research and Treatment of Cancer-Soft Tissue and Bone Sarcoma Group, the National Institutes for Health, and the National Institute for Health and Care Research Biomedical Research Centre at The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research.

Radiomics in Oncology: A Practical Guide.

Radiomics refers to the extraction of mineable data from medical imaging and has been applied within oncology to improve diagnosis, prognostication, and clinical decision support, with the goal of delivering precision medicine. The authors provide a practical approach for successfully implementing a radiomic workflow from planning and conceptualization through manuscript writing. Applications in oncology typically are either classification tasks that involve computing the probability of a sample belonging to a category, such as benign versus malignant, or prediction of clinical events with a time-to-event analysis, such as overall survival. The radiomic workflow is multidisciplinary, involving radiologists and data and imaging scientists, and follows a stepwise process involving tumor segmentation, image preprocessing, feature extraction, model development, and validation. Images are curated and processed before segmentation, which can be performed on tumors, tumor subregions, or peritumoral zones. Extracted features typically describe the distribution of signal intensities and spatial relationship of pixels within a region of interest. To improve model performance and reduce overfitting, redundant and nonreproducible features are removed. Validation is essential to estimate model performance in new data and can be performed iteratively on samples of the dataset (cross-validation) or on a separate hold-out dataset by using internal or external data. A variety of noncommercial and commercial radiomic software applications can be used. Guidelines and artificial intelligence checklists are useful when planning and writing up radiomic studies. Although interest in the field continues to grow, radiologists should be familiar with potential pitfalls to ensure that meaningful conclusions can be drawn. Online supplemental material is available for this article. Published under a CC BY 4.0 license.

Radiomic features of cervical cancer on T2-and diffusion-weighted MRI: Prognostic value in low-volume tumors suitable for trachelectomy.

BACKGROUND: Textural features extracted from MRI potentially provide prognostic information additional to volume for influencing surgical management of cervical cancer. PURPOSE: To identify textural features that differ between cervical tumors above and below the volume threshold of eligibility for trachelectomy and determine their value in predicting recurrence in patients with low-volume tumors. METHODS: Of 378 patients with Stage1-2 cervical cancer imaged prospectively (3T, endovaginal coil), 125 had well-defined, histologically-confirmed squamous or adenocarcinomas with >100 voxels (>0.07 cm3) suitable for radiomic analysis. Regions-of-interest outlined the whole tumor on T2-W images and apparent diffusion coefficient (ADC) maps. Textural features based on grey-level co-occurrence matrices were compared (Mann-Whitney test with Bonferroni correction) between tumors greater (n = 46) or less (n = 79) than 4.19 cm3. Clustering eliminated correlated variables. Significantly different features were used to predict recurrence (regression modelling) in surgically-treated patients with low-volume tumors and compared with a model using clinico-pathological features. RESULTS: Textural features (Dissimilarity, Energy, ClusterProminence, ClusterShade, InverseVariance, Autocorrelation) in 6 of 10 clusters from T2-W and ADC data differed between high-volume (mean ± SD 15.3 ± 11.7 cm3) and low-volume (mean ± SD 1.3 ± 1.2 cm3) tumors. (p < 0.02). In low-volume tumors, predicting recurrence was indicated by: Dissimilarity, Energy (ADC-radiomics, AUC = 0.864); Dissimilarity, ClusterProminence, InverseVariance (T2-W-radiomics, AUC = 0.808); Volume, Depth of Invasion, LymphoVascular Space Invasion (clinico-pathological features, AUC = 0.794). Combining ADC-radiomic (but not T2-radiomic) and clinico-pathological features improved prediction of recurrence compared to the clinico-pathological model (AUC = 0.916, p = 0.006). Findings were supported by bootstrap re-sampling (n = 1000). CONCLUSION: Textural features from ADC maps and T2-W images differ between high- and low-volume tumors and potentially predict recurrence in low-volume tumors.

Growth Trajectories, Breast Size, and Breast-Tissue Composition in a British Prebirth Cohort of Young Women.

Mammographic percent density, the proportion of fibroglandular tissue in the breast, is a strong risk factor for breast cancer, but its determinants in young women are unknown. We examined associations of magnetic resonance imaging (MRI) breast-tissue composition at age 21 years with prospectively collected measurements of body size and composition from birth to early adulthood and markers of puberty (all standardized) in a sample of 500 nulliparous women from a prebirth cohort of children born in Avon, United Kingdom, in 1991-1992 and followed up to 2011-2014. Linear models were fitted to estimate relative change in MRI percent water, which is equivalent to mammographic percent density, associated with a 1-standard-deviation increase in the exposure of interest. In mutually adjusted analyses, MRI percent water was positively associated with birth weight (relative change (RC) = 1.03, 95% confidence interval (CI): 1.00, 1.06) and pubertal height growth (RC = 1.07, 95% CI: 1.02, 1.13) but inversely associated with pubertal weight growth (RC = 0.86, 95% CI: 0.84, 0.89) and changes in dual-energy x-ray absorptiometry percent body fat mass (e.g., for change between ages 11 years and 13.5 years, RC = 0.96, 95% CI: 0.93, 0.99). Ages at thelarche and menarche were positively associated with MRI percent water, but these associations did not persist upon adjustment for height and weight growth. These findings support the hypothesis that growth trajectories influence breast-tissue composition in young women, whereas puberty plays no independent role.

Extracranial Soft-Tissue Tumors: Repeatability of Apparent Diffusion Coefficient Estimates from Diffusion-weighted MR Imaging.

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.

Pre-natal exposures and breast tissue composition: findings from a British pre-birth cohort of young women and a systematic review.

BACKGROUND: Breast density, the amount of fibroglandular tissue in the adult breast for a women's age and body mass index, is a strong biomarker of susceptibility to breast cancer, which may, like breast cancer risk itself, be influenced by events early in life. In the present study, we investigated the association between pre-natal exposures and breast tissue composition. METHODS: A sample of 500 young, nulliparous women (aged approximately 21 years) from a U.K. pre-birth cohort underwent a magnetic resonance imaging examination of their breasts to estimate percent water, a measure of the relative amount of fibroglandular tissue equivalent to mammographic percent density. Information on pre-natal exposures was collected throughout the mothers' pregnancy and shortly after delivery. Regression models were used to investigate associations between percent water and pre-natal exposures. Mediation analysis, and a systematic review and meta-analysis of the published literature, were also conducted. RESULTS: Adjusted percent water in young women was positively associated with maternal height (p for linear trend [p t] = 0.005), maternal mammographic density in middle age (p t = 0.018) and the participant's birth size (p t < 0.001 for birthweight). A 1-SD increment in weight (473 g), length (2.3 cm), head circumference (1.2 cm) and Ponderal Index (4.1 g/cm3) at birth were associated with 3 % (95 % CI 2-5 %), 2 % (95 % CI 0-3 %), 3 % (95 % CI 1-4 %) and 1 % (95 % CI 0-3 %), respectively, increases in mean adjusted percent water. The effect of maternal height on the participants' percent water was partly mediated through birth size, but there was little evidence that the effect of birthweight was primarily mediated via adult body size. The meta-analysis supported the study findings, with breast density being positively associated with birth size. CONCLUSIONS: These findings provide strong evidence of pre-natal influences on breast tissue composition. The positive association between birth size and relative amount of fibroglandular tissue indicates that breast density and breast cancer risk may share a common pre-natal origin.

Evaluating the diagnostic sensitivity of computed diffusion-weighted MR imaging in the detection of breast cancer.

PURPOSE: To evaluate the diagnostic sensitivity of computed diffusion-weighted (DW)-MR imaging for the detection of breast cancer. MATERIALS AND METHODS: 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. RESULTS: 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). CONCLUSION: 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.

Repeatability and sensitivity of T2* measurements in patients with head and neck squamous cell carcinoma at 3T.

PURPOSE: To determine whether quantitation of T2* is sufficiently repeatable and sensitive to detect clinically relevant oxygenation levels in head and neck squamous cell carcinoma (HNSCC) at 3T. MATERIALS AND METHODS: Ten patients with newly diagnosed locally advanced HNSCC underwent two magnetic resonance imaging (MRI) scans between 24 and 168 hours apart prior to chemoradiotherapy treatment. A multiple gradient echo sequence was used to calculate T2* maps. A quadratic function was used to model the blood transverse relaxation rate as a function of blood oxygenation. A set of published coefficients measured at 3T were incorporated to account for tissue hematocrit levels and used to plot the dependence of fractional blood oxygenation (Y) on T2* values, together with the corresponding repeatability range. Repeatability of T2* using Bland-Altman analysis, and calculation of limits of agreement (LoA), was used to assess the sensitivity, defined as the minimum difference in fractional blood oxygenation that can be confidently detected. RESULTS: T2* LoA for 22 outlined tumor volumes were 13%. The T2* dependence of fractional blood oxygenation increases monotonically, resulting in increasing sensitivity of the method with increasing blood oxygenation. For fractional blood oxygenation values above 0.11, changes in T2* were sufficient to detect differences in blood oxygenation greater than 10% (Δ T2* > LoA for ΔY > 0.1). CONCLUSION: Quantitation of T2* at 3T can detect clinically relevant changes in tumor oxygenation within a wide range of blood volumes and oxygen tensions, including levels reported in HNSCC. J. Magn. Reson. Imaging 2016;44:72-80.

Detecting microvascular changes in the mouse spleen using optical computed tomography.

Methods of monitoring drug toxicity in off-target organs are very important in the development of effective and safe drugs. Standard 2-D techniques, such as histology, are prone to sampling errors and can miss important information. We demonstrate a novel application of optical computed tomography (CT) imaging to quantitatively assess, in 3-D, the response of adult murine spleen to off-target drug toxicity induced by treatment with the vascular disrupting agent ZD6126. Reconstructed images from optical CT scans sensitive to haemoglobin absorption reveal detailed, high-contrast 3-D maps of splenic structure and microvasculature. A significant difference in total splenic volume was found between vehicle and ZD6126-treated cohorts, with mean volumes of 61±3mm(3) and 44±3mm(3) respectively (both n=3, p=0.05). Textural statistics for each sample were calculated using grey-level co-occurrence matrices (GLCMs). Standard 2-D GLCM analysis was found to be slice-dependent while 3-D GLCM contrast and homogeneity analysis resulted in separation of the vehicle and ZD6126-treated cohorts over a range of length scales.

Isolated pulmonary artery choriocarcinoma masquerading as pulmonary embolism diagnosed by endovascular biopsy.

Pulmonary arteries may rarely be involved by primary and secondary tumors. Clinical and imaging features mimic those of PE making it challenging to diagnose. Choriocarcinoma is a malignant germ cell tumor, typically in the female genital tract. Rarely, they can present as PA thrombus. Female patients with a history of a molar pregnancy, ectopic pregnancy, abortion or in this case a miscarriage, are at a higher risk of gestational trophoblastic disease which can manifest in this way, albeit this is rare. In this report we describe the case of a 52-year-old female who presented with a 1 month history of worsening dyspnea and pleuritic lower thoracic pain. A diagnosis of pulmonary embolism (PE) was confirmed on CT pulmonary angiogram, with a large volume thrombus in the left pulmonary artery (PA). She failed to improve on standard anticoagulation therapy and was found to have a raised beta-human chorionic gonadotropin of >100,000. This leads to an extensive malignancy work-up. The only pertinent finding was that of increased fluorodeoxyglucose (FDG) accumulation in the PA thrombus. Endovascular biopsy of the thrombus was performed, and the patient was diagnosed with choriocarcinoma of the PA. This case highlights the importance of further investigation in patients failing to respond to anticoagulation therapy for PE. It also illustrates the role of interventional radiology in obtaining histological diagnosis in patient's presenting with PA tumor thrombus.

Towards 24/7 MRI: the effect of routine weekend inpatient MRI scanning on patient waiting times.

BACKGROUND: Demand for inpatient MRI outstrips capacity which results in long waiting lists. The hospital commenced a routine weekend MRI service in January 2023. AIM: The aim of this study was to investigate the effect of a limited routine weekend MRI service on MRI turnaround times. METHODS: Waiting times for inpatient MRI scans performed before and after the introduction of weekend MRI from January 1 to August 31, 2022, and January 1 to August 31, 2023, were obtained. The turnaround time (TAT) and request category for each study were calculated. Category 1 requests were required immediately, category 2 requests were urgent and category 3 requests were routine. RESULTS: There was a 6% (n = 128) increase in MRI inpatient scanning activity in 2023 (n = 2449) compared to 2022 (n = 2322). There was a significant improvement in overall mean TAT for inpatient MRIs (p < .001) in 2023 (mean 65.2 h, range 0-555 h) compared to 2022 (mean 98.3 h, range 0-816 h). There was no significant difference in the mean waiting time for category 1 MRIs between 2022 and 2023. There was a significant improvement (p < .001) in mean waiting time in 2023 (mean 37.2 h, range 0-555) compared to 2022 (mean 55.4 h, range 0-816) for category 2 MRI. The mean waiting time for category 3 studies also significantly improved (p < .001) in 2023 (mean 93.4 h, range 1-2663) when compared to 2022 (mean 154.8, range 1-1706). CONCLUSION: Routine weekend inpatient MRI significantly shortens inpatient waiting times.

A scan-specific quality control acquisition for clinical whole-body (WB) MRI protocols.

Objective.Image quality in whole-body MRI (WB-MRI) may be degraded by faulty radiofrequency (RF) coil elements or mispositioning of the coil arrays. Phantom-based quality control (QC) is used to identify broken RF coil elements but the frequency of these acquisitions is limited by scanner and staff availability. This work aimed to develop a scan-specific QC acquisition and processing pipeline to detect broken RF coil elements, which is sufficiently rapid to be added to the clinical WB-MRI protocol. The purpose of this is to improve the quality of WB-MRI by reducing the number of patient examinations conducted with suboptimal equipment.Approach.A rapid acquisition (14 s additional acquisition time per imaging station) was developed that identifies broken RF coil elements by acquiring images from each individual coil element and using the integral body coil. This acquisition was added to one centre's clinical WB-MRI protocol for one year (892 examinations) to evaluate the effect of this scan-specific QC. To demonstrate applicability in multi-centre imaging trials, the technique was also implemented on scanners from three manufacturers.Main results. Over the course of the study RF coil elements were flagged as potentially broken on five occasions, with the faults confirmed in four of those cases. The method had a precision of 80% and a recall of 100% for detecting faulty RF coil elements. The coil array positioning measurements were consistent across scanners and have been used to define the expected variation in signal.Significance. The technique demonstrated here can identify faulty RF coil elements and positioning errors and is a practical addition to the clinical WB-MRI protocol. This approach was fully implemented on systems from two manufacturers and partially implemented on a third. It has potential to reduce the number of clinical examinations conducted with suboptimal hardware and improve image quality across multi-centre studies.

Curation of myeloma observational study MALIMAR using XNAT: solving the challenges posed by real-world data.

OBJECTIVES: MAchine Learning In MyelomA Response (MALIMAR) is an observational clinical study combining "real-world" and clinical trial data, both retrospective and prospective. Images were acquired on three MRI scanners over a 10-year window at two institutions, leading to a need for extensive curation. METHODS: Curation involved image aggregation, pseudonymisation, allocation between project phases, data cleaning, upload to an XNAT repository visible from multiple sites, annotation, incorporation of machine learning research outputs and quality assurance using programmatic methods. RESULTS: A total of 796 whole-body MR imaging sessions from 462 subjects were curated. A major change in scan protocol part way through the retrospective window meant that approximately 30% of available imaging sessions had properties that differed significantly from the remainder of the data. Issues were found with a vendor-supplied clinical algorithm for "composing" whole-body images from multiple imaging stations. Historic weaknesses in a digital video disk (DVD) research archive (already addressed by the mid-2010s) were highlighted by incomplete datasets, some of which could not be completely recovered. The final dataset contained 736 imaging sessions for 432 subjects. Software was written to clean and harmonise data. Implications for the subsequent machine learning activity are considered. CONCLUSIONS: MALIMAR exemplifies the vital role that curation plays in machine learning studies that use real-world data. A research repository such as XNAT facilitates day-to-day management, ensures robustness and consistency and enhances the value of the final dataset. The types of process described here will be vital for future large-scale multi-institutional and multi-national imaging projects. CRITICAL RELEVANCE STATEMENT: This article showcases innovative data curation methods using a state-of-the-art image repository platform; such tools will be vital for managing the large multi-institutional datasets required to train and validate generalisable ML algorithms and future foundation models in medical imaging. KEY POINTS: • Heterogeneous data in the MALIMAR study required the development of novel curation strategies. • Correction of multiple problems affecting the real-world data was successful, but implications for machine learning are still being evaluated. • Modern image repositories have rich application programming interfaces enabling data enrichment and programmatic QA, making them much more than simple "image marts".

Validated machine learning tools to distinguish immune checkpoint inhibitor, radiotherapy, COVID-19 and other infective pneumonitis.

BACKGROUND: Pneumonitis is a well-described, potentially disabling, or fatal adverse effect associated with both immune checkpoint inhibitors (ICI) and thoracic radiotherapy. Accurate differentiation between checkpoint inhibitor pneumonitis (CIP) radiation pneumonitis (RP), and infective pneumonitis (IP) is crucial for swift, appropriate, and tailored management to achieve optimal patient outcomes. However, correct diagnosis is often challenging, owing to overlapping clinical presentations and radiological patterns. METHODS: In this multi-centre study of 455 patients, we used machine learning with radiomic features extracted from chest CT imaging to develop and validate five models to distinguish CIP and RP from COVID-19, non-COVID-19 infective pneumonitis, and each other. Model performance was compared to that of two radiologists. RESULTS: Models to distinguish RP from COVID-19, CIP from COVID-19 and CIP from non-COVID-19 IP out-performed radiologists (test set AUCs of 0.92 vs 0.8 and 0.8; 0.68 vs 0.43 and 0.4; 0.71 vs 0.55 and 0.63 respectively). Models to distinguish RP from non-COVID-19 IP and CIP from RP were not superior to radiologists but demonstrated modest performance, with test set AUCs of 0.81 and 0.8 respectively. The CIP vs RP model performed less well on patients with prior exposure to both ICI and radiotherapy (AUC 0.54), though the radiologists also had difficulty distinguishing this test cohort (AUC values 0.6 and 0.6). CONCLUSION: Our results demonstrate the potential utility of such tools as a second or concurrent reader to support oncologists, radiologists, and chest physicians in cases of diagnostic uncertainty. Further research is required for patients with exposure to both ICI and thoracic radiotherapy.

Relationship between hippocampal subfield volumes and cognitive decline in healthy subjects.

We examined the relationship between hippocampal subfield volumes and cognitive decline over a 4-year period in a healthy older adult population with the goal of identifying subjects at risk of progressive cognitive impairment which could potentially guide therapeutic interventions and monitoring. 482 subjects (68.1 years +/- 7.4; 52.9% female) from the Irish Longitudinal Study on Ageing underwent magnetic resonance brain imaging and a series of cognitive tests. Using K-means longitudinal clustering, subjects were first grouped into three separate global and domain-specific cognitive function trajectories; High-Stable, Mid-Stable and Low-Declining. Linear mixed effects models were then used to establish associations between hippocampal subfield volumes and cognitive groups. Decline in multiple hippocampal subfields was associated with global cognitive decline, specifically the presubiculum (estimate -0.20; 95% confidence interval (CI) -0.78 - -0.02; p = 0.03), subiculum (-0.44; -0.82 - -0.06; p = 0.02), CA1 (-0.34; -0.78 - -0.02; p = 0.04), CA4 (-0.55; -0.93 - -0.17; p = 0.005), molecular layer (-0.49; -0.87 - -0.11; p = 0.01), dentate gyrus (-0.57; -0.94 - -0.19; p = 0.003), hippocampal tail (-0.53; -0.91 - -0.15; p = 0.006) and HATA (-0.41; -0.79 - -0.03; p = 0.04), with smaller volumes for the Low-Declining cognition group compared to the High-Stable cognition group. In contrast to global cognitive decline, when specifically assessing the memory domain, cornu ammonis 1 subfield was not found to be associated with low declining cognition (-0.14; -0.37 - 0.10; p = 0.26). Previously published data shows that atrophy of specific hippocampal subfields is associated with cognitive decline but our study confirms the same effect in subjects asymptomatic at time of enrolment. This strengthens the predictive value of hippocampal subfield atrophy in risk of cognitive decline and may provide a biomarker for monitoring treatment efficacy.

Integrating Artificial Intelligence Tools in the Clinical Research Setting: The Ovarian Cancer Use Case.

Artificial intelligence (AI) methods applied to healthcare problems have shown enormous potential to alleviate the burden of health services worldwide and to improve the accuracy and reproducibility of predictions. In particular, developments in computer vision are creating a paradigm shift in the analysis of radiological images, where AI tools are already capable of automatically detecting and precisely delineating tumours. However, such tools are generally developed in technical departments that continue to be siloed from where the real benefit would be achieved with their usage. Significant effort still needs to be made to make these advancements available, first in academic clinical research and ultimately in the clinical setting. In this paper, we demonstrate a prototype pipeline based entirely on open-source software and free of cost to bridge this gap, simplifying the integration of tools and models developed within the AI community into the clinical research setting, ensuring an accessible platform with visualisation applications that allow end-users such as radiologists to view and interact with the outcome of these AI tools.

Interpretability of radiomics models is improved when using feature group selection strategies for predicting molecular and clinical targets in clear-cell renal cell carcinoma: insights from the TRACERx Renal study.

BACKGROUND: The aim of this work is to evaluate the performance of radiomics predictions for a range of molecular, genomic and clinical targets in patients with clear cell renal cell carcinoma (ccRCC) and demonstrate the impact of novel feature selection strategies and sub-segmentations on model interpretability. METHODS: Contrast-enhanced CT scans from the first 101 patients recruited to the TRACERx Renal Cancer study (NCT03226886) were used to derive radiomics classification models to predict 20 molecular, histopathology and clinical target variables. Manual 3D segmentation was used in conjunction with automatic sub-segmentation to generate radiomics features from the core, rim, high and low enhancing sub-regions, and the whole tumour. Comparisons were made between two classification model pipelines: a Conventional pipeline reflecting common radiomics practice, and a Proposed pipeline including two novel feature selection steps designed to improve model interpretability. For both pipelines nested cross-validation was used to estimate prediction performance and tune model hyper-parameters, and permutation testing was used to evaluate the statistical significance of the estimated performance measures. Further model robustness assessments were conducted by evaluating model variability across the cross-validation folds. RESULTS: Classification performance was significant (p < 0.05, H0:AUROC = 0.5) for 11 of 20 targets using either pipeline and for these targets the AUROCs were within ± 0.05 for the two pipelines, except for one target where the Proposed pipeline performance increased by > 0.1. Five of these targets (necrosis on histology, presence of renal vein invasion, overall histological stage, linear evolutionary subtype and loss of 9p21.3 somatic alteration marker) had AUROC > 0.8. Models derived using the Proposed pipeline contained fewer feature groups than the Conventional pipeline, leading to more straightforward model interpretations without loss of performance. Sub-segmentations lead to improved performance and/or improved interpretability when predicting the presence of sarcomatoid differentiation and tumour stage. CONCLUSIONS: Use of the Proposed pipeline, which includes the novel feature selection methods, leads to more interpretable models without compromising prediction performance. TRIAL REGISTRATION: NCT03226886 (TRACERx Renal).

Water and tissue equivalence of a new PRESAGE(®) formulation for 3D proton beam dosimetry: a Monte Carlo study.

PURPOSE: To evaluate the water and tissue equivalence of a new PRESAGE(®) 3D dosimeter for proton therapy. METHODS: The GEANT4 software toolkit was used to calculate and compare total dose delivered by a proton beam with mean energy 62 MeV in a PRESAGE(®) dosimeter, water, and soft tissue. The dose delivered by primary protons and secondary particles was calculated. Depth-dose profiles and isodose contours of deposited energy were compared for the materials of interest. RESULTS: The proton beam range was found to be ≈27 mm for PRESAGE(®), 29.9 mm for soft tissue, and 30.5 mm for water. This can be attributed to the lower collisional stopping power of water compared to soft tissue and PRESAGE(®). The difference between total dose delivered in PRESAGE(®) and total dose delivered in water or tissue is less than 2% across the entire water∕tissue equivalent range of the proton beam. The largest difference between total dose in PRESAGE(®) and total dose in water is 1.4%, while for soft tissue it is 1.8%. In both cases, this occurs at the distal end of the beam. Nevertheless, the authors find that PRESAGE(®) dosimeter is overall more tissue-equivalent than water-equivalent before the Bragg peak. After the Bragg peak, the differences in the depth doses are found to be due to differences in primary proton energy deposition; PRESAGE(®) and soft tissue stop protons more rapidly than water. The dose delivered by secondary electrons in the PRESAGE(®) differs by less than 1% from that in soft tissue and water. The contribution of secondary particles to the total dose is less than 4% for electrons and ≈1% for protons in all the materials of interest. CONCLUSIONS: These results demonstrate that the new PRESAGE(®) formula may be considered both a tissue- and water-equivalent 3D dosimeter for a 62 MeV proton beam. The results further suggest that tissue-equivalent thickness may provide better dosimetric and geometric accuracy than water-equivalent thickness for 3D dosimetry of this proton beam.