Multiparametric MRI Data Research Articles

Exploring the role of multimodal [18F]F-PSMA-1007 PET/CT and multiparametric MRI data in predicting ISUP grading of primary prostate cancer.

The study explores the role of multimodal imaging techniques, such as [18F]F-PSMA-1007 PET/CT and multiparametric MRI (mpMRI), in predicting the ISUP (International Society of Urological Pathology) grading of prostate cancer. The goal is to enhance diagnostic accuracy and improve clinical decision-making by integrating these advanced imaging modalities with clinical variables. In particular, the study investigates the application of few-shot learning to address the challenge of limited data in prostate cancer imaging, which is often a common issue in medical research. This study conducted a retrospective analysis of 341 prostate cancer patients enrolled between 2019 and 2023, with data collected from five imaging modalities: [18F]F-PSMA-1007 PET, CT, Diffusion Weighted Imaging (DWI), T2 Weighted Imaging (T2WI), and Apparent Diffusion Coefficient (ADC). The study compared the performance of five single-modality data sets, PET/CT dual-modality fusion data, mpMRI tri-modality fusion data, and five-modality fusion data within deep learning networks, analyzing how different modalities impact the accuracy of ISUP grading prediction. To address the issue of limited data, a few-shot deep learning network was employed, enabling training and cross-validation with only a small set of labeled samples. Additionally, the results were compared with those from preoperative biopsies and clinical prediction models to further assess the reliability of the experimental findings. The experimental results demonstrate that the multimodal model (combining [18F]F-PSMA-1007 PET/CT and multiparametric MRI) significantly outperforms other models in predicting ISUP grading of prostate cancer. Meanwhile, both the PET/CT dual-modality and mpMRI tri-modality models outperform the single-modality model, with comparable performance between the two multimodal models. Furthermore, the experimental data confirm that the few-shot learning network introduced in this study provides reliable predictions, even with limited data. This study highlights the potential of applying multimodal imaging techniques (such as PET/CT and mpMRI) in predicting ISUP grading of prostate cancer. The findings suggest that this integrated approach can enhance the accuracy of prostate cancer diagnosis and contribute to more personalized treatment planning. Furthermore, incorporating few-shot learning into the model development process allows for more robust predictions despite limited data, making this approach highly valuable in clinical settings with sparse data.

NIMG-55. DEVELOPING A COMPUTATIONAL PIPELINE TO FORECAST AND SPATIALLY MAP HIGH-GRADE GLIOMA RESPONSE TO RADIOTHERAPY USING BIOPHYSICAL MODELS AND DATA-DRIVEN STATISTICAL ANALYSIS

Abstract High-grade glioma (HGG) presents significant treatment challenges, particularly in identifying less responsive tumor areas to target with adaptive radiotherapy (RT). This study introduces a patient-specific computational pipeline designed to predict and spatially map HGG response to RT, focusing on regions demonstrating resilience to RT. Our computational pipeline is centered around a family of 3D biophysical models describing tumor cell invasion, proliferation, and response to RT which can be personalized for individual patients using longitudinal multiparametric MRI data. Utilizing longitudinal multiparametric MRI data reporting on cell density (diffusion weighted MRI) and the extent of disease (T2-FLAIR and contrast-enhanced T1-weighted MRIs) from 21 patients, ranging from baseline to three weeks into RT, we calibrated a family of biophysical models to forecast tumor cell density up to the final week of RT and one-month follow-up. By employing pareto tail analysis on empirical cumulative distribution function fits of voxel-wise changes in apparent diffusion coefficient maps (ΔADC) between the future images (end of RT, one-month follow-up) and pre-treatment images, we identified areas of increased tumor cellularity on both the ground truth and model predicted cellularity maps. The analysis differentiated tumor voxels into “non-responding” and “responding” categories, based on cutoffs determined through ROC analysis across various conditions—including assessments of both enhancing and non-enhancing tumor regions at different time points. Optimal ΔADC cutoffs for predicted and measured data were established, leading to high area under curve (AUC; ranging from 0.803 to 0.934) values and significant Youden’s index scores (ranging from 0.526 to 0.765) across all conditions. Our findings demonstrate the feasibility of this novel pipeline in delineating biologically relevant RT targets, offering a potential tool for enhancing the efficacy of adaptive RT strategies in treating HGG.

Automatic pipeline for segmentation of LV myocardium on quantitative MR T1 maps using deep learning model and computation of radial T1 and ECV values.

Native T1 mapping is a non-invasive technique used for early detection of diffused myocardial abnormalities, and it provides baseline tissue characterization. Post-contrast T1 mapping enhances tissue differentiation, enables extracellular volume (ECV) calculation, and improves myocardial viability assessment. Accurate and precise segmenting of the left ventricular (LV) myocardium on T1 maps is crucial for assessing myocardial tissue characteristics and diagnosing cardiovascular diseases (CVD). This study presents a deep learning (DL)-based pipeline for automatically segmenting LV myocardium on T1 maps and automatic computation of radial T1 and ECV values. The study employs a multicentric dataset consisting of retrospective multiparametric MRI data of 332 subjects to develop and assess the performance of the proposed method. The study compared DL architectures U-Net and Deep Res U-Net for LV myocardium segmentation, which achieved a dice similarity coefficient of 0.84 ±0.43 and 0.85 ±0.03, respectively. The dice similarity coefficients computed for radial sub-segmentation of the LV myocardium on basal, mid-cavity, and apical slices were 0.77 ±0.21, 0.81 ±0.17, and 0.61 ±0.14, respectively. The t-test performed between ground truth vs. predicted values of native T1, post-contrast T1, and ECV showed no statistically significant difference (p> 0.05) for any of the radial sub-segments. The proposed DL method leverages the use of quantitative T1 maps for automatic LV myocardium segmentation and accurately computing radial T1 and ECV values, highlighting its potential for assisting radiologists in objective cardiac assessment and, hence, in CVD diagnostics.

Phi and phiD predict adverse pathological features after radical prostatectomy for prostate cancer in Chinese population.

Anticipating the postoperative pathological stage and potential for adverse features of prostate cancer (PCa) patients before radical prostatectomy (RP) is crucial for guiding perioperative treatment. A cohort consisting of three sub-cohorts with a total of 709 patients has been enlisted from two major tertiary medical centres in China. The primary assessment parameters for adverse pathological features in this study are the pathological T stage, the AJCC prognostic stage groups and perineural invasion (PNI). Logistic regression analyses were performed to investigate the relationship between prostate specific antigen (PSA), its derivatives (incluing Prostate Health Index, phi and phi density, phiD), and the pathological outcomes after RP. Both phi and phiD showed a significant association with pathologic T stage of pT3 or above (phi, adjusted OR, AOR = 2.82, 95% confidence interval, 95% CI: 1.88-4.23, p < 0.001; phiD, AOR = 2.47, 95% CI: 1.76-3.48, p < 0.001) and PNI (phi, AOR = 2.15, 95% CI: 1.39-3.32, p < 0.001; phiD, AOR = 1.94, 95% CI: 1.38-2.73, p < 0.001). In a subgroup analysis with a total PSA value <10 ng/mL, phi and phiD continued to show a significant correlation with pT3 or above (phi, AOR = 4.70, 95% CI: 1.29-17.12, p = 0.019; phiD, AOR = 3.44, 95% CI: 1.51-7.85, p = 0.003), and phiD also maintained its predictive capability for PNI in this subgroup (AOR = 2.10, 95% CI: 1.17-3.80, p = 0.014). Sensitivity analysis indicated that the findings in the combined cohort are mainly influenced by one of the sub-cohorts, partially attributable to disparities in sample sizes between sub-cohorts. Combined analysis of phi(D) and multiparametric MRI (mpMRI) data yielded similar results. Preoperative measurement of serum phi and phiD is valuable for predicting the occurrence of adverse pathological features in Chinese PCa patients after RP.

Application of machine learning in the analysis of multiparametric MRI data for the differentiation of treatment responses in breast cancer: retrospective study.

The objective of this study is to develop and validate a multiparametric MRI model employing machine learning to predict the effectiveness of treatment and the stage of breast cancer. The study encompassed 400 female patients diagnosed with breast cancer, with 200 individuals allocated to both the control and experimental groups, undergoing examinations in Shenzhen, China, during the period 2017-2023. This study pertains to retrospective research. Multiparametric MRI was employed to extract data concerning tumor size, blood flow, and metabolism. The model achieved high accuracy, predicting treatment outcomes with an accuracy of 92%, sensitivity of 88%, and specificity of 95%. The model effectively classified breast cancer stages: stage I, 38% ( P = 0.027); stage II, 72% ( P = 0.014); stage III, 50% ( P = 0.032); and stage IV, 45% ( P = 0.041). The developed model, utilizing multiparametric MRI and machine learning, exhibits high accuracy in predicting the effectiveness of treatment and breast cancer staging. These findings affirm the model's potential to enhance treatment strategies and personalize approaches for patients diagnosed with breast cancer. Our study presents an innovative approach to the diagnosis and treatment of breast cancer, integrating MRI data with machine learning algorithms. We demonstrate that the developed model exhibits high accuracy in predicting treatment efficacy and differentiating cancer stages. This underscores the importance of utilizing MRI and machine learning algorithms to enhance the diagnosis and individualization of treatment for this disease.

MDB-94. RADIOMICS-BASED DECISION SUPPORT FOR PEDIATRIC MEDULLOBLASTOMA RISK STRATIFICATION ON MULTI-PARAMETRIC MRI

Abstract BACKGROUND Medulloblastoma (MB) is the most common pediatric brain malignancy. Accurate risk stratification is crucial for prescribing appropriate therapy. However, current risk stratification into standard or high-risk groups is limited in scope and does not accurately address the unique biological behavior of each specific tumor, with approximately 30% of children diagnosed with MB suffering from treatment failure. We developed a radiomic tool to more accurately risk stratify children with MB. METHODS We collected 1.5T/3T MRI (Siemens, Philips, or GE scanners) and clinical data from 80 patients with MB treated at Children’s Healthcare of Atlanta. Radiomic features that quantify textural heterogeneity including first and second order statistics, Haralick’s features from Gray Level Co-occurrence Matrices, and CoLIAge were extracted from tumor volumes segmented by a board-certified radiologist from multi-parametric MRI data including Apparent Diffusion Coefficient (ADC), T2, and post-contrast T1 sequences. Continuous survival models were built using the Cox proportional hazard method with 70% of the data for progression free survival (PFS) and overall survival (OS) using the 5 most predictive features obtained by univariate analysis after removing correlated features within the training data. RESULTS We developed three models with T1, T2, and ADC separately to evaluate predictive power of radiomic signatures. Three models for predicting MB OS delivered Harrel’s concordance indices (C-index) of 0.72 and 0.61 (T1); 0.70 and 0.65 (T2); and 0.80 and 0.76 (ADC) for training and test, respectively. C-indices for PFS were 0.70 and 0.69 (T1); 0.73 and 0.67 (T2); and 0.63 and 0.61 (ADC). CONCLUSIONS Our preliminary results show the feasibility and efficacy of our approach for predicting MB OS and PFS. Future studies will build an integrated model combining T1,T2, ADC, and other clinical information for predicting MB OS and PFS using a large multi-center validation cohort.

El análisis cuantitativo de las imágenes potenciadas en difusión (DWI) mejora la precisión diagnóstica del sistema de puntuación O-RADS RM en lesiones anexiales indeterminadas

IntroductionDespite the overall high accuracy of the O-RADS MRI scoring system for characterization of indeterminate adnexal lesions, a non-negligible percentage of adnexal lesions remains indeterminate. Given this, the main objective of this study was to evaluate the value of adding quantitative diffusion-weighted imaging (DWI) to Ovarian-Adnexal Reporting and Data System (O-RADS) MRI to improve the diagnostic accuracy of the scoring system in a cohort of indeterminate adnexal lesions according to International Ovarian Tumor Analysis Group Simple Rules (IOTA-SR). Methods and materialSeventy-nine women with 81 pelvic lesions classified as indeterminate according to IOTA-SR underwent 3-Tesla MRI with a conventional multiparametric protocol. DWI was quantitatively analyzed. Lesions were surgically removed or followed-up, according to a local protocol. Sensitivity, specificity, positive and negative predictive values and accuracy were determined for conventional multiparametric MRI and quantitative DWI-derived data. ResultsTwenty masses in 72 patients (27.8%) were malignant. An apparent diffusion coefficient (ADC) cut-off value of 1.30×10-3mm2/s had 89% sensitivity and 80% specificity for malignancy. Overall, adding quantitative DWI to O-RADS MRI increased the specificity (98.08%, P<.001), positive predictive value (94.12%, P<.001), and accuracy (93.06%, P=.05). In the specific O-RADS MRI score 4 subgroup, an ADC cut-off value of 1.22×10-3 mm2/s had 86% sensitivity and 67% specificity for distinguishing benign from malignant lesions. ConclusionIn IOTA-SR indeterminate adnexal lesions, quantitative DWI significantly improves the diagnostic performance of conventional multiparametric MRI in all O-RADS MRI score groups.

Predicting pathological complete response based on weakly and semi-supervised joint learning in breast cancer multi-parametric MRI

Neoadjuvant chemotherapy (NAC) is the primary treatment used to reduce the tumor size in early breast cancer. Patients who achieve a pathological complete response (pCR) after NAC treatment have a significantly higher five-year survival rate. However, accurately predicting whether patients could achieve pCR remains challenging due to the limited availability of manually annotated MRI data. This study develops a weakly and semi-supervised joint learning model that integrates multi-parametric MR images to predict pCR to NAC in breast cancer patients. First, the attention-based multi-instance learning model is designed to characterize the representation of multi-parametric MR images in a weakly supervised learning setting. The Mean-Teacher learning framework is then developed to locate tumor regions for extracting radiochemical parameters in a semi-supervised learning setting. Finally, all extracted MR imaging features are fused to predict pCR to NAC. Our experiments were conducted on a cohort of 442 patients with multi-parametric MR images and NAC outcomes. The results demonstrate that our proposed model, which leverages multi-parametric MRI data, provides the AUC value of over 0.85 in predicting pCR to NAC, outperforming other comparative methods.

Quantifying U-Net uncertainty in multi-parametric MRI-based glioma segmentation by spherical image projection.

Uncertainty quantification in deep learning is an important research topic. For medical image segmentation, the uncertainty measurements are usually reported as the likelihood that each pixel belongs to the predicted segmentation region. In potential clinical applications, the uncertainty result reflects the algorithm's robustness and supports the confidence and trust of the segmentation result when the ground-truth result is absent. For commonly studied deep learning models, novel methods for quantifying segmentation uncertainty are in demand. To develop a U-Net segmentation uncertainty quantification method based on spherical image projection of multi-parametric MRI (MP-MRI) in glioma segmentation. The projection of planar MRI data onto a spherical surface is equivalent to a nonlinear image transformation that retains global anatomical information. By incorporating this image transformation process in our proposed spherical projection-based U-Net (SPU-Net) segmentation model design, multiple independent segmentation predictions can be obtained from a single MRI. The final segmentation is the average of all available results, and the variation can be visualized as a pixel-wise uncertainty map. An uncertainty score was introduced to evaluate and compare the performance of uncertainty measurements. The proposed SPU-Net model was implemented on the basis of 369 glioma patients with MP-MRI scans (T1, T1-Ce, T2, and FLAIR). Three SPU-Net models were trained to segment enhancing tumor (ET), tumor core (TC), and whole tumor (WT), respectively. The SPU-Net model was compared with (1) the classic U-Net model with test-time augmentation (TTA) and (2) linear scaling-based U-Net (LSU-Net) segmentation models in terms of both segmentation accuracy (Dice coefficient, sensitivity, specificity, and accuracy) and segmentation uncertainty (uncertainty map and uncertainty score). The developed SPU-Net model successfully achieved low uncertainty for correct segmentation predictions (e.g., tumor interior or healthy tissue interior) and high uncertainty for incorrect results (e.g., tumor boundaries). This model could allow the identification of missed tumor targets or segmentation errors in U-Net. Quantitatively, the SPU-Net model achieved the highest uncertainty scores for three segmentation targets (ET/TC/WT): 0.826/0.848/0.936, compared to 0.784/0.643/0.872 using the U-Net with TTA and 0.743/0.702/0.876 with the LSU-Net (scaling factor=2). The SPU-Net also achieved statistically significantly higher Dice coefficients, underscoring the improved segmentation accuracy. The SPU-Net model offers a powerful tool to quantify glioma segmentation uncertainty while improving segmentation accuracy. The proposed method can be generalized to other medical image-related deep-learning applications for uncertainty evaluation.

A novel MRI-based deep learning networks combined with attention mechanism for predicting CDKN2A/B homozygous deletion status in IDH-mutant astrocytoma.

To develop a high-accuracy MRI-based deep learning method for predicting cyclin-dependent kinase inhibitor 2A/B (CDKN2A/B) homozygous deletion status in isocitrate dehydrogenase (IDH)-mutant astrocytoma. Multiparametric brain MRI data and corresponding genomic information of 234 subjects (111 positives for CDKN2A/B homozygous deletion and 123 negatives for CDKN2A/B homozygous deletion) were obtained from The Cancer Imaging Archive (TCIA) and The Cancer Genome Atlas (TCGA) respectively. Two independent multi-sequence networks (ResFN-Net and FN-Net) are built on the basis of ResNet and ConvNeXt network combined with attention mechanism to classify CDKN2A/B homozygous deletion status using MR images including contrast-enhanced T1-weighted imaging (CE-T1WI) and T2-weighted imaging (T2WI). The performance of the network is summarized by three-way cross-validation; ROC analysis is also performed. The average cross-validation accuracy (ACC) of ResFN-Net is 0.813. The average cross-validation area under curve (AUC) of ResFN-Net is 0.8804. The average cross-validation ACC and AUC of FN-Net is 0.9236 and 0.9704, respectively. Comparing all sequence combinations of the two networks (ResFN-Net and FN-Net), the sequence combination of CE-T1WI and T2WI performed the best, and the ACC and AUC were 0.8244, 0.8975 and 0.8971, 0.9574, respectively. The FN-Net deep learning networks based on ConvNeXt network achieved promising performance for predicting CDKN2A/B homozygous deletion status of IDH-mutant astrocytoma. A novel deep learning network (FN-Net) based on preoperative MRI was developed to predict the CDKN2A/B homozygous deletion status. This network has the potential to be a practical tool for the noninvasive characterization of CDKN2A/B in glioma to support personalized classification and treatment planning. • CDKN2A/B homozygous deletion status is an important marker for glioma grading and prognosis. • An MRI-based deep learning approach was developed to predict CDKN2A/B homozygous deletion status. • The predictive performance based on ConvNeXt network was better than that of ResNet network.

7-LB: Sodium–Glucose Cotransporter-2 Inhibitors (SGLT2i) Decrease Kidney Oxygen Consumption in Adults with Type 2 Diabetes (T2D)—A Randomized Clinical Trial Using 11C-Acetate Positron Emission Tomography (PET) Imaging

Kidney hypoxia has been proposed as key pathophysiological mechanism in the development of chronic kidney disease, potentially stemming from a mismatch between oxygen delivery and oxygen consumption. SGLT2i are kidney protective drugs that initially lower glomerular filtration rate (GFR). Accordingly, we hypothesized that SGLT2i lower kidney oxygen consumption by reducing GFR and associated tubular workload. Twenty adults with T2D (sex 80% male, age 68±6 years, BMI 30±4 kg/m2, HbA1c 7.5±0.9%, eGFR 76 ± 11 mL/min/1.73m2) received a 4-week treatment with SGLT2i ertugliflozin (ERTU) and matched placebo (PLB) in a randomized, double-blind cross-over study. Participants were treated with metformin and received maximal tolerable dose of an angiotensin receptor blocker. Whole-kidney oxygen consumption (Kmono) was measured by PET using 11C-acetate. GFR was measured by gold-standard iohexol clearance. Tubular sodium transport (TNa) was calculated by kidney sodium load (([arterial Na]*mGFR) - urinary sodium excretion). Kidney efficiency calculated as TNa/ Kmono. GFR was lower during ERTU (94 ± 14 mL/min) vs. PLB (99 ± 15 mL/min) treatment (p=0.02). Kmono was 0.086 ± 0.006 min-1 during ERTU and 0.091 ± 0.009 min-1 during PLB (p&amp;lt;0.01). Kidney sodium load (12.9 ± 1.90 vs 13.7 ± 2.22 mmol/min) and TNa (12.7 ± 1.87 vs 13.6 ± 2.18 mmol/min) were lower during ERTU (p=0.02 for both), while urinary sodium excretion (p=0.1) and TNa/ Kmono remained unchanged (p=0.7). TNa strongly associated with Kmono (r=0.50; p&amp;lt;0.0001). In conclusion, SGLT2i lowers mGFR, TNa, and kidney oxygen consumption, potentially decreasing hypoxia risk. SGLT2i treatment did not lower overall kidney efficiency, despite the ATP efficiency of cortical sodium transport. Additional data on cortex and medulla-specific oxygen consumption and multiparametric kidney MRI data will be presented at ADA. Disclosure A. Hesp: None. L. I. P. Snel: None. P. Schober: None. L. A. Schwarte: None. R. Boellaard: None. P. Bjornstad: Advisory Panel; AstraZeneca. Consultant; Bayer Inc., Bristol-Myers Squibb Company. Advisory Panel; Novo Nordisk, Lilly, Horizon Therapeutics plc, Boehringer Ingelheim (Canada) Ltd., LG Chem. D. van Raalte: Consultant; Boehringer Ingelheim and Eli Lilly Alliance, AstraZeneca, Merck &amp; Co., Inc. Research Support; Boehringer Ingelheim and Eli Lilly Alliance, AstraZeneca, Merck &amp; Co., Inc. Funding MSD

Research progress on deep learning in magnetic resonance imaging-based diagnosis and treatment of prostate cancer: a review on the current status and perspectives.

Multiparametric magnetic resonance imaging (mpMRI) has emerged as a first-line screening and diagnostic tool for prostate cancer, aiding in treatment selection and noninvasive radiotherapy guidance. However, the manual interpretation of MRI data is challenging and time-consuming, which may impact sensitivity and specificity. With recent technological advances, artificial intelligence (AI) in the form of computer-aided diagnosis (CAD) based on MRI data has been applied to prostate cancer diagnosis and treatment. Among AI techniques, deep learning involving convolutional neural networks contributes to detection, segmentation, scoring, grading, and prognostic evaluation of prostate cancer. CAD systems have automatic operation, rapid processing, and accuracy, incorporating multiple sequences of multiparametric MRI data of the prostate gland into the deep learning model. Thus, they have become a research direction of great interest, especially in smart healthcare. This review highlights the current progress of deep learning technology in MRI-based diagnosis and treatment of prostate cancer. The key elements of deep learning-based MRI image processing in CAD systems and radiotherapy of prostate cancer are briefly described, making it understandable not only for radiologists but also for general physicians without specialized imaging interpretation training. Deep learning technology enables lesion identification, detection, and segmentation, grading and scoring of prostate cancer, and prediction of postoperative recurrence and prognostic outcomes. The diagnostic accuracy of deep learning can be improved by optimizing models and algorithms, expanding medical database resources, and combining multi-omics data and comprehensive analysis of various morphological data. Deep learning has the potential to become the key diagnostic method in prostate cancer diagnosis and treatment in the future.

Deep learning referral suggestion and tumour discrimination using explainable artificial intelligence applied to multiparametric MRI.

An appropriate and fast clinical referral suggestion is important for intra-axial mass-like lesions (IMLLs) in the emergency setting. We aimed to apply an interpretable deep learning (DL) system to multiparametric MRI to obtain clinical referral suggestion for IMLLs, and to validate it in the setting of nontraumatic emergency neuroradiology. A DL system was developed in 747 patients with IMLLs ranging 30 diseases who underwent pre- and post-contrast T1-weighted (T1CE), FLAIR, and diffusion-weighted imaging (DWI). A DL system that segments IMLLs, classifies tumourous conditions, and suggests clinical referral among surgery, systematic work-up, medical treatment, and conservative treatment, was developed. The system was validated in an independent cohort of 130 emergency patients, and performance in referral suggestion and tumour discrimination was compared with that of radiologists using receiver operating characteristics curve, precision-recall curve analysis, and confusion matrices. Multiparametric interpretable visualisation of high-relevance regions from layer-wise relevance propagation overlaid on contrast-enhanced T1WI and DWI was analysed. The DL system provided correct referral suggestions in 94 of 130 patients (72.3%) and performed comparably to radiologists (accuracy 72.6%, McNemar test; p = .942). For distinguishing tumours from non-tumourous conditions, the DL system (AUC, 0.90 and AUPRC, 0.94) performed similarly to human readers (AUC, 0.81~0.92, and AUPRC, 0.88~0.95). Solid portions of tumours showed a high overlap of relevance, but non-tumours did not (Dice coefficient 0.77 vs. 0.33, p < .001), demonstrating the DL's decision. Our DL system could appropriately triage patients using multiparametric MRI and provide interpretability through multiparametric heatmaps, and may thereby aid neuroradiologic diagnoses in emergency settings. Our AI triages patients with raw MRI images to clinical referral pathways in brain intra-axial mass-like lesions. We demonstrate that the decision is based on the relative relevance between contrast-enhanced T1-weighted and diffusion-weighted images, providing explainability across multiparametric MRI data. • A deep learning (DL) system using multiparametric MRI suggested clinical referral to patients with intra-axial mass-like lesions (IMLLs) similar to radiologists (accuracy 72.3% vs. 72.6%). • In the differentiation of tumourousand non-tumourousconditions, the DL system (AUC, 0.90) performed similar with radiologists (AUC, 0.81-0.92). • The DL's decision basis for differentiating tumours from non-tumours can be quantified using multiparametric heatmaps obtained via the layer-wise relevance propagation method.

Hemodynamic MRI parameters to predict asymptomatic unilateral carotid artery stenosis with random forest machine learning.

Internal carotid artery stenosis (ICAS) can cause stroke and cognitive decline. Associated hemodynamic impairments, which are most pronounced within individual watershed areas (iWSA) between vascular territories, can be assessed with hemodynamic-oxygenation-sensitive MRI and may help to detect severely affected patients. We aimed to identify the most sensitive parameters and volumes of interest (VOI) to predict high-grade ICAS with random forest machine learning. We hypothesized an increased predictive ability considering iWSAs and a decreased cognitive performance in correctly classified patients. Twenty-four patients with asymptomatic, unilateral, high-grade carotid artery stenosis and 24 age-matched healthy controls underwent MRI comprising pseudo-continuous arterial spin labeling (pCASL), breath-holding functional MRI (BH-fMRI), dynamic susceptibility contrast (DSC), T2 and T2* mapping, MPRAGE and FLAIR. Quantitative maps of eight perfusion, oxygenation and microvascular parameters were obtained. Mean values of respective parameters within and outside of iWSAs split into gray (GM) and white matter (WM) were calculated for both hemispheres and for interhemispheric differences resulting in 96 features. Random forest classifiers were trained on whole GM/WM VOIs, VOIs considering iWSAs and with additional feature selection, respectively. The most sensitive features in decreasing order were time-to-peak (TTP), cerebral blood flow (CBF) and cerebral vascular reactivity (CVR), all of these inside of iWSAs. Applying iWSAs combined with feature selection yielded significantly higher receiver operating characteristics areas under the curve (AUC) than whole GM/WM VOIs (AUC: 0.84 vs. 0.90, p = 0.039). Correctly predicted patients presented with worse cognitive performances than frequently misclassified patients (Trail-making-test B: 152.5s vs. 94.4s, p = 0.034). Random forest classifiers trained on multiparametric MRI data allow identification of the most relevant parameters and VOIs to predict ICAS, which may improve personalized treatments.

Differentiation of Pilocytic Astrocytoma from Glioblastoma using a Machine-Learning framework based upon quantitative T1 perfusion MRI

Background and purposeDifferentiation of pilocytic astrocytoma (PA) from glioblastoma is difficult using conventional MRI parameters. The purpose of this study was to differentiate these two similar in appearance tumors using quantitative T1 perfusion MRI parameters combined under a machine learning framework. Materials and methodsThis retrospective study included age/sex and location matched 26 PA and 33 glioblastoma patients with tumor histopathological characterization performed using WHO 2016 classification. Multi-parametric MRI data were acquired at 3 T scanner and included T1 perfusion and DWI data along with conventional MRI images. Analysis of T1 perfusion data using a leaky-tracer-kinetic-model, first-pass-model and piecewise-linear-model resulted in multiple quantitative parameters. ADC maps were also computed from DWI data. Tumors were segmented into sub-components such as enhancing and non-enhancing regions, edema and necrotic/cystic regions using T1 perfusion parameters. Enhancing and non-enhancing regions were combined and used as an ROI. A support-vector-machine classifier was developed for the classification of PA versus glioblastoma using T1 perfusion MRI parameters/features. The feature set was optimized using a random-forest based algorithm. Classification was also performed between the two tumor types using the ADC parameter. ResultsT1 perfusion parameter values were significantly different between the two groups. The combination of T1 perfusion parameters classified tumors more accurately with a cross validated error of 9.80% against that of ADC's 17.65% error. ConclusionThe approach of using quantitative T1 perfusion parameters based upon a support-vector-machine classifier reliably differentiated PA from glioblastoma and performed better classification than ADC.

Explainable AI for CNN-based prostate tumor segmentation in multi-parametric MRI correlated to whole mount histopathology

Automatic prostate tumor segmentation is often unable to identify the lesion even if multi-parametric MRI data is used as input, and the segmentation output is difficult to verify due to the lack of clinically established ground truth images. In this work we use an explainable deep learning model to interpret the predictions of a convolutional neural network (CNN) for prostate tumor segmentation. The CNN uses a U-Net architecture which was trained on multi-parametric MRI data from 122 patients to automatically segment the prostate gland and prostate tumor lesions. In addition, co-registered ground truth data from whole mount histopathology images were available in 15 patients that were used as a test set during CNN testing. To be able to interpret the segmentation results of the CNN, heat maps were generated using the Gradient Weighted Class Activation Map (Grad-CAM) method. The CNN achieved a mean Dice Sorensen Coefficient 0.62 and 0.31 for the prostate gland and the tumor lesions -with the radiologist drawn ground truth and 0.32 with whole-mount histology ground truth for tumor lesions. Dice Sorensen Coefficient between CNN predictions and manual segmentations from MRI and histology data were not significantly different. In the prostate the Grad-CAM heat maps could differentiate between tumor and healthy prostate tissue, which indicates that the image information in the tumor was essential for the CNN segmentation.

Multiparametric quantitative renal MRI in children and young adults: comparison between healthy individuals and patients with chronic kidney disease.

Multiparametric quantitative renal MRI may provide noninvasive radiologic biomarkers of chronic kidney disease (CKD) based on investigations in animal models and adults. We aimed to (1) obtain normative multiparametric quantitative MRI data from the kidneys of healthy children and young adults, (2) compare MRI measurements between healthy control participants and patients with CKD, and (3) determine if MRI measurements correlate with clinical and laboratory data as well as histology. This was a prospective, case-control study of 20 healthy controls and 12 CKD patients who underwent percutaneous renal biopsy ranging from 12 to 23years of age between October 2018 and March 2020. Kidney function was documented and pathology assessed for fibrosis/inflammation. Utilizing a field strength of 1.5T, we examined renal T1, T2, and T2* relaxation mapping, MR elastography (MRE), and diffusion-weighted imaging (DWI). A single analyst made all manual measurements for quantitative MRI pulse sequences. Independent measurements from cortex, medulla, and whole kidney were obtained by drawing regions of interest on single slices from the upper, mid, and lower kidney. A weighted average was calculated for each kidney; if two kidneys, the right and left were averaged. Continuous variables were compared with Mann-Whitney U test; bivariate relationships were assessed using Spearman rank-order correlation. Median estimated glomerular filtration rate (eGFR) was 112.3ml/min/1.73 m2 in controls (n = 20, 10 females) and 55.0ml/min/m2 in CKD patients (n = 12, 2 females) (p < 0.0001). Whole kidney (1333 vs. 1291ms; p = 0.018) and cortical (1212 vs 1137ms; p < 0.0001) T1 values were higher in CKD patients. Cortical T1 values correlated with eGFR (rho = -0.62; p = 0.0003) and cystatin C (rho = 0.58; p = 0.0007). Whole kidney (1.87 vs. 2.02 10-3 mm2/s; p = 0.007), cortical (1.89 vs. 2.04 10-3 mm2/s; p = 0.008), and medullary (1.87 vs. 1.98 10-3 mm2/s; p = 0.0095) DWI apparent diffusion coefficients (ADC) were lower in CKD patients. Whole kidney ADC correlated with eGFR (rho = 0.45; p = 0.012) and cystatin C (rho = -0.46; p = 0.009). Cortical histologic inflammation correlated with DWI ADC (rho = -0.71; p = 0.011). Renal T1 relaxation and DWI ADC measurements differ between pediatric healthy controls and CKD patients, correlate with laboratory markers of CKD, and may have histologic correlates.

A nomogram strategy for identifying the subclassification of IDH mutation and ATRX expression loss in lower-grade gliomas.

To construct a radiomics nomogram based on multiparametric MRI data for predicting isocitrate dehydrogenase 1 mutation (IDH +) and loss of nuclear alpha thalassemia/mental retardation syndrome X-linked expression (ATRX -) in patients with lower-grade gliomas (LrGG; World Health Organization [WHO] 2016 grades II and III). A total of 111 LrGG patients (76 mutated IDH and 35 wild-type IDH) were enrolled, divided into a training set (n = 78) and a validation set (n = 33) for predicting IDH mutation. IDH + LrGG patients were further stratified into the ATRX - (n = 38) and ATRX + (n = 38) subtypes. A total of 250 radiomics features were extracted from the region of interest of each tumor, including that from T2 fluid-attenuated inversion recovery (T2 FLAIR), contrast-enhanced T1 WI, ASL-derived cerebral blood flow (CBF), DWI-derived ADC, and exponential ADC (eADC). A radiomics signature was selected using the Elastic Net regression model, and a radiomics nomogram was finally constructed using the age, gender information, and above features. The radiomics nomogram identified LrGG patients for IDH mutation (C-index: training sets = 0.881, validation sets = 0.900) and ATRX loss (C-index: training sets = 0.863, validation sets = 0.840) with good calibration. Decision curve analysis further confirmed the clinical usefulness of the two nomograms for predicting IDH and ATRX status. The nomogram incorporating age, gender, and the radiomics signature provided a clinically useful approach in noninvasively predicting IDH and ATRX mutation status for LrGG patients. The proposed method could facilitate MRI-based clinical decision-making for the LrGG patients. • Non-invasive determination of IDH and ATRX gene status of LrGG patients can be obtained with a radiomics nomogram. • The proposed nomogram is constructed by radiomics signature selected from 250 radiomics features, combined with age and gender. • The proposed radiomics nomogram exhibited good calibration and discrimination for IDH and ATRX gene mutation stratification of LrGG patients in both training and validation sets.

Personalized management of prostate cancer: from molecular and imaging markers to radiogenomics.

PurposeProstate cancer (PCa) is the second most common cancer in men. The urge to guide treatment tactics based on personal clinical risk factors has evolved in the era of human genome sequencing. To date, personalized approaches to managing PCa patients have not yet been developed. Radiogenomics is a relatively new term, used to refer to the study of genetic variation associated with imaging features of the tumour in order to improve the prognostication of the disease course.Material and methodsThe study is a review of recent knowledge regarding potential clinical applications of radiogenomics in personalized treatment of PCa.ResultsRecent investigations have proven that by combining data on individual genetic tumour features, and radiomic profiling (radiologic-molecular correlation), with traditional staging procedures in order to personalize treatment of PCa, an improved prognostication of PCa course can be performed, and overtreatment of indolent cancer can be avoided. It was found that a combination of multiparametric MRI and gene expression data allowed the detection of radiomic features of PCa, which correlated with a number of gene signatures associated with adverse outcomes. It was revealed that several molecular markers may drive tumour upstaging, allowed the distinction between the PCa stages, and correlated with aggressiveness-related radiomic features.ConclusionsThe radiogenomics of PCa is not a comprehensively investigated area of oncourology. The combination of genomics and radiomics as integrative parts of precision medicine in the future has the potential to become the foundation for a personalized approach to the management of PCa.

Classification of MR-Detected Additional Lesions in Patients With Breast Cancer Using a Combination of Radiomics Analysis and Machine Learning.

ObjectiveThis study was conducted in order to investigate the feasibility of using radiomics analysis (RA) with machine learning algorithms based on breast magnetic resonance (MR) images for discriminating malignant from benign MR-detected additional lesions in patients with primary breast cancer.Materials and MethodsOne hundred seventy-four MR-detected additional lesions (benign, n = 86; malignancy, n = 88) from 158 patients with ipsilateral primary breast cancer from a tertiary medical center were included in this retrospective study. The entire data were randomly split to training (80%) and independent test sets (20%). In addition, 25 patients (benign, n = 21; malignancy, n = 15) from another tertiary medical center were included for the external test. Radiomics features that were extracted from three regions-of-interest (ROIs; intratumor, peritumor, combined) using fat-saturated T1-weighted images obtained by subtracting pre- from postcontrast images (SUB) and T2-weighted image (T2) were utilized to train the support vector machine for the binary classification. A decision tree method was utilized to build a classifier model using clinical imaging interpretation (CII) features assessed by radiologists. Area under the receiver operating characteristic curve (AUROC), accuracy, sensitivity, and specificity were used to compare the diagnostic performance.ResultsThe RA models trained using radiomics features from the intratumor-ROI showed comparable performance to the CII model (accuracy, AUROC: 73.3%, 69.6% for the SUB RA model; 70.0%, 75.1% for the T2 RA model; 73.3%, 72.0% for the CII model). The diagnostic performance increased when the radiomics and CII features were combined to build a fusion model. The fusion model that combines the CII features and radiomics features from multiparametric MRI data demonstrated the highest performance with an accuracy of 86.7% and an AUROC of 91.1%. The external test showed a similar pattern where the fusion models demonstrated higher levels of performance compared with the RA- or CII-only models. The accuracy and AUROC of the SUB+T2 RA+CII model in the external test were 80.6% and 91.4%, respectively.ConclusionOur study demonstrated the feasibility of using RA with machine learning approach based on multiparametric MRI for quantitatively characterizing MR-detected additional lesions. The fusion model demonstrated an improved diagnostic performance over the models trained with either RA or CII alone.