Publication In Radiology Research Articles

Physics-Informed Autoencoder for Prostate Tissue Microstructure Profiling with Hybrid Multidimensional MRI.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate the performance of Physics-Informed Autoencoder (PIA), a self-supervised deep learning model, in measuring tissue-based biomarkers for prostate cancer (PCa) using hybrid multidimensional MRI. Materials and Methods This retrospective study introduces PIA, a novel self-supervised deep learning model that integrates a three-compartment diffusion-relaxation model with hybrid multidimensional MRI. PIA was trained to encode the biophysical model into a deep neural network to predict measurements of tissue-specific biomarkers for PCa without extensive training data requirements. Comprehensive in-silico and in-vivo experiments, using histopathology measurements as the reference standard, were conducted to validate the model's efficacy in comparison to the traditional Non-Linear Least Squares (NLLS) algorithm. PIA's robustness to noise was tested in in-silico experiments with varying signal-to-noise ratio (SNR) conditions, and in-vivo performance for estimating volume fractions was evaluated in 21 patients (mean age 60 (SD:6.6) years; all male) with PCa (n = 71 regions of interest). Evaluation metrics included the intraclass correlation coefficient (ICC) and Pearson correlation coefficient. Results PIA predicted the reference standard tissue parameters with high accuracy, outperforming conventional NLLS methods, especially under noisy conditions (rs = 0.80 versus 0.65, P < .001 for epithelium volume at SNR = 20:1). In in-vivo validation, PIA's noninvasive volume fraction estimates matched quantitative histology (ICC = 0.94, 0.85 and 0.92 for epithelium, stroma, and lumen compartments, respectively, P < .001 for all). PIA's measurements strongly correlated with PCa aggressiveness (r = 0.75, P < .001). Furthermore, PIA ran 10,000 faster than NLLS (0.18 seconds versus 40 minutes per image). Conclusion PIA provided accurate prostate tissue biomarker measurements from MRI data with better robustness to noise and computational efficiency compared with the NLLS algorithm. The results demonstrate the potential of PIA as an accurate, noninvasive, and explainable AI method for PCa detection. ©RSNA, 2025.

Performance of Two Deep Learning-based AI Models for Breast Cancer Detection and Localization on Screening Mammograms from BreastScreen Norway.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate cancer detection and marker placement accuracy of two artificial intelligence (AI) models developed for interpretation of screening mammograms. Materials and Methods This retrospective study included data from 129 434 screening examinations (all female, mean age 59.2, SD = 5.8) performed between January 2008 and December 2018 in BreastScreen Norway. Model A was commercially available and B was an in-house model. Area under the receiver operating characteristic curve (AUC) with 95% confidence interval (CIs) were calculated. The study defined 3.2% and 11.1% of the examinations with the highest AI scores as positive, threshold 1 and 2, respectively. A radiologic review assessed location of AI markings and classified interval cancers as true or false negative. Results The AUC was 0.93 (95% CI: 0.92-0.94) for model A and B when including screen-detected and interval cancers. Model A identified 82.5% (611/741) of the screen-detected cancers at threshold 1 and 92.4% (685/741) at threshold 2. For model B, the numbers were 81.8% (606/741) and 93.7% (694/741), respectively. The AI markings were correctly localized for all screen-detected cancers identified by both models and 82% (56/68) of the interval cancers for model A and 79% (54/68) for B. At the review, 21.6% (45/208) of the interval cancers were identified at the preceding screening by either or both models, correctly localized and classified as false negative (n = 17) or with minimal signs of malignancy (n = 28). Conclusion Both AI models showed promising performance for cancer detection on screening mammograms. The AI markings corresponded well to the true cancer locations. ©RSNA, 2025.

Performance of Lung Cancer Prediction Models for Screening-detected, Incidental, and Biopsied Pulmonary Nodules.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate the performance of eight lung cancer prediction models on patient cohorts with screening-detected, incidentally-detected, and bronchoscopically-biopsied pulmonary nodules. Materials and Methods This study retrospectively evaluated promising predictive models for lung cancer prediction in three clinical settings: lung cancer screening with low-dose CT, incidentally, detected pulmonary nodules, and nodules deemed suspicious enough to warrant a biopsy. The area under the receiver operating characteristic curve (AUC) of eight validated models including logistic regressions on clinical variables and radiologist nodule characterizations, artificial intelligence (AI) on chest CTs, longitudinal imaging AI, and multimodal approaches for prediction of lung cancer risk was assessed in 9 cohorts (n = 898, 896, 882, 219, 364, 117, 131, 115, 373) from multiple institutions. Each model was implemented from their published literature, and each cohort was curated from primary data sources collected over periods within 2002 to 2021. Results No single predictive model emerged as the highest-performing model across all cohorts, but certain models performed better in specific clinical contexts. Single timepoint chest CT AI performed well for screening-detected nodules but did not generalize well to other clinical settings. Longitudinal imaging and multimodal models demonstrated comparatively good performance on incidentally-detected nodules. When applied to biopsied nodules, all models showed low performance. Conclusion Eight lung cancer prediction models failed to generalize well across clinical settings and sites outside of their training distributions. ©RSNA, 2025.

Non-Conscious Detection of "Missed" Lung Nodules by Radiologists: Expanding the Boundaries of Successful Processing During the Visual Assessment of Chest CT Scans.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Background Diagnostic error rates for detecting small lung nodules on chest CT scans remain high at 50%, despite advances in imaging technology and radiologist training. These failure rates may stem from limitations in conscious recognition processes. However, successful visual processes may be detecting the nodule independent of the radiologist's report. Purpose To investigate visual processing in radiologists during the assessment of chest nodules to determine if radiologists have successful non-conscious processes that detect lung nodules on chest CT examinations even when not consciously recognized or considered, as evidenced by changes in how long they look (dwell time) and pupil size to missed nodules. Materials and Methods This prospective study, conducted from [8/14] to [09/23], compared 6 experienced radiologists with 6 medically naïve control participants. Participants viewed 18 chest CTs (9 abnormal with 16 nodules, 9 normal) to detect lung nodules. High-speed video eye-tracking measured gaze duration and pupil size (indicating physiological arousal) at missed nodule locations and same locations on normal CTs. The reference standard was the known presence or absence of nodules (as determined by a 4-radiologist consensus panel) in abnormal and normal CTs, respectively. Primary outcome measures were detection rates of nodules, dwell time and pupil size at nodule locations versus normal tissue. Paired t-tests were used for statistical analysis. Results Twelve participants (6 radiologists [9.3 average years of radiological experience]) 6 controls (with no radiological experience) were evaluated. Radiologists missed on average 59% of these lung nodules. For missed nodules, radiologists exhibited longer dwell times (Mean: 228 milliseconds vs 175 milliseconds, P=.005) and larger pupil area (Mean: 1446 pixels vs. 1349 pixels, P=.04.) than normal tissue. Control participants showed no differences in dwell time (Mean: 197 milliseconds vs 180 milliseconds, P= .64) or pupil size (Mean: 1426 pixels vs. 1714 pixels, P=.23) for missed nodules than normal tissue locations. Conclusion Radiologists non-conscious processes during visual assessment of a CT examination can detect lung nodules on chest CTs even when conscious recognition fails, as evidenced by increased dwell time and larger pupil size. This successful non-conscious detection is a result of general radiology training.

CMRxRecon2024: A Multimodality, Multiview k-Space Dataset Boosting Universal Machine Learning for Accelerated Cardiac MRI.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. The released CMRxRecon2024 dataset is currently the largest and most protocol-diverse publicly available k-space dataset including multi-modality and multi-view cardiac MRI data from 330 healthy volunteers, and each one covers standardized and commonly used clinical protocols. ©RSNA, 2025.

Impact of Scanner Manufacturer, Endorectal Coil Use, and Clinical Variables on Deep Learning-assisted Prostate Cancer Classification Using Multiparametric MRI.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To assess the impact of scanner manufacturer and scan protocol on the performance of deep learning models to classify prostate cancer (PCa) aggressiveness on biparametric MRI (bpMRI). Materials and Methods In this retrospective study, 5,478 cases from ProstateNet, a PCa bpMRI dataset with examinations from 13 centers, were used to develop five deep learning (DL) models to predict PCa aggressiveness with minimal lesion information and test how using data from different subgroups-scanner manufacturers and endorectal coil (ERC) use (Siemens, Philips, GE with and without ERC and the full dataset)-impacts model performance. Performance was assessed using the area under the receiver operating characteristic curve (AUC). The impact of clinical features (age, prostate-specific antigen level, Prostate Imaging Reporting and Data System [PI-RADS] score) on model performance was also evaluated. Results DL models were trained on 4,328 bpMRI cases, and the best model achieved AUC = 0.73 when trained and tested using data from all manufacturers. Hold-out test set performance was higher when models trained with data from a manufacturer were tested on the same manufacturer (within-and between-manufacturer AUC differences of 0.05 on average, P < .001). The addition of clinical features did not improve performance (P = .24). Learning curve analyses showed that performance remained stable as training data increased. Analysis of DL features showed that scanner manufacturer and scan protocol heavily influenced feature distributions. Conclusion In automated classification of PCa aggressiveness using bpMRI data, scanner manufacturer and endorectal coil use had a major impact on DL model performance and features. Published under a CC BY 4.0 license.

Evaluating the Impact of Changes in AI-derived Case Scores over Time on Digital Breast Tomosynthesis Screening Outcomes.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate the change in DBT-AI (digital breast tomosynthesis-artificial intelligence) case scores over sequential screens. Materials and Methods This retrospective review included 21,108 female patients (mean age, 58.1 ± [SD] 11.5 years) with 31,741 DBT screening examinations performed at a single site from 2/3/2020 to 9/12/2022. Among 7,000 patients with two or more DBT-AI screenings, 1,799 had one year follow up and were included in the analysis. DBT-AI case scores and differences in case score over time were determined. Case scores ranged from 0-100. For each screening outcome (true positive (TP), false positive (FP), true negative (TN), false negative (FN)), mean and median case score change was calculated. Results The highest average case score was seen in TP examinations (average 75, range 7-100, n = 41), and the lowest average case score was seen in TN examinations (average 34, range 0-100, n = 1640). The largest positive case score change was seen in TP examinations (mean case score change 21.1, median case score change 17). FN examinations included mammographically occult cancers diagnosed following supplemental screening and those found on symptomatic diagnostic imaging. Differences between TP and TN mean case score change (P < .001) and between TP and FP mean case score change (P = .02) were statistically significant. Conclusion Using the combination of DBT-AI case score with change in case score over time may help radiologists make recall decisions in DBT screening. All studies with high case score and/or case score changes should be carefully scrutinized to maximize screening performance. ©RSNA, 2025.

Posttraining Network Compression for 3D Medical Image Segmentation: Reducing Computational Efforts via Tucker Decomposition.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To investigate whether the computational effort of 3D CT-based multiorgan segmentation with TotalSegmentator can be reduced via Tucker decomposition-based network compression. Materials and Methods In this retrospective study, Tucker decomposition was applied to the convolutional kernels of the TotalSegmentator model, an nnU-Net model trained on a comprehensive CT dataset for automatic segmentation of 117 anatomic structures. The proposed approach reduced the floating-point operations (FLOPs) and memory required during inference, offering an adjustable trade-off between computational efficiency and segmentation quality. This study utilized the publicly available TotalSegmentator dataset containing 1228 segmented CTs and a test subset of 89 CTs, employing various downsampling factors to explore the relationship between model size, inference speed, and segmentation accuracy, evaluated using the Dice score. Results The application of Tucker decomposition to the TotalSegmentator model substantially reduced the model parameters and FLOPs across various compression ratios, with limited loss in segmentation accuracy. Up to 88% of the model's parameters were removed, with no evidence of differences in performance compared with the original model for 113 of 117 classes after fine-tuning. Practical benefits varied across different graphics processing unit architectures, with more distinct speed-ups on less powerful hardware. Conclusion The study demonstrates that posthoc network compression via Tucker decomposition presents a viable strategy for reducing the computational demand of medical image segmentation models without substantially impacting model accuracy. ©RSNA, 2025.

NNFit: A Self-Supervised Deep Learning Method for Accelerated Quantification of High- Resolution Short Echo Time MR Spectroscopy Datasets.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop and evaluate the performance of NNFit, a self-supervised deep-learning method for quantification of high-resolution short echo-time (TE) echo-planar spectroscopic imaging (EPSI) datasets, with the goal of addressing the computational bottleneck of conventional spectral quantification methods in the clinical workflow. Materials and Methods This retrospective study included 89 short-TE whole-brain EPSI/GRAPPA scans from clinical trials for glioblastoma (Trial 1, May 2014-October 2018) and major-depressive-disorder (Trial 2, 2022- 2023). The training dataset included 685k spectra from 20 participants (60 scans) in Trial 1. The testing-dataset included 115k spectra from 5 participants (13 scans) in Trial 1 and 145k spectra from 7 participants (16 scans) in Trial 2. A comparative analysis was performed between NNFit and a widely used parametric-modeling spectral quantitation method (FITT). Metabolite maps generated by each method were compared using the structural- similarity-index-measure (SSIM) and linear-correlation-coefficient (R2). Radiation treatment volumes for glioblastoma based on the metabolite maps were compared with the Dice-coefficient and a two-tailed t test. Results Average SSIM and R2 scores for Trial 1 test set data were 0.91/0.90 (choline), 0.93/0.93 (creatine), 0.93/0.93 (n-acetylaspartate), 0.80/0.72 (myo-inositol), and 0.59/0.47 (glutamate + glutamine). Average scores for Trial 2 test set data were 0.95/0.95, 0.98/0.97, 0.98/0.98, 0.92/0.92, and 0.79/0.81 respectively. The treatment volumes had average Dice coefficient of 0.92. NNFit's average processing time was 90.1 seconds, whereas FITT took 52.9 minutes on average. Conclusion This study demonstrates that a deep learning approach to spectral quantitation offers comparable performance to conventional quantification methods for EPSI data, but with faster processing at short-TE. ©RSNA, 2025.

A Serial MRI-based Deep Learning Model to Predict Survival in Patients with Locoregionally Advanced Nasopharyngeal Carcinoma.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop and evaluate a deep learning-based prognostic model for predicting survival in locoregionally- advanced nasopharyngeal carcinoma (LA-NPC) using serial MRI before and after induction chemotherapy (IC). Materials and Methods This multicenter retrospective study included 1039 LA-NPC patients (779 male, 260 female, mean age 44 [standard deviation: 11]) diagnosed between April 2009 and December 2015. A radiomics- clinical prognostic model (Model RC) was developed using pre-and post-IC MRI and other clinical factors using graph convolutional neural networks (GCN). The concordance index (C-index) was used to evaluate model performance in predicting disease-free survival (DFS). The survival benefits of concurrent chemoradiation therapy (CCRT) were analyzed in model-defined risk groups. Results The C-indexes of Model RC for predicting DFS were significantly higher than those of TNM staging in the internal (0.79 versus 0.53) and external (0.79 versus 0.62, both P < .001) testing cohorts. The 5-year DFS for the Model RC-defined low-risk group was significantly better than that of the high-risk group (90.6% versus 58.9%, P < .001). In high-risk patients, those who received CCRT had a higher 5-year DFS rate than those who did not (58.7% versus 28.6%, P = .03). There was no evidence of a difference in 5-year DFS rate in low-risk patients who did or did not receive CCRT (91.9% versus 81.3%, P = .19). Conclusion Serial MRI before and after IC can effectively predict survival in LA-NPC. The radiomics-clinical prognostic model developed using a GCN-based deep learning method showed good risk discrimination capabilities and may facilitate risk-adaptive therapy. ©RSNA, 2025.

A retrospective analysis of medical student authorship in highest impact radiology journal publications.

To assess medical student authorship in radiology research, focusing on the prevalence and impact of student articles, demographic trends, and potential barriers and opportunities for involvement. We retrospectively assessed original research and review papers from 2018 to 2022 in the top five radiology journals based on their 2022 Impact Factor. Three reviewers manually validated and classified articles by medical student (MS) and non-MS authorship. Data collected included publication and authorship metrics. Thematic analysis of articles was performed from keywords and SCOPUS topic clusters. Significant associations were identified using chi-squared and two-tailed z-tests. Of the 2533 publications from all five journals, only 0.47 % were MS-authored, of which 83.3 % had a Field Weighted Citation Impact (FWCI) > 1. Of the 19 MS authors, 68.42 % had prior publications, with the same proportion holding prior first authorship. Female students averaged a higher FWCI (2.47 ± 2.31) but comprised only 26.32 % of all students. Only 16.67 % of MS articles had a female senior author (SA). This was consistent in non-MS articles, with females representing 19.9 % to 25.2 % of SAs across all journals. Of the 2521 non-MS articles, 92 % were classified into themes, revealing that 47.89 % of publications primarily focused on machine learning, while another 42.26 % incorporated some machine learning concepts. Compared to non-medical student publications in radiology, those by medical students were disproportionately limited but impactful. Both demographic and academic factors hinder student engagement in publishing. Hence promoting, supporting, and sponsoring student involvement in research is pivotal for the discipline.

Automated Deep Learning-Based Detection and Segmentation of Lung Tumors at CT.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Background Detection and segmentation of lung tumors on CT scans are critical for monitoring cancer progression, evaluating treatment responses, and planning radiation therapy; however, manual delineation is labor-intensive and subject to physician variability. Purpose To develop and evaluate an ensemble deep learning model for automating identification and segmentation of lung tumors on CT scans. Materials and Methods A retrospective study was conducted between July 2019 and November 2024 using a large dataset of CT simulation scans and clinical lung tumor segmentations from radiotherapy plans. This dataset was used to train a 3D U-Net-based, image-multiresolution ensemble model to detect and segment lung tumors on CT scans. Model performance was evaluated on internal and external test sets composed of CT simulation scans and lung tumor segmentations from two affiliated medical centers, including single primary and metastatic lung tumors. Performance metrics included sensitivity, specificity, false positive rate, and Dice similarity coefficient (DSC). Model-predicted tumor volumes were compared with physician-delineated volumes. Group comparisons were made with Wilcoxon signed-rank test or one-way ANOVA. P < 0.05 indicated statistical significance. Results The model, trained on 1,504 CT scans with clinical lung tumor segmentations, achieved 92% sensitivity (92/100) and 82% specificity (41/50) in detecting lung tumors on the combined 150-CT scan test set. For a subset of 100 CT scans with a single lung tumor each, the model achieved a median model-physician DSC of 0.77 (IQR: 0.65-0.83) and an interphysician DSC of 0.80 (IQR: 0.72-0.86). Segmentation time was shorter for the model than for physicians (mean 76.6 vs. 166.1-187.7 seconds; p<0.001). Conclusion Routinely collected radiotherapy data were useful for model training. The key strengths of the model include a 3D U-Net ensemble approach for balancing volumetric context with resolution, robust tumor detection and segmentation performance, and the ability to generalize to an external site.

FastMRI Breast: A Publicly Available Radial k-Space Dataset of Breast Dynamic Contrast-enhanced MRI.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. The fastMRI breast dataset is the first large-scale dataset of radial k-space and DICOM data for breast dynamic contrast-enhanced MRI with case-level labels. Its public availability aims to advance fast and quantitative machine learning research. ©RSNA, 2025.

Insights into Radiology Publications.

The evolution of modern medicine has been significantly driven by medical and health care research, underscoring the importance of disseminating findings to advance health care. Medical literature, encompassing various publication types such as case reports, review articles, and original research, plays a crucial role in this process by facilitating the communication and discussion of new discoveries. This review article provides a comprehensive guide to understanding and navigating radiologic publications. It examines the various types of radiologic research articles, including case reports and series, pictorial reviews, original research, systematic reviews, and meta-analyses, each of which serve distinct purposes in contributing to the field of radiology. The study adopts the "six honest men" approach-addressing why, who, what, when, where, and how-to elucidate the essential elements of successful radiology research and publication. Key topics include the motivations for publishing, the types of articles suited for different research questions, and strategic considerations for selecting appropriate journals. Additionally, the review highlights the importance of understanding publication timing, journal selection criteria, and the overall publication process, including manuscript preparation and peer review. By offering these insights, the review aims to equip early-career researchers with the knowledge and skills necessary to effectively contribute to radiology literature and advance their academic and professional careers.

Applying Conformal Prediction to a Deep Learning Model for Intracranial Hemorrhage Detection to Improve Trustworthiness.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To apply conformal prediction to a deep learning (DL) model for intracranial hemorrhage (ICH) detection and evaluate model performance in detection as well as model accuracy in identifying challenging cases. Materials and Methods This was a retrospective (November 2017 through December 2017) study of 491 noncontrast head CT volumes from the CQ500 dataset in which three senior radiologists annotated sections containing ICH. The dataset was split into definite and challenging (uncertain) subsets, where challenging images were defined as those in which there was disagreement among readers. A DL model was trained on 146 patients (mean age = 45.7, 70 females, 76 males) from the definite data (training dataset) to perform ICH localization and classification into five classes. To develop an uncertainty-aware DL model, 1,546 sections of the definite data (calibration dataset) was used for Mondrian conformal prediction (MCP). The uncertainty-aware DL model was tested on 8,401 definite and challenging sections to assess its ability to identify challenging sections. The difference in predictive performance (P value) and ability to identify challenging sections (accuracy) were reported. Results After the MCP procedure, the model achieved an F1 score of 0.920 for ICH classification on the test dataset. Additionally, it correctly identified 6,837 of the 6,856 total challenging sections as challenging (99.7% accuracy). It did not incorrectly label any definite sections as challenging. Conclusion The uncertainty-aware MCP-augmented DL model achieved high performance in ICH detection and high accuracy in identifying challenging sections, suggesting its usefulness in automated ICH detection and potential to increase trustworthiness of DL models in radiology. ©RSNA, 2024.

Deep Learning Applied to Diffusion-weighted Imaging for Differentiating Malignant from Benign Breast Tumors without Lesion Segmentation.

Purpose To evaluate and compare the performance of different artificial intelligence (AI) models in differentiating between benign and malignant breast tumors at diffusion-weighted imaging (DWI), including comparison with radiologist assessments. Materials and Methods In this retrospective study, patients with breast lesions underwent 3-T breast MRI from May 2019 to March 2022. In addition to T1-weighted imaging, T2-weighted imaging, and contrast-enhanced imaging, DWI was performed with five b values (0, 200, 800, 1000, and 1500 sec/mm2). DWI data split into training and tuning and test sets were used for the development and assessment of AI models, including a small two-dimensional (2D) convolutional neural network (CNN), ResNet-18, EfficientNet-B0, and a three-dimensional (3D) CNN. Performance of the DWI-based models in differentiating between benign and malignant breast tumors was compared with that of radiologists assessing standard breast MR images, with diagnostic performance assessed using receiver operating characteristic analysis. The study also examined data augmentation effects (augmentation A: random elastic deformation, augmentation B: random affine transformation and random noise, and augmentation C: mixup) on model performance. Results A total of 334 breast lesions in 293 patients (mean age, 54.9 years ± 14.3 [SD]; all female) were analyzed. The 2D CNN models outperformed the 3D CNN on the test dataset (area under the receiver operating characteristic curve [AUC] with different data augmentation methods: range, 0.83-0.88 vs 0.75-0.76). There was no evidence of a difference in performance between the small 2D CNN with augmentations A and B (AUC: 0.88) and the radiologists (AUC: 0.86) on the test dataset (P = .64). When comparing the small 2D CNN to radiologists, there was no evidence of a difference in specificity (81.4% vs 72.1%, P = .64) or sensitivity (85.9% vs 98.8%, P = .64). Conclusion AI models, particularly a small 2D CNN, showed good performance in differentiating between malignant and benign breast tumors using DWI, without needing manual segmentation. Keywords: MR Imaging, Breast, Comparative Studies, Feature Detection, Diagnosis Supplemental material is available for this article. ©RSNA, 2024.

Combining Biology-based and MRI Data-driven Modeling to Predict Response to Neoadjuvant Chemotherapy in Patients with Triple-Negative Breast Cancer.

Purpose To combine deep learning and biology-based modeling to predict the response of locally advanced, triple-negative breast cancer before initiating neoadjuvant chemotherapy (NAC). Materials and Methods In this retrospective study, a biology-based mathematical model of tumor response to NAC was constructed and calibrated on a patient-specific basis using imaging data from patients enrolled in the MD Anderson A Robust TNBC Evaluation FraMework to Improve Survival trial (ARTEMIS; ClinicalTrials.gov registration no. NCT02276443) between April 2018 and May 2021. To relate the calibrated parameters in the biology-based model and pretreatment MRI data, a convolutional neural network (CNN) was employed. The CNN predictions of the calibrated model parameters were used to estimate tumor response at the end of NAC. CNN performance in the estimations of total tumor volume (TTV), total tumor cellularity (TTC), and tumor status was evaluated. Model-predicted TTC and TTV measurements were compared with MRI-based measurements using the concordance correlation coefficient and area under the receiver operating characteristic curve (for predicting pathologic complete response at the end of NAC). Results The study included 118 female patients (median age, 51 years [range, 29-78 years]). For comparison of CNN predicted to measured change in TTC and TTV over the course of NAC, the concordance correlation coefficient values were 0.95 (95% CI: 0.90, 0.98) and 0.94 (95% CI: 0.87, 0.97), respectively. CNN-predicted TTC and TTV had an area under the receiver operating characteristic curve of 0.72 (95% CI: 0.34, 0.94) and 0.72 (95% CI: 0.40, 0.95) for predicting tumor status at the time of surgery, respectively. Conclusion Deep learning integrated with a biology-based mathematical model showed good performance in predicting the spatial and temporal evolution of a patient's tumor during NAC using only pre-NAC MRI data. Keywords: Triple-Negative Breast Cancer, Neoadjuvant Chemotherapy, Convolutional Neural Network, Biology-based Mathematical Model Supplemental material is available for this article. Clinical trial registration no. NCT02276443 ©RSNA, 2024 See also commentary by Mei and Huang in this issue.

SCIseg: Automatic Segmentation of Intramedullary Lesions in Spinal Cord Injury on T2-weighted MRI Scans.

Purpose To develop a deep learning tool for the automatic segmentation of the spinal cord and intramedullary lesions in spinal cord injury (SCI) on T2-weighted MRI scans. Materials and Methods This retrospective study included MRI data acquired between July 2002 and February 2023. The data consisted of T2-weighted MRI scans acquired using different scanner manufacturers with various image resolutions (isotropic and anisotropic) and orientations (axial and sagittal). Patients had different lesion etiologies (traumatic, ischemic, and hemorrhagic) and lesion locations across the cervical, thoracic, and lumbar spine. A deep learning model, SCIseg (which is open source and accessible through the Spinal Cord Toolbox, version 6.2 and above), was trained in a three-phase process involving active learning for the automatic segmentation of intramedullary SCI lesions and the spinal cord. The segmentations from the proposed model were visually and quantitatively compared with those from three other open-source methods (PropSeg, DeepSeg, and contrast-agnostic, all part of the Spinal Cord Toolbox). The Wilcoxon signed rank test was used to compare quantitative MRI biomarkers of SCI (lesion volume, lesion length, and maximal axial damage ratio) derived from the manual reference standard lesion masks and biomarkers obtained automatically with SCIseg segmentations. Results The study included 191 patients with SCI (mean age, 48.1 years ± 17.9 [SD]; 142 [74%] male patients). SCIseg achieved a mean Dice score of 0.92 ± 0.07 and 0.61 ± 0.27 for spinal cord and SCI lesion segmentation, respectively. There was no evidence of a difference between lesion length (P = .42) and maximal axial damage ratio (P = .16) computed from manually annotated lesions and the lesion segmentations obtained using SCIseg. Conclusion SCIseg accurately segmented intramedullary lesions on a diverse dataset of T2-weighted MRI scans and automatically extracted clinically relevant lesion characteristics. Keywords: Spinal Cord, Trauma, Segmentation, MR Imaging, Supervised Learning, Convolutional Neural Network (CNN) Published under a CC BY 4.0 license.

Integrated Deep Learning Model for the Detection, Segmentation, and Morphologic Analysis of Intracranial Aneurysms Using CT Angiography.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop a deep learning model for the morphologic measurement of unruptured intracranial aneurysms (UIAs) based on CT angiography (CTA) data and validate its performance using a multicenter dataset. Materials and Methods In this retrospective study, patients with CTA examinations, including those with and without UIAs, in a tertiary referral hospital from February 2018 to February 2021 were included as the training dataset. Patients with UIAs who underwent CTA at multiple centers between April 2021 to December 2022 were included as the multicenter external testing set. An integrated deep-learning (IDL) model was developed for UIA detection, segmentation and morphologic measurement using an nnU-net algorithm. Model performance was evaluated using the Dice similarity coefficient (DSC) and intraclass correlation coefficient (ICC), with measurements by senior radiologists serving as the reference standard. The ability of the IDL model to improve performance of junior radiologists in measuring morphologic UIA features was assessed. Results The study included 1182 patients with UIAs and 578 controls without UIAs as the training dataset (55 years [IQR, 47-62], 1,012 [57.5%] females) and 535 patients with UIAs as the multicenter external testing set (57 years [IQR, 50-63], 353 [66.0%] females). The IDL model achieved 97% accuracy in detecting UIAs and achieved a DSC of 0.90 (95%CI, 0.88-0.92) for UIA segmentation. Model-based morphologic measurements showed good agreement with reference standard measurements (all ICCs > 0.85). Within the multicenter external testing set, the IDL model also showed agreement with reference standard measurements (all ICCs > 0.80). Junior radiologists assisted by the IDL model showed significantly improved performance in measuring UIA size (ICC improved from 0.88 [0.80-0.92] to 0.96 [0.92-0.97], P < .001). Conclusion The developed integrated deep learning model using CTA data showed good performance in UIA detection, segmentation and morphologic measurement and may be used to assist less experienced radiologists in morphologic analysis of UIAs. ©RSNA, 2024.

AI-integrated Screening to Replace Double Reading of Mammograms: A Population-wide Accuracy and Feasibility Study.

Mammography screening supported by deep learning-based artificial intelligence (AI) solutions can potentially reduce workload without compromising breast cancer detection accuracy, but the site of deployment in the workflow might be crucial. This retrospective study compared three simulated AI-integrated screening scenarios with standard double reading with arbitration in a sample of 249 402 mammograms from a representative screening population. A commercial AI system replaced the first reader (scenario 1: integrated AIfirst), the second reader (scenario 2: integrated AIsecond), or both readers for triaging of low- and high-risk cases (scenario 3: integrated AItriage). AI threshold values were chosen based partly on previous validation and setting the screen-read volume reduction at approximately 50% across scenarios. Detection accuracy measures were calculated. Compared with standard double reading, integrated AIfirst showed no evidence of a difference in accuracy metrics except for a higher arbitration rate (+0.99%, P < .001). Integrated AIsecond had lower sensitivity (-1.58%, P < .001), negative predictive value (NPV) (-0.01%, P < .001), and recall rate (-0.06%, P = .04) but a higher positive predictive value (PPV) (+0.03%, P < .001) and arbitration rate (+1.22%, P < .001). Integrated AItriage achieved higher sensitivity (+1.33%, P < .001), PPV (+0.36%, P = .03), and NPV (+0.01%, P < .001) but lower arbitration rate (-0.88%, P < .001). Replacing one or both readers with AI seems feasible; however, the site of application in the workflow can have clinically relevant effects on accuracy and workload. Keywords: Mammography, Breast, Neoplasms-Primary, Screening, Epidemiology, Diagnosis, Convolutional Neural Network (CNN) Supplemental material is available for this article. Published under a CC BY 4.0 license.