<i>"Just Accepted" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. 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.</i> Purpose To develop a transformer-based deep learning model-MR-Transformer-that leverages ImageNet pretraining and three-dimensional (3D) spatial correlations to predict the progression of knee osteoarthritis to TKR using MRI. Materials and Methods This retrospective study included 353 case-control matched pairs of coronal intermediate-weighted turbo spin-echo (COR-IW-TSE) and sagittal intermediate-weighted turbo spin-echo with fat suppression (SAG-IW-TSE-FS) knee MRIs from the Osteoarthritis Initiative (OAI) database, with a follow-up period up to 9 years, and 270 case-control matched pairs of coronal short-tau inversion recovery (COR-STIR) and sagittal proton density fat-saturated (SAG-PD-FAT-SAT) knee MRIs from the Multicenter Osteoarthritis Study (MOST) database, with a follow-up period up to 7 years. Performance of the MR-Transformer to predict the progression of knee osteoarthritis was compared with that of existing state-of-the-art deep learning models (TSE-Net, 3DMeT, and MRNet) using sevenfold nested cross-validation across the four MRI tissue sequences. Results MR-Transformer achieved areas under the receiver operating characteristic curves (AUCs) of 0.88 (95% CI: 0.85, 0.91), 0.88 (95% CI: 0.85, 0.90), 0.86 (95% CI: 0.82, 0.89), and 0.84 (95% CI: 0.81, 0.87) for COR-IW-TSE, SAG-IW-TSE-FS, COR-STIR, and SAG-PD-FAT-SAT, respectively. The model achieved a higher AUC than that of 3DMeT for all MRI sequences (<i>P</i> < .001). The model showed the highest sensitivity of 83% (95% CI: 78, 87%) and specificity of 83% (95% CI: 76, 88%) for the COR-IW-TSE MRI sequence. Conclusion Compared with the existing deep learning models, the MR-Transformer exhibited state-of-the-art performance in predicting the progression of knee osteoarthritis to TKR using MRIs. ©RSNA, 2025.

The prediction and management of sudden cardiac death risk continue to pose significant challenges in cardiovascular care despite advances in therapies over the last two decades. Late gadolinium enhancement (LGE) on cardiac magnetic resonance-a marker of myocardial fibrosis-is a powerful non-invasive tool with the potential to aid the prediction of sudden death and direct the use of preventative therapies in several cardiovascular conditions. In this state-of-the-art review, we provide a critical appraisal of the current evidence base underpinning the utility of LGE in both ischaemic and non-ischaemic cardiomyopathies together with a focus on future perspectives and the role for machine learning and digital twin technologies.

Sparse autoencoders (SAEs) emerged as a promising tool for mechanistic interpretability of transformer-based foundation models. Very recently, SAEs were also adopted for the visual domain, enabling the discovery of visual concepts and their patch-wise attribution to tokens in the transformer model. While a growing number of foundation models emerged for medical imaging, tools for explaining their inferences are still lacking. In this work, we show the applicability of SAEs for hematology. We propose CytoSAE, a sparse autoencoder which is trained on over 40,000 peripheral blood single-cell images. CytoSAE generalizes to diverse and out-of-domain datasets, including bone marrow cytology, where it identifies morphologically relevant concepts which we validated with medical experts. Furthermore, we demonstrate scenarios in which CytoSAE can generate patient-specific and disease-specific concepts, enabling the detection of pathognomonic cells and localized cellular abnormalities at the patch level. We quantified the effect of concepts on a patient-level AML subtype classification task and show that CytoSAE concepts reach performance comparable to the state-of-the-art, while offering explainability on the sub-cellular level. Source code and model weights are available at https://github.com/dynamical-inference/cytosae.

Radiologists rely on eye movements to navigate and interpret medical images. A trained radiologist possesses knowledge about the potential diseases that may be present in the images and, when searching, follows a mental checklist to locate them using their gaze. This is a key observation, yet existing models fail to capture the underlying intent behind each fixation. In this paper, we introduce a deep learning-based approach, RadGazeIntent, designed to model this behavior: having an intention to find something and actively searching for it. Our transformer-based architecture processes both the temporal and spatial dimensions of gaze data, transforming fine-grained fixation features into coarse, meaningful representations of diagnostic intent to interpret radiologists' goals. To capture the nuances of radiologists' varied intention-driven behaviors, we process existing medical eye-tracking datasets to create three intention-labeled subsets: RadSeq (Systematic Sequential Search), RadExplore (Uncertainty-driven Exploration), and RadHybrid (Hybrid Pattern). Experimental results demonstrate RadGazeIntent's ability to predict which findings radiologists are examining at specific moments, outperforming baseline methods across all intention-labeled datasets.

<b>BACKGROUND</b>. Deep learning (DL) superresolution image reconstruction enables higher acceleration factors for combined parallel imaging-simultaneous multislice-accelerated knee MRI but requires performance validation against external reference standards. <b>OBJECTIVE</b>. The purpose of this study was to validate the clinical efficacy of six-fold-accelerated sub-5-minute 3-T knee MRI using combined threefold parallel imaging (PI) and twofold simultaneous multislice (SMS) acceleration and DL superresolution image reconstruction against arthroscopic surgery. <b>METHODS</b>. Consecutive adult patients with painful knee conditions who underwent sixfold PI-SMS-accelerated DL superresolution 3-T knee MRI and arthroscopic surgery between October 2022 and July 2023 were retrospectively included. Seven fellowship-trained musculoskeletal radiologists independently assessed the MRI studies for image-quality parameters; presence of artifacts; structural visibility (Likert scale: 1 [very bad/severe] to 5 [very good/absent]); and the presence of cruciate ligament tears, collateral ligament tears, meniscal tears, cartilage defects, and fractures. Statistical analyses included kappa-based interreader agreements and diagnostic performance testing. <b>RESULTS</b>. The final sample included 124 adult patients (mean age ± SD, 46 ± 17 years; 79 men, 45 women) who underwent knee MRI and arthroscopic surgery within a median of 28 days (range, 4-56 days). Overall image quality was good to very good (median, 4 [IQR, 4-5]) with very good interreader agreement (κ = 0.86). Motion artifacts were absent (median, 5 [IQR, 5-5]), and image noise was minimal (median, 4 [IQR, 4-5]). Visibility of anatomic structures was very good (median, 5 [IQR, 5-5]). Diagnostic performance for diagnosing arthroscopy-validated structural abnormalities was good to excellent (AUC ≥ 0.81) with at least good interreader agreement (κ ≥ 0.72). The sensitivity, specificity, accuracy, and AUC values were 100%, 99%, 99%, and 0.99 for anterior cruciate ligament tears; 100%, 100%, 100%, and 1.00 for posterior cruciate ligament tears; 90%, 95%, 94%, and 0.93 for medial meniscus tears; 76%, 97%, 90%, and 0.86 for lateral meniscus tears; and 85%, 88%, 88%, and 0.81 for articular cartilage defects, respectively. <b>CONCLUSION</b>. Sixfold PI-SMS-accelerated sub-5-minute DL superresolution 3-T knee MRI has excellent diagnostic performance for detecting internal derangement. <b>CLINICAL IMPACT</b>. Sixfold PI-SMS-accelerated PI-SMS DL superresolution 3-T knee MRI provides high efficiency through short scan times and high diagnostic performance.

Cortical lesions (CLs) have emerged as valuable biomarkers in multiple sclerosis (MS), offering high diagnostic specificity and prognostic relevance. However, their routine clinical integration remains limited due to subtle magnetic resonance imaging (MRI) appearance, challenges in expert annotation, and a lack of standardized automated methods. We propose a comprehensive multi-centric benchmark of CL detection and segmentation in MRI. A total of 656 MRI scans, including clinical trial and research data from four institutions, were acquired at 3T and 7T using MP2RAGE and MPRAGE sequences with expert-consensus annotations. We rely on the self-configuring nnU-Net framework, designed for medical imaging segmentation, and propose adaptations tailored to the improved CL detection. We evaluated model generalization through out-of-distribution testing, demonstrating strong lesion detection capabilities with an F1-score of 0.64 and 0.5 in and out of the domain, respectively. We also analyze internal model features and model errors for a better understanding of AI decision-making. Our study examines how data variability, lesion ambiguity, and protocol differences impact model performance, offering future recommendations to address these barriers to clinical adoption. To reinforce the reproducibility, the implementation and models will be publicly accessible and ready to use at https://github.com/Medical-Image-Analysis-Laboratory/ and https://doi.org/10.5281/zenodo.15911797.

Bronchopulmonary dysplasia (BPD) is a chronic lung disease affecting 35% of extremely low birth weight infants. Defined by oxygen dependence at 36 weeks postmenstrual age, it causes lifelong respiratory complications. However, preventive interventions carry severe risks, including neurodevelopmental impairment, ventilator-induced lung injury, and systemic complications. Therefore, early BPD prognosis and prediction of BPD outcome is crucial to avoid unnecessary toxicity in low risk infants. Admission radiographs of extremely preterm infants are routinely acquired within 24h of life and could serve as a non-invasive prognostic tool. In this work, we developed and investigated a deep learning approach using chest X-rays from 163 extremely low-birth-weight infants ($\leq$32 weeks gestation, 401-999g) obtained within 24 hours of birth. We fine-tuned a ResNet-50 pretrained specifically on adult chest radiographs, employing progressive layer freezing with discriminative learning rates to prevent overfitting and evaluated a CutMix augmentation and linear probing. For moderate/severe BPD outcome prediction, our best performing model with progressive freezing, linear probing and CutMix achieved an AUROC of 0.78 $\pm$ 0.10, balanced accuracy of 0.69 $\pm$ 0.10, and an F1-score of 0.67 $\pm$ 0.11. In-domain pre-training significantly outperformed ImageNet initialization (p = 0.031) which confirms domain-specific pretraining to be important for BPD outcome prediction. Routine IRDS grades showed limited prognostic value (AUROC 0.57 $\pm$ 0.11), confirming the need of learned markers. Our approach demonstrates that domain-specific pretraining enables accurate BPD prediction from routine day-1 radiographs. Through progressive freezing and linear probing, the method remains computationally feasible for site-level implementation and future federated learning deployments.

Phrase grounding, i.e., mapping natural language phrases to specific image regions, holds significant potential for disease localization in medical imaging through clinical reports. While current state-of-the-art methods rely on discriminative, self-supervised contrastive models, we demonstrate that generative text-to-image diffusion models, leveraging cross-attention maps, can achieve superior zero-shot phrase grounding performance. Contrary to prior assumptions, we show that fine-tuning diffusion models with a frozen, domain-specific language model, such as CXR-BERT, substantially outperforms domain-agnostic counterparts. This setup achieves remarkable improvements, with mIoU scores doubling those of current discriminative methods. These findings highlight the underexplored potential of generative models for phrase grounding tasks. To further enhance performance, we introduce Bimodal Bias Merging (BBM), a novel post-processing technique that aligns text and image biases to identify regions of high certainty. BBM refines cross-attention maps, achieving even greater localization accuracy. Our results establish generative approaches as a more effective paradigm for phrase grounding in the medical imaging domain, paving the way for more robust and interpretable applications in clinical practice. The source code and model weights are available at https://github.com/Felix-012/generate_to_ground.

Deep learning has significantly advanced the field of medical image classification, particularly with the adoption of Convolutional Neural Networks (CNNs). Various deep learning frameworks such as Keras, PyTorch and JAX offer unique advantages in model development and deployment. However, their comparative performance in medical imaging tasks remains underexplored. This study presents a comprehensive analysis of CNN implementations across these frameworks, using the PathMNIST dataset as a benchmark. We evaluate training efficiency, classification accuracy and inference speed to assess their suitability for real-world applications. Our findings highlight the trade-offs between computational speed and model accuracy, offering valuable insights for researchers and practitioners in medical image analysis.

Magnetic Resonance Imaging (MRI) is widely recognized as the most reliable tool for detecting tumors due to its capability to produce detailed images that reveal their presence. However, the accuracy of diagnosis can be compromised when human specialists evaluate these images. Factors such as fatigue, limited expertise, and insufficient image detail can lead to errors. For example, small tumors might go unnoticed, or overlap with healthy brain regions could result in misidentification. To address these challenges and enhance diagnostic precision, this study proposes a novel double ensembling framework, consisting of ensembled pre-trained deep learning (DL) models for feature extraction and ensembled fine-tuned hyperparameter machine learning (ML) models to efficiently classify brain tumors. Specifically, our method includes extensive preprocessing and augmentation, transfer learning concepts by utilizing various pre-trained deep convolutional neural networks and vision transformer networks to extract deep features from brain MRI, and fine-tune hyperparameters of ML classifiers. Our experiments utilized three different publicly available Kaggle MRI brain tumor datasets to evaluate the pre-trained DL feature extractor models, ML classifiers, and the effectiveness of an ensemble of deep features along with an ensemble of ML classifiers for brain tumor classification. Our results indicate that the proposed feature fusion and classifier fusion improve upon the state of the art, with hyperparameter fine-tuning providing a significant enhancement over the ensemble method. Additionally, we present an ablation study to illustrate how each component contributes to accurate brain tumor classification.