Background:Accurate and interpretable brain tumor classification from MRI images remains a key challenge in medical image analysis, particularly when using publicly available datasets of moderate size. Objectives:This study investigates the performance of a ConvNeXt-Tiny based framework for four-class brain tumor classification glioma, meningioma, pituitary tumor, and no tumor and compares it with established convolutional architectures. Materials and methods:Using transfer learning and identical experimental settings, ConvNeXt-Tiny was evaluated against DenseNet169, Xception, MobileNetV3-Large, CNN+DenseNet169, and ResNet50. Standard evaluation metrics (accuracy, precision, recall, and F1-score) were used, and Grad-CAM was applied to visualize model attention for interpretability. Generalization was further assessed using an independent dataset. Results:ConvNeXt-Tiny achieved high overall performance (accuracy = 0.9924, F1-score = 0.9918), comparable to DenseNet169 and Xception but with lower computational cost. The model maintained stable learning behavior, minimal overfitting, and consistent accuracy on unseen data. Grad-CAM visualizations confirmed that the network focused on clinically relevant tumor regions, improving transparency and reliability of predictions. Conclusion:ConvNeXt-Tiny provides a strong and efficient baseline for interpretable brain tumor classification, balancing accuracy and computational efficiency. While the results are promising, differences among recent architectures were modest, and clinical validation using multi-center MRI datasets is necessary to confirm broader applicability.

Determining whether two sets of images belong to the same or different distributions or domains is a crucial task in modern medical image analysis and deep learning; for example, to evaluate the output quality of image generative models. Currently, metrics used for this task either rely on the (potentially biased) choice of some downstream task, such as segmentation, or adopt task-independent perceptual metrics (e.g., Fréchet Inception Distance/FID) from natural imaging, which we show insufficiently capture anatomical features. To this end, we introduce a new perceptual metric tailored for medical images, FRD (Fréchet Radiomic Distance), which utilizes standardized, clinically meaningful, and interpretable image features. We show that FRD is superior to other image distribution metrics for a range of medical imaging applications, including out-of-domain (OOD) detection, the evaluation of image-to-image translation (by correlating more with downstream task performance as well as anatomical consistency and realism), and the evaluation of unconditional image generation. Moreover, FRD offers additional benefits such as stability and computational efficiency at low sample sizes, sensitivity to image corruptions and adversarial attacks, feature interpretability, and correlation with radiologist-perceived image quality. Additionally, we address key gaps in the literature by presenting an extensive framework for the multifaceted evaluation of image similarity metrics in medical imaging-including the first large-scale comparative study of generative models for medical image translation-and release an accessible codebase to facilitate future research. Our results are supported by thorough experiments spanning a variety of datasets, modalities, and downstream tasks, highlighting the broad potential of FRD for medical image analysis.

DAUS-Net: Toward Ultrasound Scanner-Agnostic Domain Generalized Robust and Accurate Segmentation.

Infection prevention and control (IPC) is essential in healthcare education, including medical imaging disciplines such as computed tomography (CT) and nuclear medicine (NM). Drug administration, including contrast media (CM) and radiopharmaceuticals, poses an increased infection risk. The aim of this study was to explore IPC curricula related to drug administration in medical imaging programs. An online cross-sectional survey was distributed to program directors from universities offering medical imaging programs in Australia, the United Kingdom (UK), New Zealand, and Ireland. The survey explored curriculum content, teaching approaches, resource utilisation, and challenges in delivering IPC education. Data were analysed using descriptive statistics, non-parametric tests, and qualitative content analysis. Twenty participants from Australia and the UK completed the survey. Most programs include practical activities, simulations, and multimedia resources. IPC curricula are routinely reviewed and aligned with national guidelines. Key challenges include limited resources and restricted clinical access. Participants highlighted collaboration with healthcare professionals, regular curriculum updates, and innovative teaching technologies as effective strategies to address these issues. Experiential learning through case studies and simulations was viewed as essential for developing practical IPC skills and supporting knowledge retention. Medical imaging programs demonstrate a strong commitment to IPC education through diverse active learning strategies. Addressing resource limitations and ensuring ongoing curriculum updates via collaboration, innovation, and experiential learning are crucial for sustainable IPC training that protects both patients and healthcare professionals. The findings emphasize the need for medical imaging departments to prioritise ongoing IPC training, incorporating simulation, case-based learning, and regular alignment with national standards. Stronger collaboration between universities and clinical sites can improve student preparedness for drug administration and enhance patient safety.

Bridging the green gap: aligning rwandan medical imaging education with national and global sustainability goals.

Federated learning (FL) enables collaborative medical image analysis across decentralized institutions while preserving data privacy. However, real-world deployment faces critical challenges: modality heterogeneity, where clients possess incomplete or varying medical image modalities, and cross-disease generalization, requiring models to adapt to unseen pathologies with anatomical consistency. Existing methods like FedAvg, FedProx, IOP-FL, and PntTranForFL assume uniform modality availability and static client participation, leading to poor performance under realistic clinical constraints. We propose TADynFed, a novel framework for the Heterogeneous Federated Learning (HFL) paradigm, which addresses both data and modality heterogeneity through, a tissue-aware disentanglement strategy that decouples modality-tailored and modality-shared features, a dynamic prototype memory bank for missing modality compensation, an adaptive aggregation mechanism that accounts for client reliability and tenure. We evaluate TADynFed using multi-disease MRI datasets including BraTS21 along with cross-domain imaging data from CheXpert (chest X-ray) and Hep-2 (microscopy), simulating a 13-client FL environment. The proposed framework achieves an average mDice score of 66.03%, significantly outperforming baseline methods such as PointTransformerFL at 58.60% and FedAvg at 53.06%. It also attains the lowest boundary alignment error with an ASD of 1.85 mm and HD95 of 8.70 mm, surpassing existing approaches by notable margins. In terms of calibration stability, TADynFed records an ECE of 0.09, indicating superior confidence reliability. Furthermore, it demonstrates high communication efficiency with only 76 MB of data exchanged per round, compared to 95-105 MB in other frameworks. These results validate TADynFed's ability to maintain high segmentation accuracy, boundary precision, and calibration stability while minimizing bandwidth usage. Outperforming existing frameworks by significant margins, TADynFed demonstrates robust boundary alignment, superior calibration, and efficient communication. It also exhibits strong cross-disease transferability without retraining. By integrating structured representation decomposition, prototype-guided fusion, and client-adaptive learning, TADynFed establishes a new benchmark in realistic, heterogeneous, and mix-modal federated medical imaging systems. Code supporting this study TADynFed.

Enhanced Multi-Scale Adaptive Binary Patterns for Robust Texture Classification

Variational autoencoders for anatomy-specific insights in medical imaging: A systematic review

Automatic anatomical landmark localization in medical imaging requires not just accurate predictions but reliable uncertainty quantification for effective clinical decision support. Current uncertainty quantification approaches often fall short, particularly when combined with normality assumptions, systematically underestimating total predictive uncertainty. This paper introduces conformal prediction as a framework for reliable uncertainty quantification in anatomical landmark localization, addressing a critical gap in automatic landmark localization. We present two novel approaches guaranteeing finite-sample validity for multi-output prediction: multi-output regression-as-classification conformal prediction (M-R2CCP) and its variant multi-output regression to classification conformal prediction set to region (M-R2C2R). Unlike conventional methods that produce axis-aligned hyperrectangular or ellipsoidal regions, our approaches generate flexible, non-convex prediction regions that better capture the underlying uncertainty structure of landmark predictions. Through extensive empirical evaluation across multiple 2D and 3D datasets, we demonstrate that our methods consistently outperform existing multi-output conformal prediction approaches in both validity and efficiency. This work represents a significant advancement in reliable uncertainty estimation for anatomical landmark localization, providing clinicians with trustworthy confidence measures for their diagnoses. While developed for medical imaging, these methods show promise for broader applications in multi-output regression problems.

In recent years, the rapid advancement and application of deep learning in medical imaging have demonstrated its effectiveness in reducing physicians’ workload and lowering the risk of misdiagnosis in pathological spine diagnosis. Nevertheless, deep learning–based models for pathological spine diagnosis have not yet matured to the level required for clinical deployment. Several challenges contribute to this limitation. First, the availability of spinal X-ray images for training is limited, and the class distribution of samples is often imbalanced. Second, conventional deep learning models rely on convolutional kernels that primarily capture local features in X-ray images, while overlooking the global morphological characteristics of the spine. To address these issues, we propose ViTST, a Vision Transformer (ViT)–based model with a self-supervised learning task for scoliosis classification. ViTST incorporates a masked strategy–based self-supervised pretext task to mitigate the challenges posed by limited training data and leverages the ViT architecture to capture global structural features of spinal X-ray images. This design enables more effective modeling of inter-regional relationships and variations within the spine. Moreover, by jointly optimizing reconstruction loss and cross-entropy loss, ViTST learns robust image representations even from relatively small datasets. In addition, we introduce a healthcare Internet of Medical Things (IoMT) architecture to enable the practical deployment of ViTST in clinical environments. Through this IoMT platform, clinicians can monitor patients’ conditions in real time and adapt treatment plans dynamically, thereby enhancing clinical decision-making and accelerating patient recovery. Finally, we conducted extensive experiments on a real-world pathological spine image dataset to validate the effectiveness of the proposed model. Experimental results demonstrate that ViTST achieved a Precision of 0.975, an Accuracy of 0.979, and an F1-score of 0.975, confirming its strong potential for application in clinical practice.

Deep feature-based approaches for brain tumor classification and segmentation in medical imaging

AdaptScanDet: Deformable mamba with multi-scan interaction for heterogeneous lesion detection in medical imaging

In medical imaging, diffusion models have shown great potential for synthetic image generation. However, these approaches often lack interpretable correspondence between generated and real images and can create anatomically implausible structures or illusions. To address these limitations, we propose the Deformation-Recovery Diffusion Model (DRDM), a novel diffusion-based generative model that emphasizes morphological transformation through deformation fields rather than direct image synthesis. DRDM introduces a topology-preserving deformation field generation strategy, which randomly samples and integrates multi-scale Deformation Velocity Fields (DVFs). DRDM is trained to learn to recover unrealistic deformation components, thus restoring randomly deformed images to a realistic distribution. This formulation enables the generation of diverse yet anatomically plausible deformations that preserve structural integrity, thereby improving data augmentation and synthesis for downstream tasks such as few-shot learning and image registration. Experiments on cardiac Magnetic Resonance Imaging and pulmonary Computed Tomography show that DRDM is capable of creating diverse, large-scale deformations, while maintaining anatomical plausibility of deformation fields. Additional evaluations on 2D image segmentation and 3D image registration tasks indicate notable performance gains, underscoring DRDM's potential to enhance both image manipulation and generative modeling in medical imaging applications. The project page: https://jianqingzheng.github.io/def_diff_rec/.

The emergence of Intent-based Networking(IBN)-enabled Healthcare Internet of Things (H-IoT) environments brings new challenges and opportunities for deploying intelligent medical image compression systems in real-time and privacy-sensitive scenarios. However, most existing medical image compression approaches are manually designed and optimized without accounting for the high dimensionality of medical data and the strict deployment constraints in heterogeneous IBN environments, resulting in suboptimal performance and limited adaptability. To address these challenges, we propose a novel implicit neural representation (INR)-based framework for automated medical image compression, leveraging the powerful continuous signal modeling capabilities of INRs to achieve high fidelity on high-dimensional medical images while maintaining compactness and adaptability. Our framework integrates an evolutionary architecture search strategy with privacy-constrained optimization and parameter quantization, enabling the architectures that balance reconstruction quality, latency, communication cost, and privacy. To validate the effectiveness of the framework, we design and conduct comprehensive simulation experiments in realistic multi-hospital IBN environments. The results demonstrate that our INR-based designs outperform traditional and data-driven baselines in both objective metrics and deployment feasibility under constrained resources.

Mapping knowledge landscapes and emerging trends in AI for coronary artery disease imaging biomarkers: A bibliometric and visualization analysis.

Learning with less: A survey of deep learning in medical imaging under varying supervision levels.

Robust non-rigid image-to-patient registration for contactless dynamic thoracic tumor localization using recursive deformable diffusion models.

Generation of virtual abdominal aortic aneurysm shape evolution using a physics-based G&R model and its applications to aneurysm growth prediction.

AMSA-YOLO: Real-time object detection with adaptive multi-scale attention mechanism.

Artificial intelligence model for cardiovascular disease risk prediction in breast cancer patients using electronic health records and computed tomography scans.