Brain aneurysm detection models, both in the literature and in industry, continue to lack generalizability during external validation, limiting clinical adoption. This challenge is largely due to extensive exclusion criteria during training data selection. The authors developed the first model to achieve generalizability using novel methodological approaches. Computed tomography angiography (CTA) scans from 2004 to 2023 at the study institution were used for model training, including untreated unruptured intracranial aneurysms without extensive cerebrovascular disease. External validation used digital subtraction angiography-verified CTAs from an international center, while prospective validation occurred at the internal institution over 9 months. A public web platform was created for further model validation. A total of 2194 CTA scans were used for this study. One thousand five hundred eighty-seven patients and 1920 aneurysms with a mean size of 5.3 ± 3.7 mm were included in the training cohort. The mean age of the patients was 69.7 ± 14.9 years, and 1203 (75.8%) were female. The model achieved a training Dice score of 0.88 and a validation Dice score of 0.76. Prospective internal validation on 304 scans yielded a lesion-level (LL) sensitivity of 82.5% (95% CI: 75.5-87.9) and specificity of 89.6 (95% CI: 84.5-93.2). External validation on 303 scans demonstrated an on-par LL sensitivity and specificity of 83.5% (95% CI: 75.1-89.4) and 92.9% (95% CI: 88.8-95.6), respectively. Radiologist LL sensitivity from the external center was 84.5% (95% CI: 76.2-90.2), and 87.5% of the missed aneurysms were detected by the model. The authors developed the first publicly testable artificial intelligence model for aneurysm detection on CTA scans, demonstrating generalizability and state-of-the-art performance in external validation. The model addresses key limitations of previous efforts and enables broader validation through a web-based platform.

Study designRetrospective cohort study.ObjectivesDevelop and validate a model combining clinical data, deep learning radiomics (DLR), and radiomic features from lumbar CT and multisequence MRI to predict high-risk patients for adjacent segment degeneration (ASDeg) post-lumbar fusion.MethodsThis study included 305 patients undergoing preoperative CT and MRI for lumbar fusion surgery, divided into training (n = 192), internal validation (n = 83), and external test (n = 30) cohorts. Vision Transformer 3D-based deep learning model was developed. LASSO regression was used for feature selection to establish a logistic regression model. ASDeg was defined as adjacent segment degeneration during radiological follow-up 6 months post-surgery. Fourteen machine learning algorithms were evaluated using ROC curves, and a combined model integrating clinical variables was developed.ResultsAfter feature selection, 21 radiomics, 12 DLR, and 3 clinical features were selected. The linear support vector machine algorithm performed best for the radiomic model, and AdaBoost was optimal for the DLR model. A combined model using these and clinical features was developed, with the multi-layer perceptron as the most effective algorithm. The areas under the curve for training, internal validation, and external test cohorts were 0.993, 0.936, and 0.835, respectively. The combined model outperformed the combined predictions of 2 surgeons.ConclusionsThis study developed and validated a combined model integrating clinical, DLR and radiomic features, demonstrating high predictive performance for identifying high-risk ASDeg patients post-lumbar fusion based on clinical data, CT, and MRI. The model could potentially reduce ASDeg-related revision surgeries, thereby reducing the burden on the public healthcare.

Software programs leveraging artificial intelligence to detect vessel occlusions are now widely available to aid in stroke triage. Given their proprietary use, there is a surprising lack of information regarding how the software works, who is using the software, and their performance in an unbiased real-world setting. In this educational review of automated vessel occlusion software, we discuss emerging evidence of their utility, underlying algorithms, real-world diagnostic performance, and limitations. The intended audience includes specialists in stroke care in neurology, emergency medicine, radiology, and neurosurgery. Practical tips for onboarding and utilization of this technology are provided based on the multidisciplinary experience of the authorship team.

Novel antibody-drug conjugates highlight the benefits for breast cancer patients with low human epidermal growth factor receptor 2 (HER2) expression. This study aims to develop and validate a Vision Transformer (ViT) model based on dynamic contrast-enhanced MRI (DCE-MRI) to classify HER2-zero, -low, and -positive breast cancer patients and to explore its interpretability. The model is trained and validated on early enhancement MRI images from 708 patients in the FUSCC cohort and tested on 80 and 101 patients in the GFPH cohort and FHCMU cohort, respectively. The ViT model achieves AUCs of 0.80, 0.73, and 0.71 in distinguishing HER2-zero from HER2-low/positive tumors across the validation set of the FUSCC cohort and the two external cohorts. Furthermore, the model effectively classifies HER2-low and HER2-positive cases, with AUCs of 0.86, 0.80, and 0.79. Transcriptomics analysis identifies significant biological differences between HER2-low and HER2-positive patients, particularly in immune-related pathways, suggesting potential therapeutic targets. Additionally, Cox regression analysis demonstrates that the prediction score is an independent prognostic factor for overall survival (HR, 2.52; p = 0.007). These findings provide a non-invasive approach for accurately predicting HER2 expression, enabling more precise patient stratification to guide personalized treatment strategies. Further prospective studies are warranted to validate its clinical utility.

Magnetic resonance continues to evolve and advance as a critical imaging modality for disease diagnosis and monitoring. Hardware and software advances continue to propel this modality to the forefront of the field of diagnostic imaging. Next generation MR contrast media, specifically gadolinium chelates with improved relaxivity and stability (relative to the provided contrast effect), have emerged providing a further boost to the field. Concern regarding gadolinium deposition in the body with primarily the weaker gadolinium chelates (which have been now removed from the market, at least in Europe) continues to be at the forefront of clinicians' minds. This has driven renewed interest in possible development of manganese-based contrast media. The development of photon counting CT and its clinical introduction have made possible a further major advance in CT image quality, along with the potential for decreasing radiation dose. The possibility of major clinical advances in thoracic, cardiac, and musculoskeletal imaging were first recognized, with its broader impact - across all organ systems - now also recognized. The utility of routine acquisition (without penalty in time or radiation dose) of full spectral multi-energy data is now also being recognized as an additional major advance made possible by photon counting CT. Artificial intelligence is now being used in the background across most imaging platforms and modalities, making possible further advances in imaging technique and image quality, although this field is nowhere yet near to realizing its full potential. And last, but not least, the field of ultrasound is on the cusp of further major advances in availability (with development of very low-cost systems) and a possible new generation of microbubble contrast media.

Segmentation of coronary arteries in Coronary Computed Tomography Angiography (CCTA) images is crucial for diagnosing coronary artery disease (CAD), but remains challenging due to small artery size, uneven contrast distribution, and issues like over-segmentation or omission. The aim of this study is to improve coronary artery segmentation in CCTA images using both conventional and deep learning techniques. We propose MHASegNet, a lightweight network for coronary artery segmentation, combined with a tailored refinement method. MHASegNet employs multi-scale hybrid attention to capture global and local features, and integrates a 3D context anchor attention module to focus on key coronary artery structures while suppressing background noise. An iterative, region-growth-based refinement addresses crown breaks and reduces false alarms. We evaluated the method on an in-house dataset of 90 subjects and two public datasets with 1060 subjects. MHASegNet, coupled with tailored refinement, outperforms state-of-the-art algorithms, achieving a Dice Similarity Coefficient (DSC) of 0.867 on the in-house dataset, 0.875 on the ASOCA dataset, and 0.827 on the ImageCAS dataset. The tailored refinement significantly reduces false positives and resolves most discontinuities, even for other networks. MHASegNet and the tailored refinement may aid in diagnosing and quantifying CAD following further validation.

Multimodal large language models (MLLMs) have shown great potential in general domains but perform poorly in some specific domains due to a lack of domain-specific data, such as image-text data or vedio-text data. In some specific domains, there is abundant graphic and textual data scattered around, but lacks standardized arrangement. In the field of medical ultrasound, there are ultrasonic diagnostic books, ultrasonic clinical guidelines, ultrasonic diagnostic reports, and so on. However, these ultrasonic materials are often saved in the forms of PDF, images, etc., and cannot be directly used for the training of MLLMs. This paper proposes a novel image-text reasoning supervised fine-tuning data generation pipeline to create specific domain quadruplets (image, question, thinking trace, and answer) from domain-specific materials. A medical ultrasound domain dataset ReMUD is established, containing over 45,000 reasoning and non-reasoning supervised fine-tuning Question Answering (QA) and Visual Question Answering (VQA) data. The ReMUD-7B model, fine-tuned on Qwen2.5-VL-7B-Instruct, outperforms general-domain MLLMs in medical ultrasound field. To facilitate research, the ReMUD dataset, data generation codebase, and ReMUD-7B parameters will be released at https://github.com/ShiDaizi/ReMUD, addressing the data shortage issue in specific domain MLLMs.

Optical Coherence Tomography (OCT) provides high-resolution, 3D, and non-invasive visualization of retinal layers in vivo, serving as a critical tool for lesion localization and disease diagnosis. However, its widespread adoption is limited by equipment costs and the need for specialized operators. In comparison, 2D color fundus photography offers faster acquisition and greater accessibility with less dependence on expensive devices. Although generative artificial intelligence has demonstrated promising results in medical image synthesis, translating 2D fundus images into 3D OCT images presents unique challenges due to inherent differences in data dimensionality and biological information between modalities. To advance generative models in the fundus-to-3D-OCT setting, the Asia Pacific Tele-Ophthalmology Society (APTOS-2024) organized a challenge titled Artificial Intelligence-based OCT Generation from Fundus Images. This paper details the challenge framework (referred to as APTOS-2024 Challenge), including: the benchmark dataset, evaluation methodology featuring two fidelity metrics-image-based distance (pixel-level OCT B-scan similarity) and video-based distance (semantic-level volumetric consistency), and analysis of top-performing solutions. The challenge attracted 342 participating teams, with 42 preliminary submissions and 9 finalists. Leading methodologies incorporated innovations in hybrid data preprocessing or augmentation (cross-modality collaborative paradigms), pre-training on external ophthalmic imaging datasets, integration of vision foundation models, and model architecture improvement. The APTOS-2024 Challenge is the first benchmark demonstrating the feasibility of fundus-to-3D-OCT synthesis as a potential solution for improving ophthalmic care accessibility in under-resourced healthcare settings, while helping to expedite medical research and clinical applications.

Ultrasound (US) imaging is a promising modality for real-time monitoring of robotic capsule endoscopes navigating through the gastrointestinal (GI) tract. It offers high temporal resolution and safety but is limited by a narrow field of view, low visibility in gas-filled regions and challenges in detecting out-of-plane motions. This work addresses these issues by proposing a novel robotic ultrasound tracking system capable of long-distance 3D tracking and active re-localization when the capsule is lost due to motion or artifacts. We develop a hybrid deep learning-based tracking framework combining convolutional neural networks (CNNs) and a transformer backbone. The CNN component efficiently encodes spatial features, while the transformer captures long-range contextual dependencies in B-mode US images. This model is integrated with a robotic arm that adaptively scans and tracks the capsule. The system's performance is evaluated using ex vivo colon phantoms under varying imaging conditions, with physical perturbations introduced to simulate realistic clinical scenarios. The proposed system achieved continuous 3D tracking over distances exceeding 90 cm, with a mean centroid localization error of 1.5 mm and over 90% detection accuracy. We demonstrated 3D tracking in a more complex workspace featuring two curved sections to simulate anatomical challenges. This suggests the strong resilience of the tracking system to motion-induced artifacts and geometric variability. The system maintained real-time tracking at 9-12 FPS and successfully re-localized the capsule within seconds after tracking loss, even under gas artifacts and acoustic shadowing. This study presents a hybrid CNN-transformer system for automatic, real-time 3D ultrasound tracking of capsule robots over long distances. The method reliably handles occlusions, view loss and image artifacts, offering millimeter-level tracking accuracy. It significantly reduces clinical workload through autonomous detection and re-localization. Future work includes improving probe-tissue interaction handling and validating performance in live animal and human trials to assess physiological impacts.

Brain tumor classification is critical for therapeutic applications that benefit from computer-aided diagnostics. Misdiagnosing a brain tumor can significantly reduce a patient's chances of survival, as it may lead to ineffective treatments. This study proposes a novel approach for classifying brain tumors in MRI images using Transfer Learning (TL) with state-of-the-art deep learning models: AlexNet, MobileNetV2, and GoogleNet. Unlike previous studies that often focus on a single model, our work comprehensively compares these architectures, fine-tuned specifically for brain tumor classification. We utilize a publicly available dataset of 4,517 MRI scans, consisting of three prevalent types of brain tumors-glioma (1,129 images), meningioma (1,134 images), and pituitary tumors (1,138 images)-as well as 1,116 images of normal brains (no tumor). Our approach addresses key research gaps, including class imbalance, through data augmentation and model efficiency, leveraging lightweight architectures like MobileNetV2. The GoogleNet model achieves the highest classification accuracy of 99.2%, outperforming previous studies using the same dataset. This demonstrates the potential of our approach to assist physicians in making rapid and precise decisions, thereby improving patient outcomes. The results highlight the effectiveness of TL in medical diagnostics and its potential for real-world clinical deployment. This study advances the field of brain tumor classification and provides a robust framework for future research in medical image analysis.