Periventricular-intraventricular haemorrhage (IVH) is the most prevalent type of neonatal intracranial haemorrhage. It is especially threatening to preterm infants, in whom it is associated with significant morbidity and mortality. Cranial ultrasound has become an important means of screening periventricular IVH in infants. The integration of artificial intelligence with neonatal ultrasound is promising for enhancing diagnostic accuracy, reducing physician workload, and consequently improving periventricular IVH outcomes. The study investigated whether deep learning-based analysis of the cranial ultrasound images of infants could detect and grade periventricular IVH. This multicentre observational study included 1,060 cases and healthy controls from two hospitals. The retrospective modelling dataset encompassed 773 participants from January 2020 to July 2023, while the prospective two-centre validation dataset included 287 participants from August 2023 to January 2024. The periventricular IVH net model, a deep learning model incorporating the convolutional block attention module mechanism, was developed. The model's effectiveness was assessed by randomly dividing the retrospective data into training and validation sets, followed by independent validation with the prospective two-centre data. To evaluate the model, we measured its recall, precision, accuracy, F1-score, and area under the curve (AUC). The regions of interest (ROI) that influenced the detection by the deep learning model were visualised in significance maps, and the t-distributed stochastic neighbour embedding (t-SNE) algorithm was used to visualise the clustering of model detection parameters. The final retrospective dataset included 773 participants (mean (standard deviation (SD)) gestational age, 32.7 (4.69) weeks; mean (SD) weight, 1,862.60 (855.49) g). For the retrospective data, the model's AUC was 0.99 (95% confidence interval (CI), 0.98-0.99), precision was 0.92 (0.89-0.95), recall was 0.93 (0.89-0.95), and F1-score was 0.93 (0.90-0.95). For the prospective two-centre validation data, the model's AUC was 0.961 (95% CI, 0.94-0.98) and accuracy was 0.89 (95% CI, 0.86-0.92). The two-centre prospective validation results of the periventricular IVH net model demonstrated its tremendous potential for paediatric clinical applications. Combining artificial intelligence with paediatric ultrasound can enhance the accuracy and efficiency of periventricular IVH diagnosis, especially in primary hospitals or community hospitals.

In recent years, artificial intelligence has been increasingly applied in the field of medical imaging. Among these applications, fundus image analysis presents special challenges, including small lesion areas in certain fundus diseases and subtle inter-disease differences, which can lead to reduced prediction accuracy and overfitting in the models. To address these challenges, this paper proposes the Transformer-based model SwinECAT, which combines the Shifted Window (Swin) Attention with the Efficient Channel Attention (ECA) Attention. SwinECAT leverages the Swin Attention mechanism in the Swin Transformer backbone to effectively capture local spatial structures and long-range dependencies within fundus images. The lightweight ECA mechanism is incorporated to guide the SwinECAT's attention toward critical feature channels, enabling more discriminative feature representation. In contrast to previous studies that typically classify fundus images into 4 to 6 categories, this work expands fundus disease classification to 9 distinct types, thereby enhancing the granularity of diagnosis. We evaluate our method on the Eye Disease Image Dataset (EDID) containing 16,140 fundus images for 9-category classification. Experimental results demonstrate that SwinECAT achieves 88.29\% accuracy, with weighted F1-score of 0.88 and macro F1-score of 0.90. The classification results of our proposed model SwinECAT significantly outperform the baseline Swin Transformer and multiple compared baseline models. To our knowledge, this represents the highest reported performance for 9-category classification on this public dataset.

Foundation models have demonstrated remarkable potential in medical domain. However, their application to complex cardiovascular diagnostics remains underexplored. In this paper, we present Cardiac-CLIP, a multi-modal foundation model designed for 3D cardiac CT images. Cardiac-CLIP is developed through a two-stage pre-training strategy. The first stage employs a 3D masked autoencoder (MAE) to perform self-supervised representation learning from large-scale unlabeled volumetric data, enabling the visual encoder to capture rich anatomical and contextual features. In the second stage, contrastive learning is introduced to align visual and textual representations, facilitating cross-modal understanding. To support the pre-training, we collect 16641 real clinical CT scans, supplemented by 114k publicly available data. Meanwhile, we standardize free-text radiology reports into unified templates and construct the pathology vectors according to diagnostic attributes, based on which the soft-label matrix is generated to supervise the contrastive learning process. On the other hand, to comprehensively evaluate the effectiveness of Cardiac-CLIP, we collect 6,722 real-clinical data from 12 independent institutions, along with the open-source data to construct the evaluation dataset. Specifically, Cardiac-CLIP is comprehensively evaluated across multiple tasks, including cardiovascular abnormality classification, information retrieval and clinical analysis. Experimental results demonstrate that Cardiac-CLIP achieves state-of-the-art performance across various downstream tasks in both internal and external data. Particularly, Cardiac-CLIP exhibits great effectiveness in supporting complex clinical tasks such as the prospective prediction of acute coronary syndrome, which is notoriously difficult in real-world scenarios.

Ultrasound is widely used in clinical care, yet standard deep learning methods often struggle with full video analysis due to non-standardized acquisition and operator bias. We offer a new perspective on ultrasound video analysis through implicit neural representations (INRs). We build on Functa, an INR framework in which each image is represented by a modulation vector that conditions a shared neural network. However, its extension to the temporal domain of medical videos remains unexplored. To address this gap, we propose VidFuncta, a novel framework that leverages Functa to encode variable-length ultrasound videos into compact, time-resolved representations. VidFuncta disentangles each video into a static video-specific vector and a sequence of time-dependent modulation vectors, capturing both temporal dynamics and dataset-level redundancies. Our method outperforms 2D and 3D baselines on video reconstruction and enables downstream tasks to directly operate on the learned 1D modulation vectors. We validate VidFuncta on three public ultrasound video datasets -- cardiac, lung, and breast -- and evaluate its downstream performance on ejection fraction prediction, B-line detection, and breast lesion classification. These results highlight the potential of VidFuncta as a generalizable and efficient representation framework for ultrasound videos. Our code is publicly available under https://github.com/JuliaWolleb/VidFuncta_public.

Medical image-language pre-training aims to align medical images with clinically relevant text to improve model performance on various downstream tasks. However, existing models often struggle with the variability and ambiguity inherent in medical data, limiting their ability to capture nuanced clinical information and uncertainty. This work introduces an uncertainty-aware medical image-text pre-training model that enhances generalization capabilities in medical image analysis. Building on previous methods and focusing on Chest X-Rays, our approach utilizes structured text reports generated by a large language model (LLM) to augment image data with clinically relevant context. These reports begin with a definition of the disease, followed by the `appearance' section to highlight critical regions of interest, and finally `observations' and `verdicts' that ground model predictions in clinical semantics. By modeling both inter- and intra-modal uncertainty, our framework captures the inherent ambiguity in medical images and text, yielding improved representations and performance on downstream tasks. Our model demonstrates significant advances in medical image-text pre-training, obtaining state-of-the-art performance on multiple downstream tasks.

Vertebral recompression after percutaneous kyphoplasty (PKP) for osteoporotic vertebral compression fractures (OVCFs) may lead to recurrent pain, deformity, and neurological impairment, compromising prognosis and quality of life. To identify independent risk factors for postoperative recompression and develop predictive models for risk assessment. We retrospectively analyzed 284 OVCF patients treated with PKP, grouped by recompression status. Predictors were screened using univariate and correlation analyses. Multicollinearity was assessed using variance inflation factor (VIF). A multivariable logistic regression model was constructed and validated via 10-fold cross-validation and temporal validation. Five independent predictors were identified: incomplete anterior cortex (odds ratio [OR] = 9.38), high paravertebral muscle fat infiltration (OR = 218.68), low vertebral CT value (OR = 0.87), large Cobb change (OR = 1.45), and high vertebral height recovery rate (OR = 22.64). The logistic regression model achieved strong performance: accuracy 97.67%, precision 97.06%, recall 97.06%, F1 score 97.06%, specificity 98.08%, area under the receiver operating characteristic curve (AUC) 0.998. Machine learning models (e.g., random forest) were also evaluated but did not outperform logistic regression in accuracy or interpretability. Five imaging-based predictors of vertebral recompression were identified. The logistic regression model showed excellent predictive accuracy and generalizability, supporting its clinical utility for early risk stratification and personalized decision-making in OVCF patients undergoing PKP.

Deep learning (DL) demonstrates high sensitivity but low specificity in lung cancer (LC) detection during CT screening, and the seven Tumor-associated antigens autoantibodies (7-TAAbs), known for its high specificity in LC, was employed to improve the DL's specificity for the efficiency of LC screening in China. To develop and evaluate a risk model combining 7-TAAbs test and DL scores for diagnosing LC with pulmonary lesions < 70 mm. Four hundreds and six patients with 406 lesions were enrolled and assigned into training set (n = 313) and test set (n = 93) randomly. The malignant lesions were defined as those lesions with high malignant risks by DL or those with positive expression of 7-TAAbs panel. Model performance was assessed using the area under the receiver operating characteristic curves (AUC). In the training set, the AUCs for DL, 7-TAAbs, combined model (DL and 7-TAAbs) and combined model (DL or 7-TAAbs) were 0.771, 0.638, 0.606, 0.809 seperately. In the test set, the combined model (DL or 7-TAAbs) achieved achieved the highest sensitivity (82.6%), NPV (81.8%) and accuracy (79.6%) among four models, and the AUCs of DL model, 7-TAAbs model, combined model (DL and 7-TAAbs), and combined model (DL or 7-TAAbs) were 0.731, 0.679, 0.574, and 0.794, respectively. The 7-TAAbs test significantly enhances DL performance in predicting LC with pulmonary leisons < 70 mm in China.

Physiological brain aging is associated with cognitive impairment and neuroanatomical changes. Brain age prediction of routine clinical 2D brain MRI scans were understudied and often unsuccessful. We developed a novel brain age prediction framework for clinical 2D T1-weighted MRI scans using a deep learning-based model trained with research grade 3D MRI scans mostly from publicly available datasets (N = 8681; age = 51.76 ± 21.74). Our model showed accurate and fast brain age prediction on clinical 2D MRI scans from cognitively unimpaired (CU) subjects (N = 175) with MAE of 2.73 years after age bias correction (Pearson's r = 0.918). Brain age gap of Alzheimer's disease (AD) subjects was significantly greater than CU subjects (p < 0.001) and increase in brain age gap was associated with disease progression in both AD (p < 0.05) and Parkinson's disease (p < 0.01). Our framework can be extended to other MRI modalities and potentially applied to routine clinical examinations, enabling early detection of structural anomalies and improve patient outcome.

Alzheimer's disease is a progressive neurological disorder that profoundly affects cognitive functions and daily activities. Rapid and precise identification is essential for effective intervention and improved patient outcomes. This research introduces an innovative hybrid filtering approach with a deep transfer learning model for detecting Alzheimer's disease utilizing brain imaging data. The hybrid filtering method integrates the Adaptive Non-Local Means filter with a Sharpening filter for image preprocessing. Furthermore, the deep learning model used in this study is constructed on the EfficientNetV2B3 architecture, augmented with additional layers and fine-tuning to guarantee effective classification among four categories: Mild, moderate, very mild, and non-demented. The work employs Grad-CAM++ to enhance interpretability by localizing disease-relevant characteristics in brain images. The experimental assessment, performed on a publicly accessible dataset, illustrates the ability of the model to achieve an accuracy of 99.45%. These findings underscore the capability of sophisticated deep learning methodologies to aid clinicians in accurately identifying Alzheimer's disease.

Assessing MGMT promoter methylation is crucial for determining appropriate glioblastoma therapy. Previous studies have focused on intratumoral regions, overlooking the peritumoral area. This study aimed to develop a radiomic model using MRI-derived features from both regions. We included 96 glioblastoma patients randomly allocated to training and testing sets. Radiomic features were extracted from intratumoral and peritumoral regions. We constructed and compared radiomic models based on intratumoral, peritumoral, and combined features. Model performance was evaluated using the area under the receiver-operating characteristic curve (AUC). The combined radiomic model achieved an AUC of 0.814 (95% CI: 0.767-0.862) in the training set and 0.808 (95% CI: 0.736-0.859) in the testing set, outperforming models based on intratumoral or peritumoral features alone. Calibration and decision curve analyses demonstrated excellent model fit and clinical utility. The radiomic model incorporating both intratumoral and peritumoral features shows promise in differentiating MGMT methylation status, potentially informing clinical treatment strategies for glioblastoma.