Cardiovascular disease is a leading global cause of death, requiring accurate heart segmentation for diagnosis and surgical planning. Deep learning methods have been demonstrated to achieve superior performances in cardiac structures segmentation. However, there are still limitations in 3D whole heart segmentation, such as inadequate spatial context modeling, difficulty in capturing long-distance dependencies, high computational complexity, and limited representation of local high-level semantic information. To tackle the above problems, we propose a lightweight Pyramid Pooling Transformer-CNN (P2TC) network for accurate 3D whole heart segmentation. The proposed architecture comprises a dual encoder-decoder structure with a 3D pyramid pooling Transformer for multi-scale information fusion and a lightweight large-kernel Convolutional Neural Network (CNN) for local feature extraction. The decoder has two branches for precise segmentation and contextual residual handling. The first branch is used to generate segmentation masks for pixel-level classification based on the features extracted by the encoder to achieve accurate segmentation of cardiac structures. The second branch highlights contextual residuals across slices, enabling the network to better handle variations and boundaries. Extensive experimental results on the Multi-Modality Whole Heart Segmentation (MM-WHS) 2017 challenge dataset demonstrate that P2TC outperforms the most advanced methods, achieving the Dice scores of 92.6% and 88.1% in Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) modalities respectively, which surpasses the baseline model by 1.5% and 1.7%, and achieves state-of-the-art segmentation results.

Cervical length (CL) measurement using transvaginal ultrasound is an effective screening tool to assess the risk of preterm birth. An adequate assessment of CL is crucial, however, manual sonographic CL measurement is highly operator-dependent and cumbersome. Therefore, a reliable and reproducible automatic method for CL measurement is in high demand to reduce inter-rater variability and improve workflow. Despite the increasing use of artificial intelligence techniques in ultrasound, applying deep learning (DL) to analyze ultrasound images of the cervix remains a challenge due to low signal-to-noise ratios and difficulties in capturing the cervical canal, which appears as a thin line and with extremely low contrast against the surrounding tissues. To address these challenges, we have developed CL-Net, a novel DL network that incorporates expert anatomical knowledge to identify the cervix, similar to the approach taken by clinicians. CL-Net captures anatomical features related to CL measurement, facilitating the identification of the cervical canal. It then identifies the cervical canal and automatically provides reproducible and reliable CL measurements. CL-Net achieved a success rate of 95.5% in recognizing the cervical canal, comparable to that of human experts (96.4%). Furthermore, the differences between the CL measurements of CL-Net and ground truth were considerably smaller than those made by non-experts and were comparable to those made by experts (median 1.36 mm, IQR 0.87-2.82 mm, range 0.06-6.95 mm for straight cervix; median 1.31 mm, IQR 0.61-2.65 mm, range 0.01-8.18 mm for curved one).

The size of an individual's tongue relative to the oral cavity is associated with articulation speed [Feng, Lu, Zheng, Chi, and Honda, in Proceedings of the 10th Biennial Asia Pacific Conference on Speech, Language, and Hearing (2017), pp. 17-19)] and may affect speech clarity. This study introduces an ultrasound-based method for measuring relative tongue size, termed ultrasound-based relative tongue size (uRTS), as a cost-effective alternative to the magnetic resonance imaging (MRI) based method. Using deep learning to extract the tongue contour, uRTS was calculated from tongue and oropharyngeal cavity sizes in the midsagittal plane. Results from ten speakers showed a strong correlation between uRTS and MRI-based measurements (r = 0.87) and a negative correlation with tongue movement speed (r = -0.73), indicating uRTS is a useful index for assessing tongue size.

Cardiac magnetic resonance (CMR) is a key investigation in clinical cardiovascular medicine and has been used extensively in population research. However, extracting clinically important measurements such as ejection fraction for diagnosing cardiovascular diseases remains time-consuming and subjective. We developed CineMA, a foundation AI model automating these tasks with limited labels. CineMA is a self-supervised autoencoder model trained on 74,916 cine CMR studies to reconstruct images from masked inputs. After fine-tuning, it was evaluated across eight datasets on 23 tasks from four categories: ventricle and myocardium segmentation, left and right ventricle ejection fraction calculation, disease detection and classification, and landmark localisation. CineMA is the first foundation model for cine CMR to match or outperform convolutional neural networks (CNNs). CineMA demonstrated greater label efficiency than CNNs, achieving comparable or better performance with fewer annotations. This reduces the burden of clinician labelling and supports replacing task-specific training with fine-tuning foundation models in future cardiac imaging applications. Models and code for pre-training and fine-tuning are available at https://github.com/mathpluscode/CineMA, democratising access to high-performance models that otherwise require substantial computational resources, promoting reproducibility and accelerating clinical translation.

Subclinical atrial fibrillation (SCAF), also known as atrial high-rate episodes (AHREs), refers to asymptomatic heart rate elevations associated with increased risks of atrial fibrillation and cardiovascular events. Although deep learning (DL) models leveraging echocardiographic images from ultrasound are widely used for cardiac function analysis, their application to AHRE prediction remains unexplored. This study introduces a novel DL-based framework for automatic AHRE detection using echocardiograms. The approach encompasses left atrium (LA) segmentation, LA strain feature extraction, and AHRE classification. Data from 117 patients with cardiac implantable electronic devices undergoing echocardiography were analyzed, with 80% allocated to the development set and 20% to the test set. LA segmentation accuracy was quantified using the Dice coefficient, yielding scores of 0.923 for the LA cavity and 0.741 for the LA wall. For AHRE classification, metrics such as area under the curve (AUC), accuracy, sensitivity, and specificity were employed. A transformer-based model integrating patient characteristics demonstrated robust performance, achieving mean AUC of 0.815, accuracy of 0.809, sensitivity of 0.800, and specificity of 0.783 for a 24-h AHRE duration threshold. This framework represents a reliable tool for AHRE assessment and holds significant potential for early SCAF detection, enhancing clinical decision-making and patient outcomes.

Clinical decision-making routinely demands reasoning over heterogeneous data, yet existing multimodal language models (MLLMs) remain largely vision-centric and fail to generalize across clinical specialties. To bridge this gap, we introduce QoQ-Med-7B/32B, the first open generalist clinical foundation model that jointly reasons across medical images, time-series signals, and text reports. QoQ-Med is trained with Domain-aware Relative Policy Optimization (DRPO), a novel reinforcement-learning objective that hierarchically scales normalized rewards according to domain rarity and modality difficulty, mitigating performance imbalance caused by skewed clinical data distributions. Trained on 2.61 million instruction tuning pairs spanning 9 clinical domains, we show that DRPO training boosts diagnostic performance by 43% in macro-F1 on average across all visual domains as compared to other critic-free training methods like GRPO. Furthermore, with QoQ-Med trained on intensive segmentation data, it is able to highlight salient regions related to the diagnosis, with an IoU 10x higher than open models while reaching the performance of OpenAI o4-mini. To foster reproducibility and downstream research, we release (i) the full model weights, (ii) the modular training pipeline, and (iii) all intermediate reasoning traces at https://github.com/DDVD233/QoQ_Med.

Accurate computer-aided diagnosis systems rely on precise segmentation of the vertebral column to assist physicians in diagnosing various disorders. However, segmenting spinal disks and bones becomes challenging in the presence of abnormalities and complex anatomical structures. While Deep Convolutional Neural Networks (DCNNs) achieve remarkable results in medical image segmentation, their performance is limited by data insufficiency and the high computational complexity of existing solutions. This paper introduces LiDSCUNet++, a lightweight deep learning framework based on depthwise-separable and pointwise convolutions integrated with UNet++ for vertebral column segmentation. The model segments vertebral anomalies from dog radiographs, and the results are further processed by YOLOv8 for automated detection of Spondylosis Deformans. LiDSCUNet++ delivers comparable segmentation performance while significantly reducing trainable parameters, memory usage, energy consumption, and computational time, making it an efficient and practical solution for medical image analysis.

This paper presents MSLesSeg, a new, publicly accessible MRI dataset designed to advance research in Multiple Sclerosis (MS) lesion segmentation. The dataset comprises 115 scans of 75 patients including T1, T2 and FLAIR sequences, along with supplementary clinical data collected across different sources. Expert-validated annotations provide high-quality lesion segmentation labels, establishing a reliable human-labeled dataset for benchmarking. Part of the dataset was shared with expert scientists with the aim to compare the last automatic AI-based image segmentation solutions with an expert-biased handmade segmentation. In addition, an AI-based lesion segmentation of MSLesSeg was developed and technically validated against the last state-of-the-art methods. The dataset, the detailed analysis of researcher contributions, and the baseline results presented here mark a significant milestone for advancing automated MS lesion segmentation research.

Investigate spleen diameter (d) and volume (v) relationship in patients with hematologic malignancies (HM) by determining volumetric thresholds that best correlate to established diameter thresholds for assessing response to treatment. Exploratorily, interrogate the impact of volumetric measurements in response categories and as a predictor of response. Secondary analysis of prospectively collected clinical trial data of 382 patients with HM. Spleen diameters were computed following Lugano criteria and volumes using deep learning segmentation. d and v relationship was estimated using power regression model, volumetric thresholds ([Formula: see text]) for treatment response estimated; threshold search to determine percentual change ([Formula: see text] and minimum volumetric increase ([Formula: see text]) that maximize agreement with Lugano criteria performed. Spleen diameter and volume predictive performance for clinical response investigated using random forest model. [Formula: see text] describes the relationship between spleen diameter and volume. [Formula: see text] for splenomegaly was 546 cm³. [Formula: see text], [Formula: see text], and [Formula: see text] for assessing response resulting in highest agreement with Lugano criteria were 570 cm³, 73%, and 170 cm³, respectively. Predictive performance for response between diameter and volume were not significantly different (P=0.78). This study provides empirical spleen volume threshold and percentual changes that best correlate with diameter thresholds, i.e., Lugano criteria, for assessment of response to treatment in patients with HM. In our dataset use of spleen volumetric thresholds versus diameter thresholds resulted in similar response assessment categories and did not signal differences in predictive values for response.

This study aims to develop a deep learning framework that leverages modality imputation and subregion segmentation to improve grading accuracy in high-grade gliomas. A retrospective analysis was conducted using data from 1,251 patients in the BraTS2021 dataset as the main cohort and 181 clinical cases collected from a medical center between April 2013 and June 2018 (51 years ± 17; 104 males) as the external test set. We propose a PatchGAN-based modality imputation network with an Aggregated Residual Transformer (ART) module combining Transformer self-attention and CNN feature extraction via residual links, paired with a U-Net variant for segmentation. Generative accuracy used PSNR and SSIM for modality conversions, while segmentation performance was measured with DSC and HD95 across necrotic core (NCR), edema (ED), and enhancing tumor (ET) regions. Senior radiologists conducted a comprehensive Likert-based assessment, with diagnostic accuracy evaluated by AUC. Statistical analysis was performed using the Wilcoxon signed-rank test and the DeLong test. The best source-target modality pairs for imputation were T1 to T1ce and T1ce to T2 (p < 0.001). In subregion segmentation, the overall DSC was 0.878 and HD95 was 19.491, with the ET region showing the highest segmentation accuracy (DSC: 0.877, HD95: 12.149). Clinical validation revealed an improvement in grading accuracy by the senior radiologist, with the AUC increasing from 0.718 to 0.913 (P < 0.001) when using the combined imputation and segmentation models. The proposed deep learning framework improves high-grade glioma grading by modality imputation and segmentation, aiding the senior radiologist and offering potential to advance clinical decision-making.