Cerebrovascular reactivity reflects changes in cerebral blood flow in response to an acute stimulus and is reflective of the brain's ability to match blood flow to demand. Functional MRI with a breath-hold task can be used to elicit this vasoactive response, but data validity hinges on subject compliance. Determining breath-hold compliance often requires external monitoring equipment. To develop a non-invasive and data-driven quality filter for breath-hold compliance using only measurements of head motion during imaging. Prospective cohort. Longitudinal data from healthy middle-aged subjects enrolled in the Coronary Artery Risk Development in Young Adults Brain MRI Study, N = 1141, 47.1% female. 3.0 Tesla gradient-echo MRI. Manual labelling of respiratory belt monitored data was used to determine breath hold compliance during MRI scan. A model to estimate the probability of non-compliance with the breath hold task was developed using measures of head motion. The model's ability to identify scans in which the participant was not performing the breath hold were summarized using performance metrics including sensitivity, specificity, recall, and F1 score. The model was applied to additional unmarked data to assess effects on population measures of CVR. Sensitivity analysis revealed exclusion of non-compliant scans using the developed model did not affect median cerebrovascular reactivity (Median [q1, q3] = 1.32 [0.96, 1.71]) compared to using manual review of respiratory belt data (1.33 [1.02, 1.74]) while reducing interquartile range. The final model based on a multi-layer perceptron machine learning classifier estimated non-compliance with an accuracy of 76.9% and an F1 score of 69.5%, indicating a moderate balance between precision and recall for the identification of scans in which the participant was not compliant. The developed model provides the probability of non-compliance with a breath-hold task, which could later be used as a quality filter or included in statistical analyses. TECHNICAL EFFICACY: Stage 3.

Vision foundation models like DINOv2 demonstrate remarkable potential in medical imaging despite their origin in natural image domains. However, their design inherently works best for uni-modal image analysis, limiting their effectiveness for multi-modal imaging tasks that are common in many medical fields, such as neurology and oncology. While supervised models perform well in this setting, they fail to leverage unlabeled datasets and struggle with missing modalities, a frequent challenge in clinical settings. To bridge these gaps, we introduce MM-DINOv2, a novel and efficient framework that adapts the pre-trained vision foundation model DINOv2 for multi-modal medical imaging. Our approach incorporates multi-modal patch embeddings, enabling vision foundation models to effectively process multi-modal imaging data. To address missing modalities, we employ full-modality masking, which encourages the model to learn robust cross-modality relationships. Furthermore, we leverage semi-supervised learning to harness large unlabeled datasets, enhancing both the accuracy and reliability of medical predictions. Applied to glioma subtype classification from multi-sequence brain MRI, our method achieves a Matthews Correlation Coefficient (MCC) of 0.6 on an external test set, surpassing state-of-the-art supervised approaches by +11.1%. Our work establishes a scalable and robust solution for multi-modal medical imaging tasks, leveraging powerful vision foundation models pre-trained on natural images while addressing real-world clinical challenges such as missing data and limited annotations.

Maternal-fetal Ultrasound is the primary modality for monitoring fetal development, yet automated segmentation remains challenging due to the scarcity of high-quality annotations. To address this limitation, we propose a semi-supervised learning framework that leverages information divergence for robust fetal ultrasound segmentation. Our method employs a lightweight convolutional network (1.47M parameters) and a Transformer-based network, trained jointly with labelled data through standard supervision and with unlabelled data via cross-supervision. To encourage consistent and confident predictions, we introduce an information divergence loss that combines per-pixel Kullback-Leibler divergence and Mutual Information Gap, effectively reducing prediction disagreement between the two models. In addition, we apply mixup on unlabelled samples to further enhance robustness. Experiments on two fetal ultrasound datasets demonstrate that our approach consistently outperforms seven state-of-the-art semi-supervised methods. When only 5% of training data is labelled, our framework improves the Dice score by 2.39%, reduces the 95% Hausdorff distance by 14.90, and decreases the Average Surface Distance by 4.18. These results highlight the effectiveness of leveraging information divergence for annotation-efficient and robust medical image segmentation. Our code is publicly available on GitHub.

Intraoperative 2D/3D registration aligns preoperative 3D volumes with real-time 2D radiographs, enabling accurate localization of instruments and implants. A recent fully differentiable similarity learning framework approximates geodesic distances on SE(3), expanding the capture range of registration and mitigating the effects of substantial disturbances, but existing Euclidean approximations distort manifold structure and slow convergence. To address these limitations, we explore similarity learning in non-Euclidean spherical feature spaces to better capture and fit complex manifold structure. We extract feature embeddings using a CNN-Transformer encoder, project them into spherical space, and approximate their geodesic distances with Riemannian distances in the bi-invariant SO(4) space. This enables a more expressive and geometrically consistent deep similarity metric, enhancing the ability to distinguish subtle pose differences. During inference, we replace gradient descent with fully differentiable Levenberg-Marquardt optimization to accelerate convergence. Experiments on real and synthetic datasets show superior accuracy in both patient-specific and patient-agnostic scenarios.

We aimed to develop and validate a radiomics-based machine learning nomogram using multiparametric magnetic resonance imaging to preoperatively predict substantial lymphovascular space invasion in patients with endometrial cancer. This retrospective dual-center study included patients with histologically confirmed endometrial cancer who underwent preoperative magnetic resonance imaging (MRI). The patients were divided into training and test sets. Radiomic features were extracted from multiparametric magnetic resonance imaging to generate radiomic scores using a support vector machine algorithm. Three predictive models were constructed: clinical (Model^C), radiomics-only (Model^R), and fusion (Model^N). The models' performances were evaluated by analyzing their receiver operating characteristic curves, and pairwise comparisons of the models' areas under the curves were conducted using DeLong's test and adjusted using the Bonferroni correction. Decision curve analysis with integrated discrimination improvement was used for net benefit comparison. This study enrolled 283 women (training set: n = 198; test set: n = 85). The lymphovascular space invasion groups (substantial and no/focal) had significantly different radiomic scores (P < 0.05). Model^N achieved an area under the curve of 0.818 (95% confidence interval: 0.757-0.869) and 0.746 (95% confidence interval: 0.640-0.835) for the training and test sets, respectively, demonstrating a good fit according to the Hosmer-Lemeshow test (P > 0.05). The DeLong test with Bonferroni correction indicated that Model^N demonstrated better diagnostic efficiency than Model^C in predicting substantial lymphovascular space invasion in the two sets (adjusted P < 0.05). In addition, decision curve analysis demonstrated a higher net benefit for Model^N, with integrated discrimination improvements of 0.043 and 0.732 (P < 0.01) in the training and test sets, respectively. The multiparametric magnetic resonance imaging-based radiomics machine learning nomogram showed moderate diagnostic performance for substantial lymphovascular space invasion in patients with endometrial cancer.

The purpose of this study was to assess the accuracy of a customized deep learning model based on CNN and U-Net for detecting and segmenting the second mesiobuccal canal (MB2) of maxillary first molar teeth on cone beam computed tomography (CBCT) scans. CBCT scans of 37 patients were imported into 3D slicer software to crop and segment the canals of the mesiobuccal (MB) root of the maxillary first molar. The annotated data were divided into two groups: 80% for training and validation and 20% for testing. The data were used to train the AI model in 2 separate steps: a classification model based on a customized CNN and a segmentation model based on U-Net. A confusion matrix and receiver-operating characteristic (ROC) analysis were used in the statistical evaluation of the results of the classification model, whereas the Dice-coefficient (DCE) was used to express the segmentation accuracy. F1 score, testing accuracy, recall and precision values were 0.93, 0.87, 1.0 and 0.87 respectively, for the cropped images of MB root of maxillary 1st molar teeth in the testing group. The testing loss was 0.4, and the area under the curve (AUC) value was 0.57. The segmentation accuracy results were satisfactory, where the DCE of training was 0.85 and DCE of testing was 0.79. MB2 in the maxillary first molar can be precisely detected and segmented via the developed AI algorithm in CBCT images. Current Controlled Trial Number NCT05340140. April 22, 2022.

To establish an explainable machine learning (ML) approach using patient-related and noncontrast chest CT-derived features to predict the contrast material arrival time (TARR) in CT pulmonary angiography (CTPA). This retrospective study included consecutive patients referred for CTPA between September 2023 to October 2024. Sixteen clinical and 17 chest CT-derived parameters were used as inputs for the ML approach, which employed recursive feature elimination for feature selection and XGBoost with SHapley Additive exPlanations (SHAP) for explainable modeling. The prediction target was abnormal TARR of the pulmonary artery (ie, TARR <7 seconds or >10 s), determined by the time to peak enhancement in the test bolus, with 2 models distinguishing these cases. External validation was conducted. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). A total of 666 patients (mean age, 70 [IQR, 59.3 to 78.0]; 46.8% female participants) were split into training (n = 353), testing (n = 151), and external validation (n = 162) sets. 86 cases (12.9%) had TARR <7 seconds, and 138 cases (20.7%) had TARR >10 seconds. The ML models exhibited good performance in their respective testing and external validation sets (AUC: 0.911 and 0.878 for TARR <7 s; 0.834 and 0.897 for TARR >10 s). SHAP analysis identified the measurements of the vena cava and pulmonary artery as key features for distinguishing abnormal TARR. The explainable ML algorithm accurately identified normal and abnormal TARR of the pulmonary artery, facilitating personalized CTPA scans.

This retrospective-prospective study evaluated whether a deep learning-based MRI reconstruction algorithm can preserve diagnostic quality in brain MRI scans accelerated up to fourfold, using both public and prospective clinical data. The study included 18 healthy volunteers (scans acquired at 3T, January 2024-March 2025), as well as selected fastMRI public datasets with diverse pathologies. Phase-encoding-undersampled 2D/3D T1, T2, and FLAIR sequences were reconstructed with DeepFoqus-Accelerate and compared with standard-of-care (SOC). Three board-certified neuroradiologists and two MRI technologists independently reviewed 36 paired SOC/AI reconstructions from both datasets using a 5-point Likert scale, while quantitative similarity was assessed for 408 scans and 1224 datasets using Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), and Haar wavelet-based Perceptual Similarity Index (HaarPSI). No AI-reconstructed scan scored below 3 (minimally acceptable), and 95% scored $\geq 4$. Mean SSIM was 0.95 $\pm$ 0.03 (90% cases >0.90), PSNR >41.0 dB, and HaarPSI >0.94. Inter-rater agreement was slight to moderate. Rare artifacts did not affect diagnostic interpretation. These findings demonstrate that DeepFoqus-Accelerate enables robust fourfold brain MRI acceleration with 75% reduced scan time, while preserving diagnostic image quality and supporting improved workflow efficiency.

Point-of-care ultrasonography has become a valuable tool for assessing diaphragmatic function in critically ill patients receiving invasive mechanical ventilation. However, conventional diaphragm ultrasound assessment remains highly operator-dependent and subjective. Previous research introduced automatic measurement of diaphragmatic excursion and velocity using 2D speckle-tracking technology. This study aimed to develop an artificial intelligence-multimodal learning framework to improve the prediction of weaning failure and guide individualized weaning strategies. This prospective study enrolled critically ill patients older than 18 years who received mechanical ventilation for more than 48 hours and were eligible for a spontaneous breathing trial in 2 intensive care units in Guangzhou, China. Before the spontaneous breathing trial, diaphragm ultrasound videos were collected using a standardized protocol, and automatic measurements of excursion and velocity were obtained. A total of 88 patients were included, with 50 successfully weaned and 38 experiencing weaning failure. Each patient record included 27 clinical and 6 diaphragmatic indicators, selected based on previous literature and phenotyping studies. Clinical variables were preprocessed using OneHotEncoder, normalization, and scaling. Ultrasound videos were interpolated to a uniform resolution of 224×224×96. Artificial intelligence-multimodal learning based on clinical characteristics, laboratory parameters, and diaphragm ultrasonic videos was established. Four experiments were conducted in an ablation setting to evaluate model performance using different combinations of input data: (1) diaphragmatic excursion only, (2) clinical and diaphragmatic indicators, (3) ultrasound videos only, and (4) all modalities combined (multimodal). Metrics for evaluation included classification accuracy, area under the receiver operating characteristic curve (AUC), average precision in the precision-recall curve, and calibration curve. Variable importance was assessed using SHAP (Shapley Additive Explanation) to interpret feature contributions and understand model predictions. The multimodal co-learning model outperformed all single-modal approaches. The accuracy improved when predicted through diaphragm ultrasound video data using Video Vision Transformer (accuracy=0.8095, AUC=0.852), clinical or ultrasound indicators (accuracy=0.7381, AUC=0.746), and the multimodal co-learning (accuracy=0.8331, AUC=0.894). The proposed co-learning model achieved the highest score (average precision=0.91) among the 4 experiments. Furthermore, calibration curve analysis demonstrated that the proposed colearning model was well calibrated, as the curve was closest to the perfectly calibrated line. Combining ultrasound and clinical data for colearning improved the accuracy of the weaning outcome prediction. Multimodal learning based on automatic measurement of point-of-care ultrasonography and automated collection of objective clinical indicators greatly enhanced the practical operability and user-friendliness of the system. The proposed model offered promising potential for widespread clinical application in intensive care settings.

To evaluate the performance of artificial intelligence (AI)-based models in predicting elevated neonatal insulin levels through fetal hepatic echotexture analysis. This diagnostic accuracy study analyzed ultrasound images of fetal livers from pregnancies between 37 and 42 weeks, including cases with and without gestational diabetes mellitus (GDM). Images were stored in Digital Imaging and Communications in Medicine (DICOM) format, annotated by experts, and converted to segmented masks after quality checks. A balanced dataset was created by randomly excluding overrepresented categories. Artificial intelligence classification models developed using the FastAI library-ResNet-18, ResNet-34, ResNet-50, EfficientNet-B0, and EfficientNet-B7-were trained to detect elevated C-peptide levels (>75th percentile) in umbilical cord blood at birth, based on fetal hepatic ultrasonographic images. Out of 2339 ultrasound images, 606 were excluded due to poor quality, resulting in 1733 images analyzed. Elevated C-peptide levels were observed in 34.3% of neonates. Among the 5 CNN models evaluated, EfficientNet-B0 demonstrated the highest overall performance, achieving a sensitivity of 86.5%, specificity of 82.1%, positive predictive value (PPV) of 83.0%, negative predictive value (NPV) of 85.7%, accuracy of 84.3%, and an area under the ROC curve (AUC) of 0.83 in predicting elevated neonatal insulin levels through fetal hepatic echotexture analysis. AI-based analysis of fetal liver echotexture via ultrasound effectively predicted elevated neonatal C-peptide levels, offering a promising non-invasive method for detecting insulin imbalance in newborns.