This meta-analysis aims to assess the diagnostic performance of artificial intelligence (AI) based on magnetic resonance imaging (MRI) in detecting molecular subtypes of pediatric medulloblastoma (MB) in children. A thorough review of the literature was performed using PubMed, Embase, and Web of Science to locate pertinent studies released prior to October 2024. Selected studies focused on the diagnostic performance of AI based on MRI in detecting molecular subtypes of pediatric MB. A bivariate random-effects model was used to calculate pooled sensitivity and specificity, both with 95% confidence intervals (CI). Study heterogeneity was assessed using I² statistics. Among the 540 studies determined, eight studies (involving 1195 patients) were included. For the wingless (WNT), the combined sensitivity, specificity, and receiver operating characteristic curve (AUC) based on MRI were 0.73 (95% CI: 0.61-0.83, I² = 19%), 0.94 (95% CI: 0.79-0.99, I² = 93%), and 0.80 (95% CI: 0.77-0.83), respectively. For the sonic hedgehog (SHH), the combined sensitivity, specificity, and AUC were 0.64 (95% CI: 0.51-0.75, I² = 69%), 0.84 (95% CI: 0.80-0.88, I² = 54%), and 0.85 (95% CI: 0.81-0.88), respectively. For Group 3 (G3), the combined sensitivity, specificity, and AUC were 0.89 (95% CI: 0.52-0.98, I² = 82%), 0.70 (95% CI: 0.62-0.77, I² = 44%), and 0.88 (95% CI: 0.84-0.90), respectively. For Group 4 (G4), the combined sensitivity, specificity, and AUC were 0.77 (95% CI: 0.64-0.87, I² = 54%), 0.91 (95% CI: 0.68-0.98, I² = 80%), and 0.86 (95% CI: 0.83-0.89), respectively. MRI-based artificial intelligence shows high diagnostic performance in detecting molecular subtypes of pediatric MB. However, all included studies employed retrospective designs, which may introduce potential biases. More researches using external validation datasets are needed to confirm the results and assess their clinical applicability.

BackgroundLung cancer (LC) is the leading cause of cancer mortality, often diagnosed at advanced stages. Screening reduces mortality in high-risk individuals, but its efficiency can improve with pre- and post-screening risk stratification. With recent LC screening guideline updates in Europe and the US, numerous novel risk prediction models have emerged since the last systematic review of such models. We reviewed risk-based models for selecting candidates for CT screening, and post-CT stratification. MethodsWe systematically reviewed Embase and MEDLINE (2020-2024), identifying studies proposing new LC risk models for screening selection or nodule classification. Data extraction included study design, population, model type, risk horizon, and internal/external validation metrics. In addition, we performed an exploratory meta-regression of AUCs to assess whether sample size, model class, validation type, and biomarker use were associated with discrimination. ResultsOf 1987 records, 68 were included: 41 models were for screening selection (20 without biomarkers, 21 with), and 27 for nodule classification. Regression-based models predominated, though machine learning and deep learning approaches were increasingly common. Discrimination ranged from moderate (AUC{approx}0.70) to excellent (>0.90), with biomarker and imaging-enhanced models often outperforming traditional ones. Model calibration was inconsistently reported, and fewer than half underwent external validation. Meta-regression suggested that, among pre-screening models, larger sample sizes were modestly associated with higher AUC. Conclusion75 models had been identified prior to 2020, we found 68 models since. This reflects growing interest in personalized LC screening. While many demonstrate strong discrimination, inconsistent calibration and limited external validation hinder clinical adoption. Future efforts should prioritize improving existing models rather than developing new ones, transparent evaluation, cost-effectiveness analysis, and real-world implementation.

This study aims to explore a method based on three-dimensional cervical spinal cord reconstruction, radiomics feature extraction, and machine learning to build a postoperative prognosis prediction model for patients with cervical spondylotic myelopathy (CSM). It also evaluates the predictive performance of different cervical spinal cord segmentation strategies and machine learning algorithms. A retrospective analysis is conducted on 126 CSM patients who underwent posterior single-door laminoplasty from January 2017 to December 2022. Three different cervical spinal cord segmentation strategies (narrowest segment, surgical segment, and entire cervical cord C1-C7) are applied to preoperative MRI images for radiomics feature extraction. Good clinical prognosis is defined as a postoperative JOA recovery rate ≥50%. By comparing the performance of 8 machine learning algorithms, the optimal cervical spinal cord segmentation strategy and classifier are selected. Subsequently, clinical features (smoking history, diabetes, preoperative JOA score, and cSVA) are combined with radiomics features to construct a clinical-radiomics prediction model. Among the three cervical spinal cord segmentation strategies, the SVM model based on the narrowest segment performed best (AUC=0.885). Among clinical features, smoking history, diabetes, preoperative JOA score, and cSVA are important indicators for prognosis prediction. When clinical features are combined with radiomics features, the fusion model achieved excellent performance on the test set (accuracy=0.895, AUC=0.967), significantly outperforming either the clinical model or the radiomics model alone. This study validates the feasibility and superiority of three-dimensional radiomics combined with machine learning in predicting postoperative prognosis for CSM. The combination of radiomics features based on the narrowest segment and clinical features can yield a highly accurate prognosis prediction model, providing new insights for clinical assessment and individualized treatment decisions. Future studies need to further validate the stability and generalizability of this model in multi-center, large-sample cohorts.

Mammography screening is an essential tool for early detection of breast cancer. The speed and accuracy of mammography interpretation have the potential to be improved with deep learning methods. However, the development of a foundation visual language model (VLM) is hindered by limited data and domain differences between natural and medical images. Existing mammography VLMs, adapted from natural images, often ignore domain-specific characteristics, such as multi-view relationships in mammography. Unlike radiologists who analyze both views together to process ipsilateral correspondence, current methods treat them as independent images or do not properly model the multi-view correspondence learning, losing critical geometric context and resulting in suboptimal prediction. We propose GLAM: Global and Local Alignment for Multi-view mammography for VLM pretraining using geometry guidance. By leveraging the prior knowledge about the multi-view imaging process of mammograms, our model learns local cross-view alignments and fine-grained local features through joint global and local, visual-visual, and visual-language contrastive learning. Pretrained on EMBED [14], one of the largest open mammography datasets, our model outperforms baselines across multiple datasets under different settings.

Since its introduction in 2015, the U-Net architecture has become popular for medical image segmentation. U-Net is known for its "skip connections," which transfer image details directly to its decoder branch at various levels. However, it's unclear how these skip connections affect the model's performance when the texture of input images varies. To explore this, we tested six types of U-Net-like architectures in three groups: Standard (U-Net and V-Net), No-Skip (U-Net and V-Net without skip connections), and Enhanced (AGU-Net and UNet++, which have extra skip connections). Because convolutional neural networks (CNNs) are known to be sensitive to texture, we defined a novel texture disparity (TD) metric and ran experiments with synthetic images, adjusting this measure. We then applied these findings to four real medical imaging datasets, covering different anatomies (breast, colon, heart, and spleen) and imaging types (ultrasound, histology, MRI, and CT). The goal was to understand how the choice of architecture impacts the model's ability to handle varying TD between foreground and background. For each dataset, we tested the models with five categories of TD, measuring their performance using the Dice Score Coefficient (DSC), Hausdorff distance, surface distance, and surface DSC. Our results on synthetic data with varying textures show differences between the performance of architectures with and without skip connections, especially when trained in hard textural conditions. When translated to medical data, it indicates that training data sets with a narrow texture range negatively impact the robustness of architectures that include more skip connections. The robustness gap between architectures reduces when trained on a larger TD range. In the harder TD categories, models from the No-Skip group performed the best in 5/8 cases (based on DSC) and 7/8 (based on Hausdorff distances). When measuring robustness using the coefficient of variation metric on the DSC, the No-Skip group performed the best in 7 out of 16 cases, showing superior results than the Enhanced (6/16) and Standard groups (3/16). These findings suggest that skip connections offer performance benefits, usually at the expense of robustness losses, depending on the degree of texture disparity between the foreground and background, and the range of texture variations present in the training set. This indicates careful evaluation of their use for robustness-critical tasks like medical image segmentation. Combinations of texture-aware architectures must be investigated to achieve better performance-robustness characteristics.

Precise temperature control is challenging during transcranial MR-guided focused ultrasound (MRgFUS) treatment. The aim of this study was to develop a deep learning model integrating the treatment parameters for each sonication, along with patient-specific clinical information and skull metrics, for prediction of the MRgFUS therapeutic temperature. This is a retrospective analysis of sonications from patients with essential tremor or Parkinson's disease who underwent unilateral MRgFUS thalamotomy or pallidothalamic tractotomy at a single hospital from January 2019 to June 2023. For model training, a dataset of 600 sonications (72 patients) was used, while a validation dataset comprising 199 sonications (18 patients) was used to assess model performance. Additionally, an external dataset of 146 sonications (20 patients) was used for external validation. The developed deep learning model, called Fust-Net, achieved high predictive accuracy, with normalized mean absolute errors of 1.655°C for the internal dataset and 2.432°C for the external dataset, which closely matched the actual temperature. The graded evaluation showed that Fust-Net achieved an effective temperature prediction rate of 82.6%. These results showcase the exciting potential of Fust-Net for achieving precise temperature control during MRgFUS treatment, opening new doors for enhanced precision and safety in clinical applications.

Machine learning (ML) has been increasingly applied to cervical cancer (CC) research. However, few studies have combined both clinical parameters and imaging data. At the same time, there remains an urgent need for more robust and accurate preoperative assessment of parametrial invasion and lymph node metastasis, as well as postoperative prognosis prediction. The objective of this study is to develop an integrated ML model combining clinicopathological variables and magnetic resonance image features for (1) preoperative parametrial invasion and lymph node metastasis detection and (2) postoperative recurrence and survival prediction. Retrospective data from 250 patients with CC (2014-2022; 2 tertiary hospitals) were analyzed. Variables were assessed for their predictive value regarding parametrial invasion, lymph node metastasis, survival, and recurrence using 7 ML models: K-nearest neighbor (KNN), support vector machine, decision tree, random forest (RF), balanced RF, weighted DT, and weighted KNN. Performance was assessed via 5-fold cross-validation using accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve (AUC). The optimal models were deployed in an artificial intelligence-assisted contouring and prognosis prediction system. Among 250 women, there were 11 deaths and 24 recurrences. (1) For preoperative evaluation, the integrated model using balanced RF achieved optimal performance (sensitivity 0.81, specificity 0.85) for parametrial invasion, while weighted KNN achieved the best performance for lymph node metastasis (sensitivity 0.98, AUC 0.72). (2) For postoperative prognosis, weighted KNN also demonstrated high accuracy for recurrence (accuracy 0.94, AUC 0.86) and mortality (accuracy 0.97, AUC 0.77), with relatively balanced sensitivity of 0.80 and 0.33, respectively. (3) An artificial intelligence-assisted contouring and prognosis prediction system was developed to support preoperative evaluation and postoperative prognosis prediction. The integration of clinical data and magnetic resonance images provides enhanced diagnostic capability to preoperatively detect parametrial invasion and lymph node metastasis detection and prognostic capability to predict recurrence and mortality for CC, facilitating personalized, precise treatment strategies.

Aortic stenosis (AS), a prevalent and serious heart valve disorder, requires early detection but remains difficult to diagnose in routine practice. Although echocardiography with Doppler imaging is the clinical standard, these assessments are typically limited to trained specialists. Point-of-care ultrasound (POCUS) offers an accessible alternative for AS screening but is restricted to basic 2D B-mode imaging, often lacking the analysis Doppler provides. Our project introduces MultiASNet, a multimodal machine learning framework designed to enhance AS screening with POCUS by combining 2D B-mode videos with structured data from echocardiography reports, including Doppler parameters. Using contrastive learning, MultiASNet aligns video features with report features in tabular form from the same patient to improve interpretive quality. To address misalignment where a single report corresponds to multiple video views, some irrelevant to AS diagnosis, we use cross-attention in a transformer-based video and tabular network to assign less importance to irrelevant report data. The model integrates structured data only during training, enabling independent use with B-mode videos during inference for broader accessibility. MultiASNet also incorporates sample selection to counteract label noise from observer variability, yielding improved accuracy on two datasets. We achieved balanced accuracy scores of 93.0% on a private dataset and 83.9% on the public TMED-2 dataset for AS detection. For severity classification, balanced accuracy scores were 80.4% and 59.4% on the private and public datasets, respectively. This model facilitates reliable AS screening in non-specialist settings, bridging the gap left by Doppler data while reducing noise-related errors. Our code is publicly available at github.com/DeepRCL/MultiASNet.

Large Language Models (LLMs) have shown strong performance in text-based healthcare tasks. However, their utility in image-based applications remains unexplored. We investigate the effectiveness of LLMs for medical imaging tasks, specifically glioma classification and segmentation, and compare their performance to that of traditional convolutional neural networks (CNNs). Using the BraTS 2020 dataset of multi-modal brain MRIs, we evaluated a general-purpose vision-language LLM (LLaMA 3.2 Instruct) both before and after fine-tuning, and benchmarked its performance against custom 3D CNNs. For glioma classification (Low-Grade vs. High-Grade), the CNN achieved 80% accuracy and balanced precision and recall. The general LLM reached 76% accuracy but suffered from a specificity of only 18%, often misclassifying Low-Grade tumors. Fine-tuning improved specificity to 55%, but overall performance declined (e.g., accuracy dropped to 72%). For segmentation, three methods - center point, bounding box, and polygon extraction, were implemented. CNNs accurately localized gliomas, though small tumors were sometimes missed. In contrast, LLMs consistently clustered predictions near the image center, with no distinction of glioma size, location, or placement. Fine-tuning improved output formatting but failed to meaningfully enhance spatial accuracy. The bounding polygon method yielded random, unstructured outputs. Overall, CNNs outperformed LLMs in both tasks. LLMs showed limited spatial understanding and minimal improvement from fine-tuning, indicating that, in their current form, they are not well-suited for image-based tasks. More rigorous fine-tuning or alternative training strategies may be needed for LLMs to achieve better performance, robustness, and utility in the medical space.

3D structural Magnetic Resonance Imaging (MRI) brain scans are commonly acquired in clinical settings to monitor a wide range of neurological conditions, including neurodegenerative disorders and stroke. While deep learning models have shown promising results analyzing 3D MRI across a number of brain imaging tasks, most are highly tailored for specific tasks with limited labeled data, and are not able to generalize across tasks and/or populations. The development of self-supervised learning (SSL) has enabled the creation of large medical foundation models that leverage diverse, unlabeled datasets ranging from healthy to diseased data, showing significant success in 2D medical imaging applications. However, even the very few foundation models for 3D brain MRI that have been developed remain limited in resolution, scope, or accessibility. In this work, we present a general, high-resolution SimCLR-based SSL foundation model for 3D brain structural MRI, pre-trained on 18,759 patients (44,958 scans) from 11 publicly available datasets spanning diverse neurological diseases. We compare our model to Masked Autoencoders (MAE), as well as two supervised baselines, on four diverse downstream prediction tasks in both in-distribution and out-of-distribution settings. Our fine-tuned SimCLR model outperforms all other models across all tasks. Notably, our model still achieves superior performance when fine-tuned using only 20% of labeled training samples for predicting Alzheimer's disease. We use publicly available code and data, and release our trained model at https://github.com/emilykaczmarek/3D-Neuro-SimCLR, contributing a broadly applicable and accessible foundation model for clinical brain MRI analysis.