Inter reader variability and cross site domain shift challenge the automatic segmentation of prostate anatomy using T2 weighted MRI images. This study investigates whether transformer models can retain precision amid such heterogeneity. We compare the performance of UNETR and SwinUNETR in prostate gland segmentation against our previous 3D UNet model [1], based on 546 MRI (T2weighted) volumes annotated by two independent experts. Three training strategies were analyzed: single cohort dataset, 5 fold cross validated mixed cohort, and gland size based dataset. Hyperparameters were tuned by Optuna. The test set, from an independent population of readers, served as the evaluation endpoint (Dice Similarity Coefficient). In single reader training, SwinUNETR achieved an average dice score of 0.816 for Reader#1 and 0.860 for Reader#2, while UNETR scored 0.8 and 0.833 for Readers #1 and #2, respectively, compared to the baseline UNets 0.825 for Reader #1 and 0.851 for Reader #2. SwinUNETR had an average dice score of 0.8583 for Reader#1 and 0.867 for Reader#2 in cross-validated mixed training. For the gland size-based dataset, SwinUNETR achieved an average dice score of 0.902 for Reader#1 subset and 0.894 for Reader#2, using the five-fold mixed training strategy (Reader#1, n=53; Reader#2, n=87) at larger gland size-based subsets, where UNETR performed poorly. Our findings demonstrate that global and shifted-window self-attention effectively reduces label noise and class imbalance sensitivity, resulting in improvements in the Dice score over CNNs by up to five points while maintaining computational efficiency. This contributes to the high robustness of SwinUNETR for clinical deployment.

Synthesizing high quality CT projection data remains a significant challenge due to the limited availability of annotated data and the complex nature of CT imaging. In this work, we present PRO, a projection domain synthesis foundation model for CT imaging. To the best of our knowledge, this is the first study that performs CT synthesis in the projection domain. Unlike previous approaches that operate in the image domain, PRO learns rich structural representations from raw projection data and leverages anatomical text prompts for controllable synthesis. This projection domain strategy enables more faithful modeling of underlying imaging physics and anatomical structures. Moreover, PRO functions as a foundation model, capable of generalizing across diverse downstream tasks by adjusting its generative behavior via prompt inputs. Experimental results demonstrated that incorporating our synthesized data significantly improves performance across multiple downstream tasks, including low-dose and sparse-view reconstruction. These findings underscore the versatility and scalability of PRO in data generation for various CT applications. These results highlight the potential of projection domain synthesis as a powerful tool for data augmentation and robust CT imaging. Our source code is publicly available at: https://github.com/yqx7150/PRO.

Kernel size selection in Convolutional Neural Networks (CNNs) is a critical but often overlooked design decision that affects receptive field, feature extraction, computational cost, and model accuracy. This paper proposes the Best Kernel Size Estimation Function (BKSEF), a mathematically grounded and empirically validated framework for optimal, layer-wise kernel size determination. BKSEF balances information gain, computational efficiency, and accuracy improvements by integrating principles from information theory, signal processing, and learning theory. Extensive experiments on CIFAR-10, CIFAR-100, ImageNet-lite, ChestX-ray14, and GTSRB datasets demonstrate that BKSEF-guided architectures achieve up to 3.1 percent accuracy improvement and 42.8 percent reduction in FLOPs compared to traditional models using uniform 3x3 kernels. Two real-world case studies further validate the approach: one for medical image classification in a cloud-based setup, and another for traffic sign recognition on edge devices. The former achieved enhanced interpretability and accuracy, while the latter reduced latency and model size significantly, with minimal accuracy trade-off. These results show that kernel size can be an active, optimizable parameter rather than a fixed heuristic. BKSEF provides practical heuristics and theoretical support for researchers and developers seeking efficient and application-aware CNN designs. It is suitable for integration into neural architecture search pipelines and real-time systems, offering a new perspective on CNN optimization.

High tumor recurrence after surgery remains a significant challenge in managing hepatocellular carcinoma (HCC). We aimed to construct a multimodal model to forecast the early recurrence of HCC after surgical resection and explore the associated biological mechanisms. Overall, 519 patients with HCC were included from three medical centers. 433 patients from Nanfang Hospital were used as the training cohort, and 86 patients from the other two hospitals comprised validation cohort. Radiomics and deep learning (DL) models were developed using contrast-enhanced computed tomography images. Radiomics feature visualization and gradient-weighted class activation mapping were applied to improve interpretability. A multimodal model (MM-RDLM) was constructed by integrating radiomics and DL models. Associations between MM-RDLM and recurrence-free survival (RFS) and overall survival were analyzed. Gene set enrichment analysis (GSEA) and multiplex immunohistochemistry (mIHC) were used to investigate the biological mechanisms. Models based on hepatic arterial phase images exhibited the best predictive performance, with radiomics and DL models achieving areas under the curve (AUCs) of 0.770 (95 % confidence interval [CI]: 0.725-0.815) and 0.846 (95 % CI: 0.807-0.886), respectively, in the training cohort. MM-RDLM achieved an AUC of 0.955 (95 % CI: 0.937-0.972) in the training cohort and 0.930 (95 % CI: 0.876-0.984) in the validation cohort. MM-RDLM (high vs. low) was notably linked to RFS in the training (hazard ratio [HR] = 7.80 [5.74 - 10.61], P < 0.001) and validation (HR = 10.46 [4.96 - 22.68], P < 0.001) cohorts. GSEA revealed enrichment of the natural killer cell-mediated cytotoxicity pathway in the MM-RDLM low cohort. mIHC showed significantly higher percentages of CD3-, CD56-, and CD8-positive cells in the MM-RDLM low group. The MM-RDLM model demonstrated strong predictive performance for early postoperative recurrence of HCC. These findings contribute to identifying patients at high risk for early recurrence and provide insights into the potential underlying biological mechanisms.

We aimed to develop a machine-learning(ML) algorithm consisting of physical examination, sonographic findings, and laboratory markers. The data of 70 patients with confirmed ovarian torsion followed and treated in our clinic for ovarian torsion and 73 patients for control group that presented to the emergency department with similar complaints but didn't have ovarian torsion detected on ultrasound as the control group between 2013-2023 were retrospectively analyzed. Sonographic findings, laboratory values, and clinical status of patients were examined and fed into three supervised ML systems to identify and develop viable decision algorithms. Presence of nausea/vomiting and symptom duration was statistically significant(p<0.05) for ovarian torsion. Presence of abdominal pain and palpable mass on physical examination weren't significant(p>0.05). White blood cell count(WBC), neutrophile/lymphocyte ratio(NLR), systemic immune-inflammation index(SII) and systemic inflammation response index(SIRI), high values of C-reactive protein was highly significant in prediction of torsion( p<0.001,p<0.05). Ovarian size ratio, medialization, follicular ring sign, presence of free fluid in pelvis in ultrasound demonstrated statistical significance in the torsion group(p<0.001). We used supervised ML algorithms, including decision trees, random forests, and LightGBM, to classify patients as either control or having torsion. We evaluated the models using 5-fold cross-validation, achieving an average F1-score of 98%, an accuracy of 98%, and a specificity of 100% across each fold with the decision tree model. This study represents the first development of a ML algorithm that integrates clinical, laboratory and ultrasonographic findings for the diagnosis of pediatric ovarian torsion with over 98% accuracy.

The human brain exhibits intricate spatiotemporal dynamics, which can be described and understood through the framework of complex dynamic systems theory. In this study, we leverage functional magnetic resonance imaging (fMRI) data to investigate reaction-diffusion processes in the brain. A reaction-diffusion process refers to the interaction between two or more substances that spread through space and react with each other over time, often resulting in the formation of patterns or waves of activity. Building on this empirical foundation, we apply a reaction-diffusion framework inspired by theoretical physics to simulate the emergence of brain spacetime vortices within the brain. By exploring this framework, we investigate how reaction-diffusion processes can serve as a compelling model to govern the formation and propagation of brain spacetime vortices, which are dynamic, swirling patterns of brain activity that emerge and evolve across both time and space within the brain. Our approach integrates computational modeling with fMRI data to investigate the spatiotemporal properties of these vortices, offering new insights into the fundamental principles of brain organization. This work highlights the potential of reaction-diffusion models as an alternative framework for understanding brain spacetime dynamics.

Background and purposeAccurate estimation of Overall Survival (OS) in Non-Small Cell Lung Cancer (NSCLC) patients provides critical insights for treatment planning. While previous studies showed that radiomics and Deep Learning (DL) features increased prediction accuracy, this study aimed to examine whether a model that combines the radiomics and DL features with the clinical and dosimetric features outperformed other models. Materials and methodsWe collected pre-treatment lung CT scans and clinical data for 225 NSCLC patients from the Maastro Clinic: 180 for training and 45 for testing. Radiomics features were extracted using the Python radiomics feature extractor, and DL features were obtained using a 3D ResNet model. An ensemble model comprising XGB and NN classifiers was developed using: (1) clinical features only; (2) clinical and radiomics features; (3) clinical and DL features; and (4) clinical, radiomics, and DL features. The performance metrics were evaluated for the test and K-fold cross-validation data sets. ResultsThe prediction model utilizing only clinical variables provided an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.64 and a test accuracy of 77.55%. The best performance came from combining clinical, radiomics, and DL features (AUC: 0.84, accuracy: 85.71%). The prediction improvement of this model was statistically significant compared to models trained with clinical features alone or with a combination of clinical and radiomics features. ConclusionIntegrating radiomics and DL features with clinical characteristics improved the prediction of OS after radiotherapy for NSCLC patients. The increased accuracy of our integrated model enables personalized, risk-based treatment planning, guiding clinicians toward more effective interventions, improved patient outcomes and enhanced quality of life.

We evaluated the accuracy of ChatGPT-3.5's responses to common questions regarding acute breast symptoms and explored whether using lay language, as opposed to medical language, affected the accuracy of the responses. Questions were formulated addressing acute breast conditions, informed by the American College of Radiology (ACR) Appropriateness Criteria (AC) and our clinical experience at a tertiary referral breast center. Of these, seven addressed the most common acute breast symptoms, nine addressed pregnancy-associated breast symptoms, and four addressed specific management and imaging recommendations for a palpable breast abnormality. Questions were submitted three times to ChatGPT-3.5 and all responses were assessed by five fellowship-trained breast radiologists. Evaluation criteria included clinical judgment and adherence to the ACR guidelines, with responses scored as: 1) "appropriate," 2) "inappropriate" if any response contained inappropriate information, or 3) "unreliable" if responses were inconsistent. A majority vote determined the appropriateness for each question. ChatGPT-3.5 generated responses were appropriate for 7/7 (100 %) questions regarding common acute breast symptoms when phrased both colloquially and using standard medical terminology. In contrast, ChatGPT-3.5 generated responses were appropriate for 3/9 (33 %) questions about pregnancy-associated breast symptoms and 3/4 (75 %) questions about management and imaging recommendations for a palpable breast abnormality. ChatGPT-3.5 can automate healthcare information related to appropriate management of acute breast symptoms when prompted with both standard medical terminology or lay phrasing of the questions. However, physician oversight remains critical given the presence of inappropriate recommendations for pregnancy associated breast symptoms and management of palpable abnormalities.

This study focuses on artificial intelligence-driven classification of glioma and Ki-67 leveling using T2w-FLAIR MRI, exploring the association of Ki-67 biomarkers with deep learning (DL) features through explainable artificial intelligence (XAI) and SHapley Additive exPlanations (SHAP). This IRB-approved study included 101 patients with glioma brain tumor acquired MR images with the T2W-FLAIR sequence. We extracted DL bottleneck features using ResNet50 from glioma MR images. Principal component analysis (PCA) was deployed for dimensionality reduction. XAI was used to identify potential features. The XGBosst classified the histologic grades of the glioma and the level of Ki-67. We integrated potential DL features with patient demographics (age and sex) and Ki-67 biomarkers, utilizing SHAP to determine the model's essential features and interactions. Glioma grade classification and Ki-67 level predictions achieved overall accuracies of 0.94 and 0.91, respectively. It achieved precision scores of 0.92, 0.94, and 0.96 for glioma grades 2, 3, and 4, and 0.88, 0.94, and 0.97 for Ki-67 levels (low: 5%≤Ki-67<10%, moderate: 10%≤Ki-67≤20, and high: Ki-67>20%). Corresponding F1-scores were 0.95, 0.88, and 0.96 for glioma grades and 0.92, 0.93, and 0.87 for Ki-67 levels. SHAP analysis further highlighted a strong association between bottleneck DL features and Ki-67 biomarkers, demonstrating their potential to differentiate glioma grades and Ki-67 levels while offering valuable insights into glioma aggressiveness. This study demonstrates the precise classification of glioma grades and the prediction of Ki-67 levels to underscore the potential of AI-driven MRI analysis to enhance clinical decision-making in glioma management.

Three-dimensional (3D) ultrasound (US) aims to provide sonographers with the spatial relationships of anatomical structures, playing a crucial role in clinical diagnosis. Recently, deep-learning-based freehand 3D US has made significant advancements. It reconstructs volumes by estimating transformations between images without external tracking. However, image-only reconstruction poses difficulties in reducing cumulative drift and further improving reconstruction accuracy, particularly in scenarios involving complex motion trajectories. In this context, we propose an enhanced motion network (MoNetV2) to enhance the accuracy and generalizability of reconstruction under diverse scanning velocities and tactics. First, we propose a sensor-based temporal and multi-branch structure that fuses image and motion information from a velocity perspective to improve image-only reconstruction accuracy. Second, we devise an online multi-level consistency constraint that exploits the inherent consistency of scans to handle various scanning velocities and tactics. This constraint exploits both scan-level velocity consistency, path-level appearance consistency, and patch-level motion consistency to supervise inter-frame transformation estimation. Third, we distill an online multi-modal self-supervised strategy that leverages the correlation between network estimation and motion information to further reduce cumulative errors. Extensive experiments clearly demonstrate that MoNetV2 surpasses existing methods in both reconstruction quality and generalizability performance across three large datasets.