This study evaluates whether integrating curriculum learning with diffusion-based synthetic augmentation can enhance the detection of difficult pulmonary nodules in chest radiographs, particularly those with low size, brightness, and contrast, which often challenge conventional AI models due to data imbalance and limited annotation. A Faster R-CNN with a Feature Pyramid Network (FPN) backbone was trained on a hybrid dataset comprising expert-labeled NODE21 (1,213 patients; 52.4 percent male; mean age 63.2 +/- 11.5 years), VinDr-CXR, CheXpert, and 11,206 DDPM-generated synthetic images. Difficulty scores based on size, brightness, and contrast guided curriculum learning. Performance was compared to a non-curriculum baseline using mean average precision (mAP), Dice score, and area under the curve (AUC). Statistical tests included bootstrapped confidence intervals, DeLong tests, and paired t-tests. The curriculum model achieved a mean AUC of 0.95 versus 0.89 for the baseline (p < 0.001), with improvements in sensitivity (70 percent vs. 48 percent) and accuracy (82 percent vs. 70 percent). Stratified analysis demonstrated consistent gains across all difficulty bins (Easy to Very Hard). Grad-CAM visualizations confirmed more anatomically focused attention under curriculum learning. These results suggest that curriculum-guided synthetic augmentation enhances model robustness and generalization for pulmonary nodule detection.

Background: We aimed to quantify hepatic vessel volumes across chronic liver disease stages and healthy controls using deep learning-based magnetic resonance imaging (MRI) analysis, and assess correlations with biomarkers for liver (dys)function and fibrosis/portal hypertension. Methods: We assessed retrospectively healthy controls, non-advanced and advanced chronic liver disease (ACLD) patients using a 3D U-Net model for hepatic vessel segmentation on portal venous phase gadoxetic acid-enhanced 3-T MRI. Total (TVVR), hepatic (HVVR), and intrahepatic portal vein-to-volume ratios (PVVR) were compared between groups and correlated with: albumin-bilirubin (ALBI) and model for end-stage liver disease-sodium (MELD-Na) score, and fibrosis/portal hypertension (Fibrosis-4 [FIB-4] score, liver stiffness measurement [LSM], hepatic venous pressure gradient [HVPG], platelet count [PLT], and spleen volume). Results: We included 197 subjects, aged 54.9 $\pm$ 13.8 years (mean $\pm$ standard deviation), 111 males (56.3\%): 35 healthy controls, 44 non-ACLD, and 118 ACLD patients. TVVR and HVVR were highest in controls (3.9; 2.1), intermediate in non-ACLD (2.8; 1.7), and lowest in ACLD patients (2.3; 1.0) ($p \leq 0.001$). PVVR was reduced in both non-ACLD and ACLD patients (both 1.2) compared to controls (1.7) ($p \leq 0.001$), but showed no difference between CLD groups ($p = 0.999$). HVVR significantly correlated indirectly with FIB-4, ALBI, MELD-Na, LSM, and spleen volume ($\rho$ ranging from -0.27 to -0.40), and directly with PLT ($\rho = 0.36$). TVVR and PVVR showed similar but weaker correlations. Conclusions: Deep learning-based hepatic vessel volumetry demonstrated differences between healthy liver and chronic liver disease stages and shows correlations with established markers of disease severity.

Nasal septum deviation (NSD) is one of the contributing factors to impaired nasal function and dentofacial developmental abnormalities. Although cone-beam computed tomography (CBCT) is clinically valuable for NSD diagnosis, manual interpretation remains labor-intensive and expertise-dependent. Our study included 330 CBCT scans diagnosed with either NSD or non-NSD to develop an automated 2-stage artificial intelligence (AI) framework integrating real-time detection and classification for NSD screening. In the first stage, the YOLOv11 (You Only Look Once) object detection algorithm was employed to detect the region of interest containing the nasal septum. In the second stage, 3 convolutional neural network architectures, ResNet, EfficientNet, and MobileNet, were evaluated for classifying CBCT images into NSD and normal categories. Among the YOLOv11 variants, YOLOv11n demonstrated superior performance with a precision of 0.996, a recall of 1.000, an mAP50 of 0.995, and an mAP50-95 of 0.873. For the classification task, Mobile_small emerged as the top-performing model, achieving an area under the curve of 0.817, an area under the precision-recall curve of 0.845, and an accuracy of 0.749. An AI-assisted diagnostic tool was developed based on YOLOv11n and MobileNet models and validated on 50 internal and 50 external CBCT scans. With AI assistance, orthodontists' diagnostic accuracy increased by 20.12% and 21.49%, respectively, whereas average diagnosis time decreased by 23.75 seconds, improving efficiency by 53.92%. The proposed system enables rapid NSD screening with diagnostic-level accuracy, demonstrating the viability of lightweight AI models for clinical CBCT analysis. AI-assisted diagnosis improves orthodontists' accuracy and time efficiency in identifying NSD.

Ultrasound (US) imaging is the primary choice for diagnosing and triaging patients in the battlefield as well as emergency medicine due to ease of portability and low-power requirements. Interpretation and acquisition of ultrasound images can be challenging and requires personnel with specialized training. Incorporating artificial intelligence (AI) can enhance the imaging process while improving diagnostic accuracy. To accomplish this goal, we have developed a full torso tissue-mimicking phantom for simulating US image capture at each site of the extended-focused assessment with sonography for trauma (eFAST) exam and is suitable for developing AI guidance and classification models. The US images taken from the phantom were used to train AI models for detection of specific anatomical features and injury state diagnosis. The tissue-mimicking phantom successfully simulated full thoracic motion as well as modular injuries at each scan site. AI models trained from the tissue phantom were able to achieve IOU's greater than 0.80 and accuracy of 71.5% on blind inferences. In summary, the tissue mimicking phantom is a reliable tool for acquiring eFAST images for training AI models. Furthermore, the tissue phantom could be implemented for training personnel on ultrasound examination techniques as well as developing image acquisition automation techniques.

This review explores the role of artificial intelligence (AI) in visceral adipose tissue (VAT) and ectopic fat imaging. It aims to evaluate how AI may be used to enhance the efficiency and accuracy of cardiovascular disease (CVD) risk assessment. It addresses key questions regarding AI's capabilities in risk prediction, segmentation, and integration with large volume data for CVD risk assessment. Recent studies demonstrate that AI, powered by deep learning models, significantly improve VAT and ectopic fat segmentation. AI can also be used to facilitate early detection of cardiometabolic risks and allows integration of imaging with clinical data for a more personalized approach to medicine. Emerging applications include AI-enabled telehealth and continuous monitoring through wearable technologies. AI is transforming VAT and ectopic fat imaging by enabling more precise, personalized, and scalable assessments of fat distribution and cardiovascular risk. While challenges remain, such as model interpretability, future research will likely focus on refining algorithms and expanding AI's clinical applications, potentially redefining obesity and CVD risk management.

Imaging triage of stroke patients is primarily based on perfusion imaging. Simplified triage based on non-contrast CT are limited (NCCT). To evaluate the predictive capability of a deep learning algorithm, "Triage Stroke" (Brainomix 360) in identifying anterior circulation large vessel occlusions (LVO) on NCCT in patients with suspected acute ischemic stroke (AIS). This multi-institutional study analyzed 612 patients with suspected AIS at 3 US comprehensive stroke centers. A balanced cohort of consecutive patients with and without anterior circulation LVO was analyzed. Ground truth was based on concurrent CTA evaluated by site neuroradiologists. The primary outcome was predictive performance for LVO detection. The secondary outcomes were 1) prospective comparison of NCCT LVO detection against general radiologists and subspecialty neuroradiologists, and 2) the influence of NIHSS on the model. Triage Stroke software detected an LVO on NCCT with a 67% sensitivity and 93% specificity. The positive and negative predictive values were 59% and 95%, respectively, with an area under the curve (AUC) of 0.8. The software's sensitivity for LVO detection was significantly higher than the group average of all radiologists (difference = 20.5%; CI, 8.26-32.78; <i>P</i> = .001) and was also higher when separated into general and neuroradiology subgroups. The AUC for NCCT LVO was significantly higher than the group of all readers (difference = 11%; CI, 4%-17%; <i>P</i> < .001), and the nonexpert readers (difference = 13%, CI, 7%-20%; <i>P</i> < .001). The addition of NIHSS to the model yielded a high specificity (99%) and similar sensitivity (65%), resulting in the optimum positive predictive value of all models tested (91%). Triage Stroke software demonstrated strong predictive capabilities for NCCT detection of anterior circulation LVOs outperforming radiologists. Coupled with NIHSS it may simplify identification of endovascular candidates especially in resource-constrained environments worldwide.

To present a publicly available deep learning-based torso segmentation model that provides comprehensive voxel-wise coverage, including delineations that extend to the boundaries of anatomical compartments. We extracted preliminary segmentations from TotalSegmentator, spine, and body composition models for magnetic resonance tomography (MR) images, then improved them iteratively and retrained an nnUNet model. Using a random retrospective subset of German National Cohort (NAKO), UK Biobank, internal MR and computed tomography (CT) data (Training: 2897 series from 626 subjects, 290 female; mean age 53 ± 16; 3-fold-cross validation (20% hold-out). Internal testing 36 series from 12 subjects, 6 male; mean age 60 ± 11), we segmented 71 structures in torso MR and 72 in CT images: 20 organs, 10 muscles, 19 vessels, 16 bones, ribs in CT, intervertebral discs, spinal cord, spinal canal and body composition (subcutaneous fat, unclassified muscles and visceral fat). For external validation, we used existing automatic organ segmentations, independent ground truth segmentations on gradient echo images, and the Amos data. We used non-parametric bootstrapping for confidence intervals and the Wilcoxon rank-sum test for computing statistical significance. We achieved an average Dice score of 0.90 ± 0.06 on our internal gradient echo test set, which included 71 semantic segmentation labels. Our model ties with the best model on Amos with a Dice of 0,81 ± 0.14, while having a larger field of view and a considerably higher number of structures included. Our work presents a publicly available full-torso segmentation model for MRI and CT images that classifies almost all subject voxels to date. Question No completed MRI segmentation model exists that delineates the true transition boundaries of the anatomical structures of bone and muscles. Findings We provide a simple-to-use model that automatically segments MRI images, that can be utilized as a backbone for computer-aided automatic analysis. Clinical relevance Our segmentation model enables accurate and detailed full-torso segmentation on MRI and CT, improving automated analysis in large-scale epidemiological studies and facilitating more precise body composition and organ assessments for clinical and research applications.

To assess the efficacy of super-resolution deep learning reconstruction (SR-DLR) in enhancing the visualization of pancreatic cystic lesions (PCLs) on magnetic resonance cholangiopancreatography (MRCP). This retrospective study included 85 patients who underwent MRCP, comprising 52 patients with PCLs and 33 without. Images reconstructed using SR-DLR were compared with original images. Quantitative metrics included signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the common bile duct (CBD) and PCLs, as well as full width at half maximum (FWHM), edge rise distance (ERD), and edge rise slope (ERS) of the CBD and main pancreatic duct (MPD). Qualitative evaluation was conducted by three radiologists, assessing the depiction of PCLs and the MPD, image sharpness, noise, artifacts, overall image quality, and the connection of PCLs and MPD. Quantitative and qualitative metrics were compared using paired t-test and the Wilcoxon signed rank test. SR-DLR significantly enhanced SNR and CNR (p < 0.001). Image sharpness was also enhanced, as shown by lower ERD and higher ERS in both CBD and MPD, together with reduced FWHM of the MPD (p < 0.005). Qualitative assessments indicated improved depiction of PCLs and image sharpness with SR-DLR across all readers (p ≤ 0.017). Most readers also reported improved visualization of the MPD and reduced noise, and overall quality. There was no statistically significant difference in determining the connectivity between PCLs and the MPD. SR-DLR significantly enhances image quality in MRCP, improving visualization of PCLs. These findings suggest that SR-DLR can contribute to appropriate management of PCLs.

While a larger fraction of head and neck squamous cell carcinoma (HNSCC) genomes is characterized by a high prevalence of copy number alterations (CNA-positive), a smaller subset with more favorable oncologic outcome is instead driven by somatic mutations (CNA-negative). We aimed to investigate the radiomic phenotypes of CNA-positive and -negative HNSCCs in contrast CT images. Single nucleotide polymorphism (SNP)-array copy number data were utilized and CNA-based hierarchical clustering of patients was performed to define CNA subclasses. Radiomic features (n=1037) quantifying shape, first-order intensity, and texture were extracted from HNSCC primary tumors in pretherapeutic neck CTs. We performed univariate association analyses and trained, optimized and validated radiomics-based CNA prediction models by combining feature selection algorithms with machine learning classifiers. A total of 522 and 114 patients were included in the copy number and radiomic analyses, respectively. Univariate analysis revealed 190 features from all feature subtypes (shape, first-order, texture) were significantly associated with the CNA status; after multiple testing correction, 29 texture or first-order features remained significant. The best-performing CNA status prediction model utilized a support vector machine classifier, achieving an AUC of 0.71 (95% confidence interval: 0.60-0.83). CNA subgroups exhibit distinct radiomic phenotypes, primarily reflected in texture and intensity characteristics. These findings enhance our understanding of the biological significance of radiomic information in HNSCC. In the clinical setting, as CNA-positive and -negative HNSCCs may emerge as distinct subclasses with unique staging schemes and treatment implications, improved CT radiomics-based prediction models could offer a noninvasive, cost-effective method for CNA subtyping.

Alzheimer's disease (AD) is a progressive neurodegenerative disorder and the primary cause of dementia, responsible for 60-70% of global cases. It severely affects memory, cognitive function, and daily independence, placing a substantial emotional and economic burden on patients and caregivers. Early and accurate prediction remains difficult due to the high cost of neuroimaging, scarcity of annotated datasets, and the "black-box" nature of most artificial intelligence (AI) models. With the emergence of Healthcare 5.0, the Internet of Medical Things (IoMT) offers new opportunities for patient-centric, real-time monitoring and data-driven diagnosis. This study proposes an IoMT-driven Alzheimer's prediction framework that combines transfer learning (ResNet152) with explainable AI (XAI) to provide both accuracy and interpretability. The publicly available Kaggle Alzheimer's MRI dataset, comprising 33,984 images across four classes (Non-Demented, Very Mild, Mild, and Moderate Demented) was employed. To address class imbalance, a Conditional Wasserstein GAN was applied for synthetic image generation and balanced sampling. The proposed ResNet152-TL-XAI model achieved 97.77% accuracy, with a precision of 0.981, recall of 0.987, F1-score of 0.983, and specificity of 99.13%, outperforming several state-of-the-art methods. Interpretability was ensured through Grad-CAM, SHAP, and LIME, which consistently highlighted clinically relevant brain regions such as the Hippocampus and ventricles, confirming biological plausibility and increasing clinician trust. By integrating IoMT-enabled data acquisition, transfer learning for efficient training, and multi-method XAI for transparency, the proposed pipeline demonstrates strong potential for early, accurate, and interpretable Alzheimer's staging. These results position the framework as a practical candidate for integration into Healthcare 5.0 ecosystems, supporting timely diagnosis, patient monitoring, and personalized interventions.