CNNs have demonstrated superior performance in medical image segmentation. To overcome the limitation of only using local receptive field, previous work has attempted to integrate Transformers into convolutional network components such as encoders, decoders, or skip connections. However, these methods can only establish long-distance dependencies for some specific patterns and usually neglect the loss of fine-grained details during downsampling in multi-scale feature extraction. To address the issues, we present a novel hybrid Transformer network called FocalTransNet. specifically, we construct a focal-enhanced (FE) Transformer module by introducing dense cross-connections into a CNN-Transformer dual-path structure and deploy the FE Transformer throughout the entire encoder. Different from existing hybrid networks that employ embedding or stacking strategies, the proposed model allows for a comprehensive extraction and deep fusion of both local and global features at different scales. Besides, we propose a symmetric patch merging (SPM) module for downsampling, which can retain the fine-grained details by stablishing a specific information compensation mechanism. We evaluated the proposed method on four different medical image segmentation benchmarks. The proposed method outperforms previous state-of-the-art convolutional networks, Transformers, and hybrid networks. The code for FocalTransNet is publicly available at https://github.com/nemanjajoe/FocalTransNet.

Precise and rapid identification of knee osteoarthritis (OA) is essential for efficient management and therapy planning. Conventional diagnostic techniques frequently depend on subjective interpretation, which have shortcomings, particularly during the first phases of the illness. In this study, magnetic resonance imaging (MRI) was used to create knee datasets as novel techniques for evaluating knee OA. This methodology utilizes artificial intelligence (AI) algorithms to identify and evaluate important indications of knee osteoarthritis, including osteophytes, eburnation, bone marrow lesions (BMLs), and cartilage thickness. We conducted training and evaluation on multiple deep learning models, including ResNet50, DenseNet121, VGG16 and ResNet101 utilizing annotated MRI data. By conducting thorough statistical analysis and validation, we have proven the efficacy of our models in precisely diagnosing and grading knee OA. This research presents a new grading method, verified by experienced radiologists, that uses eburnation as a significant indicator of the severity of knee OA. This study provides a new method for an AI-powered automated system designed to diagnose knee OA. This system will simplify the diagnostic process, minimize mistakes made by humans, and enhance the effectiveness of clinical treatment. Through the integration of AI-ML (machine learning) technologies, our goal is to improve patient outcomes, optimize the utilization of healthcare resources, and enable personalized knee OA therapy.

Magnetic resonance imaging (MRI) is a cornerstone of medical imaging, celebrated for its non-invasiveness, high spatial and temporal resolution, and exceptional soft tissue contrast, with over 100 million clinical procedures performed annually worldwide. In this field, MRI-based nanosensors have garnered significant interest in biomedical research due to their tunable sensing mechanisms, high permeability, rapid kinetics, and surface functionality. Extensive studies in the field have reported the use of superparamagnetic iron oxide nanoparticles (SPIONs) and proteins as a proof-of-concept for sensing critical neurochemicals via MRI. However, the signal change ratio and response rate of our SPION-protein-based in vitro dopamine and in vivo calcium sensors need to be further enhanced to detect the subtle and transient fluctuations in neurochemical levels associated with neural activities, starting from in vitro diagnostics. In this paper, we present an advanced reinforcement-learning-based computational model that treats sensor design as an optimal decision-making problem by choosing sensor performance as a weighted reward objective function. The adjustments of the SPION's and protein's three-dimensional configuration and magnetic moment establish a set of actions that can autonomously maximize the cumulative reward in the computational environment. Our new model first elucidates the sensor's conformation alteration behind the increment in T₂ contrast observed experimentally in MRI in the presence and absence of calcium and dopamine neurochemicals. Additionally, our enhanced machine-learning algorithm can autonomously learn the performance trends of SPION-protein-based sensors and identify their optimal structural parameters. Experimental in vitro validation with TEM and MR relaxometry confirmed the predicted optimal SPION diameters, demonstrating the highest sensing performance at 9 nm for calcium and 11 nm for dopamine detection. Beginning with in vitro diagnostics, these results demonstrate a versatile modeling platform for the development of MRI-based neurochemical sensors, providing insights into their behavior under operational conditions. This platform also enables the autonomous design of improved sensor sizes and geometries, providing a roadmap for the future optimization of MRI sensors.

Post-stroke cognitive impairment (PSCI) is a common and debilitating consequence of stroke that often arises from complex interactions between diverse brain alterations. The accurate early prediction of PSCI is critical for guiding personalized interventions. However, existing methods often struggle to capture complex structural disruptions and integrate multimodal information effectively. This study proposes the multimodal dynamic hierarchical clustering network (MDHCNet), a graph neural network designed for accurate and interpretable PSCI prediction. MDHCNet constructs brain graphs from diffusion-weighted imaging, magnetic resonance angiography, and T1- and T2-weighted images and integrates them with clinical features using a hierarchical cross-modal fusion module. Experimental results using a real-world stroke cohort demonstrated that MDHCNet consistently outperformed deep learning baselines. Ablation studies validated the benefits of multimodal fusion, while saliency-based interpretation highlighted discriminative brain regions associated with cognitive decline. These findings suggest that MDHCNet is an effective and explainable tool for early PSCI prediction, with the potential to support individualized clinical decision-making in stroke rehabilitation.

This study explored the feasibility of preoperatively predicting perineural invasion (PNI) of intrahepatic cholangiocarcinoma (ICC) through machine learning based on clinical and CT image features, which may help in individualized clinical decision making and modification of further treatment strategies. This study enrolled 199 patients with histologically confirmed ICC from three institutions for final analysis. 111 patients from Institution I were recruited as the training cohort and internal validation cohort. Significant clinical and CT image features for predicting PNI were screened using the least absolute shrinkage and selection operator (LASSO) to construct machine learning models. 72 patients from Institutions II and III were recruited as two external validation cohorts, and 16 patients from Institution I were enrolled as a prospective cohort to assess model performance. Tumor location (perihilar), intrahepatic bile duct dilatation, and arterial enhancement pattern were selected using LASSO for model construction. Machine learning models were developed based on these three features using five algorithms: multilayer perceptron, random forest, support vector machine, logistic regression, and XGBoost. The AUCs of the models exceeded 0.86, 0.84, 0.79, and 0.72 in the training cohort, internal validation cohort, external validation cohorts, and prospective cohort, respectively. Machine learning models based on CT were accurate in predicting PNI of ICC, which may help in treatment decision making.

This study aimed to evaluate and compare the diagnostic performance of various Thyroid Imaging Reporting and Data Systems (TIRADS), with a particular focus on the artificial intelligence-based TIRADS (AI-TIRADS), in characterizing thyroid nodules. In this retrospective study conducted between April 2016 and May 2022, 1,322 thyroid nodules from 1,139 patients with confirmed cytopathological diagnoses were included. Each nodule was assessed using TIRADS classifications defined by the American College of Radiology (ACR-TIRADS), the American Thyroid Association (ATA-TIRADS), the European Thyroid Association (EU-TIRADS), the Korean Thyroid Association (K-TIRADS), and the AI-TIRADS. Three radiologists independently evaluated the ultrasound (US) characteristics of the nodules using all classification systems. Diagnostic performance was assessed using sensitivity, specificity, positive predictive value (PPV), and negative predictive value, and comparisons were made using the McNemar test. Among the nodules, 846 (64%) were benign, 299 (22.6%) were of intermediate risk, and 147 (11.1%) were malignant. The AI-TIRADS demonstrated a PPV of 21.2% and a specificity of 53.6%, outperforming the other systems in specificity without compromising sensitivity. The specificities of the ACR-TIRADS, the ATA-TIRADS, the EU-TIRADS, and the K-TIRADS were 44.6%, 39.3%, 40.1%, and 40.1%, respectively (all pairwise comparisons with the AI-TIRADS: <i>P</i> < 0.001). The PPVs for the ACR-TIRADS, the ATA-TIRADS, the EU-TIRADS, and the K-TIRADS were 18.5%, 17.9%, 17.9%, and 17.4%, respectively (all pairwise comparisons with the AI-TIRADS, excluding the ACR-TIRADS: <i>P</i> < 0.05). The AI-TIRADS shows promise in improving diagnostic specificity and reducing unnecessary biopsies in thyroid nodule assessment while maintaining high sensitivity. The findings suggest that the AI-TIRADS may enhance risk stratification, leading to better patient management. Additionally, the study found that the presence of multiple suspicious US features markedly increases the risk of malignancy, whereas isolated features do not substantially elevate the risk. The AI-TIRADS can enhance thyroid nodule risk stratification by improving diagnostic specificity and reducing unnecessary biopsies, potentially leading to more efficient patient management and better utilization of healthcare resources.

Age estimation plays a major role in the identification of unknown dead bodies, including skeletal remains. We present a novel age estimation method developed by applying a deep-learning network to the coxal bone and lumbar vertebrae on post-mortem computed tomography (PMCT) images. The coxal bone and lumbar vertebrae were targeted in this study. Volume-rendered images of these bones from 1,229 individuals were captured and input to a convolutional neural network based on the visual geometry group 16 network. A transfer learning strategy was employed. The predictive capabilities of age estimation models were assessed by a 10-fold cross-validation procedure, with mean absolute error (MAE) and correlation coefficients between chronological and estimated ages calculated for validation. In addition, gradient-weighted class activation mapping (Grad-CAM) was conducted to visualize the regions of interest in learning. The estimation models created showed low MAE (range, 7.27-6.44 years) and high correlation coefficients (range, 0.84-0.91) in the validation. Aging-induced shape changes were grossly observed at the vertebral body, coxal bone surface, and other sites. The Grad-CAM results identified these as regions of interest in learning. The present method has the potential to become an age estimation tool that is routinely applied in the examination of unknown dead bodies, including skeletal remains.

The respiratory system depends on complex biomechanical processes to enable gas exchange. The mechanical properties of the lung parenchyma, airways, vasculature, and surrounding structures play an essential role in overall ventilation efficacy. These complex biomechanical processes however are significantly altered in chronic obstructive pulmonary disease (COPD) due to emphysematous destruction of lung parenchyma, chronic airway inflammation, and small airway obstruction. Recent advancements computed tomography (CT) and magnetic resonance imaging (MRI) acquisition techniques, combined with sophisticated image post-processing algorithms and deep neural network integration, have enabled comprehensive quantitative assessment of lung structure, tissue deformation, and lung function at the tissue level. These methods have led to better phenotyping, therapeutic strategies and refined our understanding of pathological processes that compromise pulmonary function in COPD. In this review, we discuss recent developments in imaging and image processing methods for studying pulmonary biomechanics with specific focus on clinical applications for chronic obstructive pulmonary disease (COPD) including the assessment of regional ventilation, planning of endobronchial valve treatment, prediction of disease onset and progression, sizing of lungs for transplantation, and guiding mechanical ventilation. These advanced image-based biomechanical measurements when combined with clinical expertise play a critical role in disease management and personalized therapeutic interventions for patients with COPD.

Microscopic tumor cell infiltration beyond contrast-enhancing regions influences glioblastoma prognosis but remains undetectable using conventional MRI. To develop and evaluate the glioblastoma infiltrating area interactive detection framework (GIAIDF), an interactive deep-learning framework that integrates diffusion tensor imaging (DTI) biomarkers for identifying microscopic infiltration within peritumoral edema. Retrospective. A total of 73 training patients (51.13 ± 13.87 years; 47 M/26F) and 25 internal validation patients (52.82 ± 10.76 years; 14 M/11F) from Center 1; 25 external validation patients (47.29 ± 11.39 years; 16 M/9F) from Center 2; 13 prospective biopsy patients (45.62 ± 9.28 years; 8 M/5F) from Center 1. 3.0 T MRI including three-dimensional contrast-enhanced T1-weighted BRAVO sequence (repetition time = 7.8 ms, echo time = 3.0 ms, inversion time = 450 ms, slice thickness = 1 mm), three-dimensional T2-weighted fluid-attenuated inversion recovery (repetition time = 7000 ms, echo time = 120 ms, inversion time = 2000 ms, slice thickness = 1 mm), and diffusion tensor imaging (repetition time = 8500 ms, echo time = 63 ms, slice thickness = 2 mm). Histopathology of 25 stereotactic biopsy specimens served as the reference standard. Primary metrics included AUC, accuracy, sensitivity, and specificity. GIAIDF heatmaps were co-registered to biopsy trajectories using Ratio-FAcpcic (0.16-0.22) as interactive priors. ROC analysis (DeLong's method) for AUC; recall, precision, and F1 score for prediction validation. GIAIDF demonstrated recall = 0.800 ± 0.060, precision = 0.915 ± 0.057, F1 = 0.852 ± 0.044 in internal validation (n = 25) and recall = 0.778 ± 0.053, precision = 0.890 ± 0.051, F1 = 0.829 ± 0.040 in external validation (n = 25). Among 13 patients undergoing stereotactic biopsy, 25 peri-ED specimens were analyzed: 18 without tumor cell infiltration and seven with infiltration, achieving AUC = 0.929 (95% CI: 0.804-1.000), sensitivity = 0.714, specificity = 0.944, and accuracy = 0.880. Infiltrated sites showed significantly higher risk scores (0.549 ± 0.194 vs. 0.205 ± 0.175 in non-infiltrated sites, p < 0.001). This study has provided a potential tool, GIAIDF, to identify regions of GBM infiltration within areas of peri-ED based on preoperative MR images.

Metabolomics has been associated with cognitive decline and dementia, but the relationship between metabolites and brain aging remains unclear. We aimed to investigate the associations of metabolomics with brain age assessed by neuroimaging and to explore whether these relationships vary according to apolipoprotein E (APOE) ε4. This study included 17,770 chronic brain disorder-free participants aged 40-69 years from UK Biobank who underwent neuroimaging scans an average of 9 years after baseline. A total of 249 plasma metabolites were measured using nuclear magnetic resonance spectroscopy at baseline. Brain age was estimated using LASSO regression and 1079 brain MRI phenotypes and brain age gap (BAG; i.e., brain age minus chronological age) was calculated. Data were analyzed using linear regression. We identified 64 and 77 metabolites associated with brain age and BAG, respectively, of which 55 overlapped. Lipids (including cholesterol, cholesteryl esters, free cholesterol, phospholipids, and total lipids) in S/M-HDL, as well as phospholipids and triglycerides as a percentage of total lipids in different-density lipoproteins, were associated with larger BAG. The percentages of cholesterol, cholesteryl esters, and free cholesterol to total lipids in VLDL, LDL, and HDL of different particle sizes were associated with smaller BAG. The associations of LA/FA, omega-6/FA, SFA/FA, and phospholipids to total lipids in L-HDL with brain age were consistent across APOE ε4 carriers and non-carriers (all p for interaction > 0.05). Plasma metabolites show remarkably widespread associations with brain aging regardless of APOE ε4 genetic risk. Metabolic profiles could serve as an early indicator of accelerated brain aging.