Abstract In order to improve patient outcomes and treatment processes, early and precise diagnosis of brain tumors is crucial. Despite the strong performance shown by deep learning models in classifying brain tumors from MRI scans, integrating these models into user-friendly clinical workflows remains challenging. This study presents a web-based AI platform designed for brain tumor classification using MRI images. The platform enables users to upload MRI scans, apply standardized preprocessing, select between different deep learning models, and receive real-time predictions. A custom convolutional neural network (CNN), pretrained VGG19, and pretrained Xception were evaluated using a publicly available T1-weighted contrast-enhanced brain MRI dataset of approximately 7,000 images across four classes: glioma, meningioma, pituitary tumor, and no tumor. The dataset lacks patient demographic and clinical metadata, limiting assessment of clinical generalizability. The models achieved test accuracies of 96.03%, 98%, and 96%, respectively, with VGG19 demonstrating the best performance. The study is limited by the use of a single public, slice-based dataset and the absence of external clinical validation; future work will focus on multi-sequence MRI and independent clinical cohorts.
An efficient workflow in the cardiac catheterization laboratory is important to guarantee quality of care and keep healthcare affordable. Automated surgical process monitoring of diagnostic coronary angiograms (CAGs) has the potential to increase workflow efficiency, reduce non-clinical workload, and optimize scheduling. The aim of this study is to automatically classify workflow phases of CAGs using deep learning models trained on a real-world dataset. With an observational study we collected a dataset, composed of an integration of video recorded CAGs, clinical data, and system logging data. Every procedural phase was annotated by a medical doctor based on clinically relevant key points in time. A Scaled-YOLOv4 object detection algorithm was employed to extract features from the video images. We implemented Long short-term memory (LSTM) and LSTM-Fully convolutional network (LSTM-FCN) on the integrated time series data to classify the workflow phases. The accuracy in the classification was weighted on the duration of the phase. Results: A dataset including video, clinical and logging data of 222 CAG procedures was obtained. A weighted accuracy for phase classification of 91.3
Gait speed is a key clinical indicator in neurological and orthopaedic conditions, yet waveform-level adaptations in ground reaction forces (GRF) and multi-muscle electromyography (EMG) remain poorly characterised. Existing approaches often analyse discrete outcomes or individual modalities, leaving limited integration of continuous waveform inference, dimensionality reduction, explainable machine learning, and equivalence testing within a unified multimodal framework. To compare three-axis GRF and six-muscle EMG between slow (0.5 m/s) and fast (1.0 m/s) treadmill walking using statistical parametric mapping (SPM), functional principal component analysis (fPCA), explainable machine learning, and equivalence testing. Fifty-eight healthy adults were analysed (55 with complete EMG). Paired SPM with cluster-based permutation assessed waveform differences. fPCA-derived features entered a Random Forest with leave-one-subject-out cross-validation and SHAP interpretability. Two one-sided tests (TOST) assessed equivalence of the vertical GRF. No significant cluster-level SPM differences were found for any GRF component. In contrast, significant EMG clusters were detected in tibialis anterior (ten clusters), gastrocnemius medial and lateral, vastus lateralis, rectus femoris, and semitendinosus. The Random Forest achieved 87.2
Tracheal reconstruction remains a significant clinical challenge due to the difficulty of maintaining airway patency and promoting epithelial regeneration following tracheal injury or resection. Recent advances in tissue engineering have explored 3D-printed scaffolds as potential solutions; however, limitations remain regarding scaffold biocompatibility, mechanical stability, and integration with host tissue. This study evaluated the feasibility of a non-biodegradable polyurethane–polyvinyl chloride (PU-PVC) composite scaffold for partial tracheal wall reconstruction in a rabbit model. Nine New Zealand white rabbits underwent anterior tracheal wall reconstruction using a customized half-pipe PU-PVC scaffold. Scaffold mechanical properties were characterized before implantation. Animals were monitored for 30 days, after which bronchoscopic evaluation was performed to assess luminal patency. Histological analysis was conducted to evaluate epithelialization and inflammatory response. Explanted scaffolds were also assessed for structural integrity and mechanical stability. The 30-day survival rate was 88.9
Epilepsy remains a major global health concern, particularly in regions where continuous medical monitoring is difficult to implement. This study introduces a wearable system powered by edge-based artificial intelligence, designed to detect epileptic seizures in real time. The device integrates multiple sensors—accelerometers for motion tracking, photoplethysmography (PPG) for cardiovascular monitoring, and GPS for location detection—to enhance reliability through sensor fusion. Multiple machine learning models, including support vector machines (SVM), neural networks (NN), and random forest classifiers, were deployed and assessed directly on the device. Among these, the optimized random forest algorithm achieved the highest accuracy and fastest response time. The fusion of sensor data significantly improved specificity and maintained a very low false alarm rate. When a seizure is detected, the system instantly sends SMS alerts with precise location details to assigned caregivers, enabling prompt medical intervention. Simulation results demonstrate that this cost-effective and self-contained platform offers strong potential for improving patient safety and facilitating rapid response in low-resource settings where conventional monitoring tools are unavailable.
