2787 papers
This study employs two Long Short-Term Memory (LSTM) based machine learning models to detect the severity of keratoconus by analyzing patient eye parameters from OCT images, effectively monitoring disease progression by considering both unilateral and bilateral eye data.
This before-and-after study demonstrates that the Persian-language mobile application AIMedEdu significantly improved healthcare students' perceived medical AI readiness and received excellent usability ratings, though future research is needed to confirm long-term educational effectiveness and generalizability.
The paper introduces HAG-EUNet, a hybrid deep learning model combining an EfficientNet-B3 encoder with a transformer-enhanced U-Net decoder to achieve automated, high-precision choroidal segmentation in OCT images and extract 32 clinically relevant biomarkers, significantly outperforming standard U-Net baselines in both accuracy and efficiency.
KORAL-Net is an ordered 3D segmentation architecture that separates volumetric representation from multi-scale feature reconstruction to preserve boundary evidence in three-phase DCE-MRI, achieving superior Dice scores and reduced false positives compared to nnU-Net v2 on both internal and cross-cohort I-SPY datasets.
This paper proposes a Universal Autonomous Driving System (UADS) that integrates a high-precision, evolved YOLO detector with an Artificial Neural Expert System (ANES) and Large Language Models (LLM) to achieve state-of-the-art object detection and autonomous decision-making while addressing NP-hard challenges in autonomous driving.
This paper introduces PASER-NLL, a teacher-free data selection method that replaces the failing KL-divergence signal of pruned models with the student's negative log-likelihood to achieve superior recovery performance and significant resource savings under aggressive structural pruning.
This paper presents an AI-assisted control architecture for the Elettra 2.0 particle accelerator that integrates Large Language Models and Model Context Protocol within a hierarchical framework to enable safe, natural-language supervisory control via a Digital Twin, achieving over 95% command success rates without unsafe actions.
This study proposes a Bidirectional Long Short-Term Memory (BiLSTM) model enhanced with Multi-Head Attention (MHA) to address limitations in Neural Machine Translation, demonstrating superior performance in alignment accuracy, BLEU, and METEOR scores compared to existing methods.
This study introduces MAL-BAYES-III, a partial-Bayesian framework that applies Hamiltonian Monte Carlo exclusively to a compact classification head to generate calibrated, uncertainty-aware probabilities for detecting obfuscated malware in memory forensics, offering diagnostic value and selective review capabilities despite not surpassing traditional gradient boosting in raw accuracy.
This study introduces an unsupervised Graph Attention Network Autoencoder (GAT-AE) framework that fuses multi-source geospatial data with street network topology to model dynamic functional urban centrality, demonstrating superior spatial coherence and the ability to identify complex non-linear urban phenomena compared to traditional linear methods.