Myungeun Lee

Accurate QT-interval measurement requires reliable identification of QRS onset and T-wave offset. We present QTANet, a single-lead framework that classifies ECG samples as baseline, QRS or T wave and derives QT from the detected landmarks. We used 179 recordings from the Lobachevsky University Electrocardiography Database, with a fixed test set and patient-wise cross-validation, and external evaluation without fine-tuning on the QT Database (QTDB) and 33 clinical recordings. On the expert-annotated QTDB interval subset of 101 recordings, QTANet achieved a mean absolute error of 34.94 ms and measurement coverage of 95.2%, compared with 40.55 ms and 88.2% for the control without attention and 52.41 ms and 68.7% for NeuroKit2. In a separate expert-annotated subset permitting frame-level evaluation, attention increased per-recording T-wave F1 by 0.061, with improvement in 30 of 41 recordings. Attention most consistently reduced T waves misread as QRS complexes. QTANet outperformed an identically trained published U-Net externally at frame level but had higher internal QT error; a larger modern U-Net outperformed QTANet at frame level and in QTDB QT error. Internal interval-level gains were small and clinical results varied across training seeds. These findings support boundary-derived QT measurement with jointly reported error and coverage; clinical utility requires further validation.

Automated radiology report generation requires localized visual evidence and broader anatomical context. We propose ADAPT, an anatomy-guided framework that constructs compact visual prompts for large language model (LLM)-based report generation. Anatomy-Guided Token Pooling (AGTP) introduces anatomical segmentation masks as soft additive biases in cross-attention, steering learnable region queries toward relevant patches while retaining access to the complete patch sequence. A Dynamic Scale Predictor (DSP) estimates the bias strength for each examination, region, and attention head from mask-weighted visual features. The resulting tokens are projected into a frozen LLM to generate reports. Experiments on the Indiana University Chest X-Ray (IU-Xray) and Medical Information Mart for Intensive Care–Chest X-Ray (MIMIC-CXR) datasets show gains over an R2GenGPT baseline reproduced under the same training protocol, with all reported NLG and clinical label-agreement metrics improving on MIMIC-CXR and six of seven NLG metrics improving on IU-Xray. On MIMIC-CXR, ADAPT improves the Bilingual Evaluation Understudy (BLEU) scores BLEU-3 and BLEU-4 from 0.163 and 0.117 to 0.184 and 0.134, respectively, and CheXbert precision from 0.506 to 0.545. On IU-Xray, it achieves the highest Consensus-based Image Description Evaluation (CIDEr) score among the compared methods reporting that metric. Component ablations show that Region Token Pooling (RTP) accounts for much of the improvement over the baseline. On MIMIC-CXR, dynamic anatomical guidance yields the highest NLG point estimates among the tested fixed-scale alternatives, while learned-scale analysis reveals region- and input-dependent modulation. These findings support compact region-organized pooling with examination-conditioned anatomical guidance as an effective approach to visual-prompt construction.

With the increasing prevalence of neurodegenerative diseases driven by population aging, imaging-based biomarkers are needed to quantify brain aging at an early stage. Brain age, which estimates structural brain aging relative to chronological age, has emerged as a useful indicator. Prior work has mainly used T1-weighted MRI with deep learning models such as convolutional neural networks (CNNs) or transformers; however, many approaches insufficiently capture three-dimensional structural continuity and localized anatomical patterns, and tissue-specific aging in gray matter (GM) and white matter (WM) is often treated as auxiliary. To address these limitations, we propose a 3D Global–Patch Transformer framework for brain age prediction that directly processes volumetric data while jointly learning global brain structure and local anatomical features. Our model runs global and patch pathways in parallel and explicitly incorporates GM and WM structural maps alongside T1-weighted MRI to encode tissue-specific aging signals. Experiments on multiple public datasets, including IXI and OASIS, show that the proposed method reduces mean absolute error (MAE) by approximately 10–15% compared with CNN-based and single-input transformer baselines, with notably improved performance in older populations, highlighting the value of tissue-level structural information for brain age estimation.

Background and ObjectivesAccurate measurement of the QT interval from electrocardiogram (ECG) signals is essential for assessing ventricular repolarization and arrhythmia risk; however, it remains a challenging task due to inter-observer variability in manual annotation and the limited robustness of conventional automated methods, particularly in detecting the T-wave offset. Many existing automated approaches estimate the QT interval as a single regression value, which restricts clinical interpretability because waveform boundaries cannot be explicitly verified. This study proposes QTANet, a boundary-aware framework that provides clinically interpretable QT interval estimation by explicitly detecting QRS onset and T-wave offset.MethodsQTANet performs frame-level ECG waveform classification and derives QT intervals based on detected physiological landmarks. The framework integrates a ResNet-based convolutional encoder for morphological feature extraction, discrete wavelet transform for noise-robust frequency analysis, bidirectional LSTM for temporal context modeling, and self-attention mechanisms that emphasize boundary transitions. A dynamically weighted focal loss was employed to address class imbalance. The model was trained and validated on the Lobachevsky University Database (LUDB) using five-fold cross-validation, and externally evaluated on the QT Database (QTDB; n=105) and an independent clinical dataset from Chonnam National University Hospital (CNUH; n=23).ResultsOn LUDB, QTANet achieved a frame-level classification accuracy of 94.5% (F1: 92.8%) and a QT interval estimation MAE of 12.08 ± 17.65 ms (Pearson’s r = 0.84). On external datasets, the model maintained consistent performance (QTDB: MAE 28.63 ± 36.11 ms, r = 0.63; CNUH: MAE 17.47 ± 23.17 ms, r = 0.78), outperforming both the ResNet+LSTM regression baseline and the rule-based NeuroKit2 approach across all metrics.ConclusionsBoundary-aware QT interval estimation through explicit waveform landmark detection provides a clinically interpretable and robust alternative to black-box regression methods. The proposed framework enables visual verification of detected boundaries and demonstrates consistent performance across heterogeneous recording conditions, suggesting its potential utility as a decision-support tool in cardiac monitoring and pharmacovigilance.

Large language model (LLM)-based virtual patient (VP) simulations are emerging to complement traditional medical and nursing education by enabling safe, repeatable, and context-rich clinical practice. This review synthesizes recent developments from 2023 to 2025, mapping implementation approaches, data practices, evaluation methods, and cross-cutting challenges across forty studies. Six implementation categories are identified: scenario generation; prompt-driven VPs; feedback-integrated automated scoring; realism- and adaptability-enhanced systems; knowledge-driven and multi-agent hybrids; and mental health-oriented systems. The analysis summarizes dataset usage (including knowledge sources and governance) and evaluation frameworks, and it introduces quantitative indicators for reproducible assessment. Persistent challenges include factual accuracy, role consistency, emotional realism, and ethical and legal accountability. Overall, LLM-based VP systems show growing potential to extend simulation-based learning, but stronger evidence from multi-site controlled studies, standardized metrics, transparent reporting (model versions, prompts), and robust data governance is needed to establish educational validity and generalizability.

Federated Learning (FL) has emerged as a promising framework for multi-institutional medical artificial intelligence, enabling collaborative model development while preserving data privacy and security. Despite increasing research on federated approaches for cardiovascular disease prediction, previous reviews have largely focused on disease-specific perspectives without systematically comparing data modalities. This study comprehensively examines 28 representative investigations from the past five years, including 17 biosignal-based and 11 electronic health record (EHR)-based applications. Biosignal-based FL emphasizes personalized electrocardiogram (ECG) classification, mitigation of non-independent and identically distributed (Non-IID) data, and Internet of Things (IoT)-based monitoring using methods such as client clustering, asynchronous learning, and Bayesian inference. In contrast, EHR-based studies prioritize large-scale hospital collaboration, adaptive optimization, and secure aggregation through distributed frameworks. By systematically comparing methodological strategies, performance trade-offs, and clinical feasibility, this review highlights the complementary strengths of biosignal- and EHR-based approaches. Biosignal frameworks show strong potential for personalized, low-latency cardiac monitoring, whereas EHR frameworks excel in scalable and privacy-preserving decision support. Building upon the limitations of earlier reviews, this paper introduces data-type-centric design guidelines to enhance the reliability, interpretability, and clinical scalability of FL in cardiovascular diagnosis and prediction.

TumorPrism3D software was developed to segment brain tumors with a straightforward and user-friendly graphical interface applied to two- and three-dimensional brain magnetic resonance (MR) images. The MR images of 185 patients (103 males, 82 females) with glioblastoma multiforme were downloaded from The Cancer Imaging Archive (TCIA) to test the tumor segmentation performance of this software. Regions of interest (ROIs) corresponding to contrast-enhancing lesions, necrotic portions, and non-enhancing T2 high signal intensity components were segmented for each tumor. TumorPrism3D demonstrated high accuracy in segmenting all three tumor components in cases of glioblastoma multiforme. They achieved a better Dice similarity coefficient (DSC) ranging from 0.83 to 0.91 than 3DSlicer with a DSC ranging from 0.80 to 0.84 for the accuracy of segmented tumors. Comparative analysis with the widely used 3DSlicer software revealed TumorPrism3D to be approximately 37.4% faster in the segmentation process from initial contour drawing to final segmentation mask determination. The semi-automated nature of TumorPrism3D facilitates reproducible tumor segmentation at a rapid pace, offering the potential for quantitative analysis of tumor characteristics and artificial intelligence-assisted segmentation in brain MR imaging.