Hwiyoung Kim

Alcohol is a carcinogen; however, the risk associated with low-level drinking remains uncertain. We aimed to evaluate the associations of alcohol amount and drinking patterns with gastric cancer (GC) risk. We analyzed 6,007,469 Korean adults aged 50-69 who participated in the national health screening program in 2010-2011 and were followed for up to 10 years. Alcohol intake (g/day) was categorized as none, light, moderate, high, or heavy drinking. Targeted maximum likelihood estimation (TMLE) with Super Learner was applied to estimate risk ratios (RRs) for GC, adjusting for confounders selected using a directed acyclic graph. Drinking patterns were assessed based on drinking frequency, independent of total alcohol intake. GC risk increased progressively with higher alcohol intake, from light to heavy drinking, compared with that in non-drinkers, with the highest risk among heavy drinkers (5-year RR: 1.46; 95% CI: 1.41-1.51; 10-year RR: 1.43; 95% CI: 1.39-1.47). Even light drinking was associated with modest, significant increases in risk (5-year RR: 1.08; 95% confidence interval [CI]: 1.05-1.11; 10-year RR: 1.07; 95% CI: 1.04-1.10), although the low-level association was attenuated in a sensitivity analysis using an expanded adjustment set. Analyses incorporating repeated alcohol measurements from three time points yielded consistent results. Among light drinkers, associations were more pronounced among men and among former and current smokers. Consuming ≥ 1 drink per day was associated with higher GC risk than < 1 drink per day (5-year RR: 1.16; 95% CI: 1.12-1.19; 10-year RR: 1.19; 95% CI: 1.16-1.22). Within the same intake categories, individuals who drank more frequently showed higher GC risk than those with binge-type drinking patterns. These findings suggest that alcohol consumption is associated with increased GC risk in a dose-response manner. The association at low-level intake was modest and should be interpreted with caution, but these findings support recommendations to limit alcohol consumption for cancer prevention.

Introduction: Atrial fibrillation and flutter are heart rhythm disorders frequently associated with multiple other chronic conditions, complicating their management and requiring optimized care. Analyzing pre-atrial fibrillation and flutter comorbidity patterns could enable proactive, preventive, and personalized healthcare. Methods: This population-based nested case-control study analyzed data from the Korean National Health Insurance Corporation (2002–2019). Adults aged ≥19 years with at least three years of recorded claims were included. Cases were individuals newly diagnosed with atrial fibrillation and flutter between 2007 and 2019 following a washout period (2002–2006). Controls were matched 1:4 using stratified random sampling. Using 5-year disease histories, BEHRT, a transformer-based model, predicted atrial fibrillation and flutter, while BERTopic identified sex-specific multimorbidity patterns. Predictive performance was evaluated using the area under the receiver operating characteristic curve (AUC). Results: BEHRT achieved an AUC of 0.80 for predicting atrial fibrillation and flutter among 600,030 participants (8,661 cases and 591,369 controls). BERTopic analysis revealed sex-specific multimorbidity patterns: aortic aneurysm, hypertensive heart disease, and chronic obstructive pulmonary disease were common in males, while Alzheimer's disease, Parkinson's disease, and rheumatic heart disease were prominent in females. Discussion: The combination of BEHRT and BERTopic demonstrated the ability to predict atrial fibrillation and flutter based on multimorbid histories while identifying distinct sex-specific disease patterns. These findings underscore the potential for artificial intelligence to enhance personalized healthcare and optimize prevention and management strategies for chronic conditions.

Current liver cancer screening in Korea focuses on viral hepatitis or cirrhosis, despite rising risks from metabolic and alcohol-related liver disease. We

Early Autism Spectrum Disorder (ASD) identification is crucial but resource-intensive. This study evaluated a novel two-stage multimodal AI framework for scalable ASD screening using data from 1242 children (18–48 months). A mobile application collected parent-child interaction audio and screening tool data (MCHAT, SCQ-L, SRS). Stage 1 differentiated typically developing from high-risk/ASD children, integrating MCHAT/SCQ-L text with audio features (AUROC 0.942). Stage 2 distinguished high-risk from ASD children by combining task success data with SRS text (AUROC 0.914, Accuracy 0.852). The model’s predicted risk categories strongly agreed with gold-standard ADOS-2 assessments (79.59% accuracy) and correlated significantly (Pearson r = 0.830, p < 0.001). Leveraging mobile data and deep learning, this framework demonstrates potential for accurate, scalable early ASD screening and risk stratification, supporting timely interventions.

The accurate and early detection of vertebral metastases is crucial for improving patient outcomes. Although deep-learning models have shown potential in this area, their lack of prediction reliability and robustness limits their clinical utility. To address these challenges, we propose a novel technique called Ensemble Monte Carlo Dropout (EMCD) for uncertainty quantification (UQ), which combines the Monte Carlo dropout and deep ensembles. In this retrospective study, we analyzed 11,468 abdominal computed tomography images from 116 patients diagnosed with vertebral metastases and 957 images from 11 healthy controls. Uncertainty was quantified and visualized using single number, predictive probability interval, posterior distribution and uncertainty class activation maps to provide a detailed understanding of prediction confidence. The EMCD model demonstrated superior performance compared with traditional UQ methods, achieving an area under the receiver operating characteristic curve (AUC) of 0.93 and an expected calibration error of 0.09, indicating high predictive accuracy and reliability. In addition, the model exhibited strong performance in handling out-of-distribution data. When data retention was applied based on uncertainty values, the AUC of the model improved to 0.96, highlighting the potential of uncertainty-driven data selection to enhance performance. The EMCD model represents a significant advancement in the automated detection of vertebral metastases, providing superior diagnostic accuracy and introducing a robust UQ framework to aid clinicians in making informed decisions.

Hwang EJ, et al. Korean J Radiol. 2024 Jul;25(7):613-622. https://doi.org/10.3348/kjr.2023.1246