Seoul National University Hospital

Objective We aimed to prospectively investigate whether bladder volume measured using deep learning artificial intelligence (AI) algorithms (AI‐BV) is more accurate than that measured using conventional methods (C‐BV) if using a portable ultrasound bladder scanner (PUBS). Patients and Methods Patients who underwent filling cystometry because of lower urinary tract symptoms between January 2021 and July 2022 were enrolled. Every time the bladder was filled serially with normal saline from 0 mL to maximum cystometric capacity in 50 mL increments, C‐BV was measured using PUBS. Ultrasound images obtained during this process were manually annotated to define the bladder contour, which was used to build a deep learning AI model. The true bladder volume (T‐BV) for each bladder volume range was compared with C‐BV and AI‐BV for analysis. Results We enrolled 250 patients (213 men and 37 women), and a deep learning AI model was established using 1912 bladder images. There was a significant difference between C‐BV (205.5 ± 170.8 mL) and T‐BV (190.5 ± 165.7 mL) ( p = 0.001), but no significant difference between AI‐BV (197.0 ± 161.1 mL) and T‐BV (190.5 ± 165.7 mL) ( p = 0.081). In bladder volume ranges of 101–150, 151–200, and 201–300 mL, there were significant differences in the percentage of volume differences between [C‐BV and T‐BV] and [AI‐BV and T‐BV] ( p < 0.05), but no significant difference if converted to absolute values ( p > 0.05). C‐BV ( R ² = 0.91, p < 0.001) and AI‐BV ( R ² = 0.90, p < 0.001) were highly correlated with T‐BV. The mean difference between AI‐BV and T‐BV (6.5 ± 50.4) was significantly smaller than that between C‐BV and T‐BV (15.0 ± 50.9) ( p = 0.001). Conclusion Following image pre‐processing, deep learning AI‐BV more accurately estimated true BV than conventional methods in this selected cohort on internal validation. Determination of the clinical relevance of these findings and performance in external cohorts requires further study. Trial Registration: The clinical trial was conducted using an approved product for its approved indication, so approval from the Ministry of Food and Drug Safety (MFDS) was not required. Therefore, there is no clinical trial registration number.

Objectives To assess the temporal evolution and interobserver agreement of the early categories per the liver imaging reporting and data system (LI-RADS) radiation treatment response assessment (TRA) algorithm in patients receiving selective internal radiation therapy (SIRT) with Yttrium-90 for hepatocellular carcinoma (HCC). Materials and methods This single-center retrospective study included consecutive patients with treatment-naïve HCC who underwent serial contrast-enhanced CT/MRI before and after SIRT. Three masked radiologists independently evaluated response at 3–6 months. Another senior radiologist assessed response at 9, 12, 15, 18, 21, 24, and > 24 months after comprehensive review of available clinical-radiological information. Results 65 patients (mean age, 66.7 ± 11.2 years; 48 men) were included. At 3–6 months after SIRT, 47.7% (31/65) of lesions were assigned to the nonprogressing category, and the remaining 52.3% (34/65) to the nonviable category. Among early nonprogressing lesions, 64.5% (20/31) regressed to the nonviable category, 25.8% (8/31) remained nonprogressing, and 9.7% (3/31) evolved into the viable category at ≥ 12 months. The nonprogressing category decreased in number over time, with 61.3% (19/31) conversion to the nonviable category at 9 months. Among the early nonviable lesions, 91.2% (31/34) remained nonviable at ≥ 12 months, and 8.8% (3/34) evolved into the viable category. Agreement for the 3–6 months LR-TR category assignment was moderate (kappa = 0.46) with CT but almost perfect (kappa = 0.85) with MRI. Conclusions SIRT induced a delayed and sustained response in the majority of HCC patients after ≥ 12 months. MRI demonstrated superior agreement over CT in assessing response at 3–6 months. Key Points Question Tumor response to SIRT can change; there is limited evidence on the evolution of the imaging appearance of HCC following SIRT . Findings Sixty-four and five-tenths of early nonprogressing lesions regressed to nonviable, and 91.2% of early nonviable lesions remained free of viability. LR-TR category assignment agreement was moderate with CT but almost perfect with MRI . Clinical relevance SIRT induced a delayed and sustained response in HCC, underscoring the necessity of dynamic evaluation of long-term changes in treated lesions. MRI with subtraction imaging may be preferred over CT for long-term monitoring, which may help prevent premature retreatment decisions . Graphical Abstract

Purpose Long-term cardiovascular risk in adolescent and young adult (AYA) survivors of non-Hodgkin lymphoma (NHL) remains insufficiently characterized. This retrospective cohort study investigated the incidence of cardiovascular disease (CVD) among AYA survivors of NHL. Methods We identified 4553 individuals aged 15–39 years diagnosed with NHL between 2006 and 2019 using the Korean National Health Insurance System database. A control group of 13,659 individuals without a history of cancer or CVD was selected using 1:3 matching based on age, sex, and residential area. The primary outcomes were major adverse cardiovascular events such as myocardial infarction, cardiomyopathy, heart failure, ischemic stroke, and hemorrhagic stroke. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazard models. Results The mean (SD) age of the participants was 29.5 (6.8) years, and 59% were male. Over a median follow-up of 6.5 years, survivors of NHL had significantly higher risks of cardiomyopathy (HR 6.77; 95% CI 3.44–13.33), heart failure (HR 4.90; 95% CI 3.45–6.97), and hemorrhagic stroke (HR 3.14; 95% CI 1.75–5.65), compared to controls. In the subgroup analyses stratified by treatment modality, the highest risks were observed among patients who underwent hematopoietic stem cell transplantation, which involved high-dose chemotherapy with or without radiotherapy. The risk of myocardial infarction and ischemic stroke did not increase significantly. Conclusion AYA survivors of NHL had a significantly higher risk of CVD, including cardiomyopathy, heart failure, and hemorrhagic stroke, than the general population. Implications for Cancer Survivors Long-term cardiovascular surveillance is essential for AYA survivors of NHL, particularly those receiving intensive treatment.

Background Retinopathy of prematurity (ROP) is influenced by factors, including gestational age (GA), oxygen exposure, and chorioamnionitis. However, the association between histologic chorioamnionitis (HCA) and ROP remains controversial. This study aimed to investigate the association between HCA and severe ROP. Methods This retrospective cohort study utilized data from the National Korean Neonatal Network registry, focusing on infants with birthweights < 1500 g and GA < 32 weeks. Univariate and multivariate logistic regression analyses assessed the association between HCA and severe ROP. Sub-cohort analyses were performed to evaluate the effect of HCA on severe ROP across different GA groups. Results Infants in the HCA cohort had lower GA and birth weights, with a higher prevalence of any-stage and severe ROP compared to those in the without-HCA cohort. However, multivariate logistic regression showed an inverse association between HCA and severe ROP. Sub-cohort analyses revealed that HCA was associated with an increased risk of severe ROP in infants born at 26–28 and 28–31 weeks, while no significant association was observed in infants born at 23–25 weeks. Conclusions HCA may reduce the risk of severe ROP, suggesting that intrauterine inflammation could play a protective role. Further research is needed to elucidate underlying mechanisms. Impact Retinopathy of prematurity (ROP) is influenced by numerous perinatal and postnatal factors, including low gestational age, oxygen exposure, and chorioamnionitis. However, the association between chorioamnionitis and ROP remains controversial. Our study showed that histologic chorioamnionitis (HCA) was negatively correlated with severe ROP in preterm infants, even after adjusting for confounding factors such as gestational age and birth weight. These findings suggest that HCA may have a protective effect against severe ROP, potentially mediated by inflammatory markers.

Background Chronic obstructive pulmonary disease (COPD) is a common and progressive respiratory condition characterized by persistent airflow limitation and symptoms such as dyspnea, cough, and sputum production. Acute exacerbations (AE) of COPD (AE-COPD) are key determinants of disease progression; yet, existing predictive models relying mainly on spirometric measurements, such as forced expiratory volume in 1 second, reflect only a fraction of the physiological information embedded in respiratory function tests. Recent advances in artificial intelligence (AI) have enabled more sophisticated analyses of full spirometric curves, including flow-volume loops and volume-time curves, facilitating the identification of complex patterns associated with increased exacerbation risk. Objective This study aimed to determine whether a predictive model that integrates clinical data and spirometry images with the use of AI improves accuracy in predicting moderate-to-severe and severe AE-COPD events compared to a clinical-only model. Methods A retrospective cohort study was conducted using COPD registry data from 2 teaching hospitals from January 2004 to December 2020. The study included a total of 10,492 COPD cases, divided into a development cohort (6870 cases) and an external validation cohort (3622 cases). The AI-enhanced model (AI-PFT-Clin) used a combination of clinical variables (eg, history of AE-COPD, dyspnea, and inhaled treatments) and spirometry image data (flow-volume loop and volume-time curves). In contrast, the Clin model used only clinical variables. The primary outcomes were moderate-to-severe and severe AE-COPD events within a year of spirometry. Results In the external validation cohort, the AI-PFT-Clin model outperformed the Clin model, showing an area under the receiver operating characteristic curve of 0.755 versus 0.730 (P<.05) for moderate-to-severe AE-COPD and 0.713 versus 0.675 (P<.05) for severe AE-COPD. The AI-PFT-Clin model demonstrated reliable predictive capability across subgroups, including younger patients and those without previous exacerbations. Higher AI-PFT-Clin scores correlated with elevated AE-COPD risk (adjusted hazard ratio for Q4 vs Q1: 4.21, P<.001), with sustained predictive stability over a 10-year follow-up period. Conclusions The AI-PFT-Clin model, by integrating clinical data with spirometry images, offers enhanced predictive accuracy for AE-COPD events compared to a clinical-only approach. This AI-based framework facilitates the early identification of high-risk individuals through the detection of physiological abnormalities not captured by conventional metrics. The model’s robust performance and long-term predictive stability suggest its potential utility in proactive COPD management and personalized intervention planning. These findings highlight the promise of incorporating advanced AI techniques into routine COPD management, particularly in populations traditionally seen as lower risk, supporting improved management of COPD through tailored patient care.

Thrombolysis in myocardial infarction frame count enables assessment of coronary flow but cannot measure coronary flow velocity (CFV), which is needed to examine microvascular function. To overcome this limitation, we introduce a semi-automated software for fast CFV computation using contrast bolus tracking techniques in angiography and compare its performance against experts. The study included patients undergoing coronary angiography. Two experts measured the CFV using the number of frames, segment length, and frame rate. Measurements were repeated for shorter segments and different projections, and their estimations were compared with the software. In total, 123 patients (152 vessels) were included. The software had excellent reproducibility in measuring CFV (intraclass correlation coefficient (ICC) = .995), which was superior to experts (ICC = .946) and provided similar estimations irrespective of the segment length (ICC = .992); conversely, the experts overestimated CFV in short segments. The reproducibility of the experts and the software was moderate when comparing CFV measurements in different projections (first expert vs software ICC = .807, second expert vs software ICC = .790, first expert vs second expert ICC = .885). The software provides reproducible CFV estimations that are close to experts’ estimations. Further validation against wire-based functional techniques is needed to examine its potential in assessing microvascular function.

This paper introduces a voice-interactive chatbot designed to provide personalized health recommendations through a knowledge base built on validated medical literature. By leveraging 17 key health variables, the chatbot addresses approximately 150 million potential health scenarios, offering advice for personal health management. The system developed as an Android application, represents the initial phase of a broader project aimed at delivering evidence-based health management services. Future advancements will focus on replacing the current knowledge base with a knowledge graph and integrating large language models (LLMs) to enable more natural and contextual conversations in healthcare.

Large Language Models (LLMs) offer potential for automating systematic reviews, a labor-intensive process in evidence-based medicine. We evaluated GPT-4o, GPT-4o-mini, and Llama 3.1:8B on abstract screening and risk of bias assessment using 12 Cochrane drug intervention reviews. GPT-4o achieved the best screening performance (recall 0.894, precision 0.492). We propose a one-shot inclusivity adjustment method enabling threshold modulation without repeated inferences. For risk of bias, accuracy varied by domain, highest in random sequence generation (0.873), and lowest in selective reporting (0.418). Our findings demonstrate LLMs’ practical utility and current limitations in automating systematic reviews.

Given its vicious cycle of victimization, early life adversity (ELA) in childhood may be associated with peer victimization during adolescence. Both ELA and peer victimization have been suggested to be major risk factors for depression. Volumetric alterations in the hippocampus implicated in stress sensitivity have been reported in individuals with ELA and peer victimization. This cross-sectional study examined the moderating role of hippocampal volume in the association between ELA and peer victimization in adolescents with major depressive disorder (MDD). The sample included 78 adolescents with MDD (age M (SD) = 14.92 (1.54) years, 56 females). The Early Trauma Inventory-Short Form and Peer Victimization Scale were used to assess participants’ ELA and peer victimization, respectively. High-resolution structural T1 images were obtained using a Siemens 3T magnetic resonance scanner. Whole hippocampal volumes were segmented and calculated using the FreeSurfer 6.0. Correlation and moderation analyses were also performed. Emotional abuse, a type of ELA, was significantly correlated with peer victimization after controlling for age and sex. The association between emotional abuse and peer victimization was moderated by bilateral hippocampal volume in adolescents with MDD after controlling for age, sex, and intracranial volume. Additionally, the association between emotional abuse and peer victimization was stronger when the bilateral hippocampal volumes were larger. Our findings partially supported the concept of a vicious cycle of victimization, which may be a critical aspect of depression in adolescents. Furthermore, the moderation results suggested that hippocampal volume plays an important role in the victimization cycle in adolescents with MDD.

Importance Amiloride has been proposed as an alternative to spironolactone for treating resistant hypertension. However, no randomized clinical trials have compared the efficacy of spironolactone and amiloride in patients with resistant hypertension. Objective To determine whether amiloride is noninferior to spironolactone in reducing home-measured systolic blood pressure (SBP) in patients with resistant hypertension. Design, Setting, and Participants Prospective, open-label, blinded end-point randomized clinical trial conducted at 14 sites in South Korea. From November 16, 2020, to February 29, 2024, 118 patients with home SBP of 130 mm Hg or greater after a 4-week run-in period with a fixed-dose triple medication combination (angiotensin receptor blocker, calcium channel blocker, and thiazide) were enrolled. Intervention Patients were randomized in a 1:1 ratio to receive 12.5 mg/d of spironolactone (n = 60) or 5 mg/d of amiloride (n = 58). If home SBP remained 130 mm Hg or greater and serum potassium was less than 5.0 mmol/L after 4 weeks, dosages were increased to 25 mg/d and 10 mg/d, respectively. Main Outcomes and Measures The primary end point was the between-group difference in home SBP change at week 12, with a noninferiority margin of −4.4 mm Hg for the lower bound of the confidence interval. Secondary end points included achievement rates of home- and office-measured SBP of less than 130 mm Hg. Results The median age of the study population was 55 years, with 70% male. There were no differences between groups in demographic characteristics other than use of α-blockers (8.6% in the amiloride group and 0% in the spironolactone group). The mean baseline home SBPs were 141.5 (SD, 7.9) mm Hg and 142.3 (SD, 8.5) mm Hg in the amiloride and spironolactone groups, respectively. At week 12, mean home SBP measurements were changed from baseline by −13.6 (SD, 8.6) mm Hg and −14.7 (SD, 11.0) mm Hg in the amiloride and spironolactone groups, respectively (between-group difference in change, −0.68 mm Hg; 90% CI, −3.50 to 2.14 mm Hg), with amiloride demonstrating noninferiority to spironolactone. Home-measured achievement rates of SBP less than 130 mm Hg in the amiloride and spironolactone groups were 66.1% and 55.2%, respectively, and office-measured achievement rates of SBP less than 130 mm Hg were 57.1% and 60.3%, respectively, with no difference between the 2 groups. One case of hyperkalemia-related discontinuation occurred in the amiloride group, with no cases of gynecomastia in either group. Conclusions and Relevance Amiloride was noninferior to spironolactone in lowering home SBP, suggesting that it could be an effective alternative for treatment of resistant hypertension. Trial Registration ClinicalTrials.gov Identifier: NCT04331691

Exhaled breath condensate (EBC) is a respiratory biofluid generated from airway lining fluid. EBC contains various biological components, such as nucleotides, proteins, lipids, and others derived from cells in the lung. EBC collection is simple and noninvasive, making it an attractive source for pulmonary disease diagnosis, prognosis, and therapeutic monitoring. However, molecular analysis of EBC is technically challenging due to low concentrations of biomolecules in EBC and inconsistent sampling environments. This underscores the need to standardize the sample collection methods and develop novel analysis techniques to explore EBC for interstitial lung diseases (ILD). Here, we provide an overview of EBC, including their biological characteristics, collection methods, and analysis techniques. We also discuss the potential clinical applications of EBC for ILD and outline future technical developments.

Objective This study aimed to compare the lipid‐lowering effect and safety of low‐intensity atorvastatin (5 mg) plus ezetimibe (10 mg) combination therapy (A5E10) with monotherapy regimens–atorvastatin 5 mg [A5], ezetimibe 10 mg [E10], and atorvastatin 10 mg [A10])–in dyslipidemia patients. Methods A randomized, double‐blind, placebo‐controlled trial involving 252 dyslipidemia patients was conducted at 25 centers in South Korea (NCT05970679). Participants aged ≥ 19 years were randomized into four groups: A5E10, A5, E10, and A10. The primary endpoint was the percentage change in low‐density lipoprotein cholesterol (LDL‐C) levels from baseline to 8 weeks. Secondary endpoints included changes in other lipid parameters, lipid ratios, LDL‐C goal achievement rates and safety assessments. Results The mean age of the patients was 63 years, and 51.2% were male. The A5E10 group showed significantly greater LDL‐C reduction (47.6%) compared with A5 (33.4%), E10 (19.4%), and A10 (40.1%) at 8 weeks (p < 0.0001). A5E10 also significantly reduced triglyceride, non‐high‐density lipoprotein cholesterol, and apolipoprotein B levels. In addition, a significant reduction in LDL‐C levels was observed over the 4 weeks, with a 46.7% reduction in LDL‐C levels after 4 weeks of A5E10 administration. No severe adverse events were observed in the A5E10 group. Conclusion The combination of low‐intensity atorvastatin and ezetimibe was more effective than moderate‐intensity atorvastatin monotherapy in lowering LDL‐C levels and improving other lipid parameters. It was well‐tolerated and demonstrated rapid benefits within a month, offering a promising alternative for patients with low to moderate cardiovascular risk who do not achieve adequate control with statin monotherapy.

Inherited bone marrow failure syndromes are genetic hematologic disorders with increased cancer risk. Accurate diagnosis is crucial for appropriate management. This study assessed the clinical usefulness of next-generation sequencing (NGS)-based target gene sequencing in pediatric and AYA (adolescent and young adult) patients with hematologic abnormalities. From December 2019 to June 2023, 93 patients with suspected congenital hematologic diseases at a single institution underwent NGS-based testing. Medical records were retrospectively reviewed. The median age at diagnosis was 9.3 years (range 0.2–31.4), with 59.1% males. Indications for testing included specific medical histories (28 patients), persistent cytopenia or recurrent neutropenic fever (22 patients), changes in cytopenia patterns (11 patients), and other reasons (32 patients). Pathogenic variants were identified in 9/28 (32.1%), 3/22 (13.6%), 4/11 (36.4%), and 0/32 (0%). Overall, 16 patients (17.2%) had pathogenic variants, including FANCA, BRCA2, PMS2, ELANE, G6PC3 and VPS13B in patients with idiopathic neutropenia, and GATA2 in patients with suspected myelodysplastic syndrome. Genetic findings led to diagnostic revisions in 12 patients (12.9%), including reclassification of aplastic anemia (AA) as Fanconi anemia, Diamond-Blackfan anemia, or Shwachman-Diamond syndrome, prompting hematopoietic stem cell transplantation and altering cancer surveillance. Pathogenic variants were more frequently observed in patients with a specific medical history or changes in cytopenia, and in those with additional clinical features (cytogenetic abnormalities or non-severe AA). This study demonstrated the diagnostic usefulness of NGS-based target gene sequencing for pediatric and AYA patients with suspected genetic hematologic disorders, supporting the need for multicenter studies and standardized guideline development.

Deep learning–based chest CT–derived metrics were comparable to forced vital capacity for predicting survival in patients with amyotrophic lateral sclerosis and were particularly valuable for those with bulbar impairment.

This perspective proposes adapting video-text generative AI to 3D medical imaging (CT/MRI) and medical videos (endoscopy/laparoscopy) by treating 3D images as videos. The approach leverages modern video models to analyze multiple sequences simultaneously and provide real-time AI assistance during procedures. The paper examines medical imaging’s unique characteristics (synergistic information, metadata, and world model), outlines applications in automated reporting, case retrieval, and education, and addresses challenges of limited datasets, benchmarks, and specialized training.

Background This study aimed to evaluate the quality of care in newly diagnosed RA patients by analyzing conventional disease-modifying antirheumatic drugs (cDMARDs) treatment patterns and healthcare utilization using a nationwide claims database. Methods This retrospective cohort study was conducted using the Korean Health Insurance Review and Assessment database. Study subjects were those newly diagnosed with RA (ICD-10 code M05, M06) and were prescribed a cDMARD in 2014, with follow-up until 2018. Demographic and clinical information, the level of healthcare facilities (LOH) at which the first prescription claim was made, and subsequent healthcare service utilization were collected. We also analyzed the initial pattern in cDMARD prescription and its retention rate. Results A total of 21,136 patients were analyzed. Diagnosis of seronegative RA (n = 14,571, 68.9%) was more common than seropositive RA. Seropositive RA was most often discovered in tertiary general hospitals (n = 2,230, 34.0%), whereas seronegative RA was most diagnosed in primary care clinics (n = 7,539, 51.7%) (P < 0.001). The most prescribed initial cDMARD was hydroxychloroquine as monotherapy (n = 9,867, 46.7%). However, methotrexate, a well-established first-line cDMARD, was prescribed in 5,447 (25.8%) patients. The discontinuation rate of cDMARD was higher in seronegative than seropositive patients (65.3% vs. 90.3%) and in patients first diagnosed in community LOH (P for trend < 0.001). The mean number of visits to any outpatient clinics (35/year) was substantially higher than that of the general population. Yet, the number of outpatient visits for RA management was only 2.8/year. Conclusion The quality of care for newly diagnosed RA patients in South Korea can be improved. Further education on accurate diagnosis and effective treatment is necessary to improve the quality of care provided by other specialists and general practitioners.