Ji Soo Park’s research while affiliated with Severance Hospital and other places

Background Pediatric respiratory diseases, including asthma and pneumonia, are major causes of morbidity and mortality in children. Auscultation of lung sounds is a key diagnostic tool but is prone to subjective variability. The integration of artificial intelligence (AI) and machine learning (ML) with electronic stethoscopes offers a promising approach for automated and objective lung sound. Objective This systematic review and meta-analysis assess the performance of ML models in pediatric lung sound analysis. The study evaluates the methodologies, model performance, and database characteristics while identifying limitations and future directions for clinical implementation. Methods A systematic search was conducted in Medline via PubMed, Embase, Web of Science, OVID, and IEEE Xplore for studies published between January 1, 1990, and December 16, 2024. Inclusion criteria are as follows: studies developing ML models for pediatric lung sound classification with a defined database, physician-labeled reference standard, and reported performance metrics. Exclusion criteria are as follows: studies focusing on adults, cardiac auscultation, validation of existing models, or lacking performance metrics. Risk of bias was assessed using a modified Quality Assessment of Diagnostic Accuracy Studies (version 2) framework. Data were extracted on study design, dataset, ML methods, feature extraction, and classification tasks. Bivariate meta-analysis was performed for binary classification tasks, including wheezing and abnormal lung sound detection. Results A total of 41 studies met the inclusion criteria. The most common classification task was binary detection of abnormal lung sounds, particularly wheezing. Pooled sensitivity and specificity for wheeze detection were 0.902 (95% CI 0.726-0.970) and 0.955 (95% CI 0.762-0.993), respectively. For abnormal lung sound detection, pooled sensitivity was 0.907 (95% CI 0.816-0.956) and specificity 0.877 (95% CI 0.813-0.921). The most frequently used feature extraction methods were Mel-spectrogram, Mel-frequency cepstral coefficients, and short-time Fourier transform. Convolutional neural networks were the predominant ML model, often combined with recurrent neural networks or residual network architectures. However, high heterogeneity in dataset size, annotation methods, and evaluation criteria were observed. Most studies relied on small, single-center datasets, limiting generalizability. Conclusions ML models show high accuracy in pediatric lung sound analysis, but face limitations due to dataset heterogeneity, lack of standard guidelines, and limited external validation. Future research should focus on standardized protocols and the development of large-scale, multicenter datasets to improve model robustness and clinical implementation.

Background : Triple-negative breast cancer (TNBC) is a frequent phenotype of BRCA-mutant tumors. Tumors with BRCAness may show characteristics of BRCA-mutant tumors and respond to similar treatments. Next-generation sequencing (NGS) is an efficient and cost-effective method for simultaneously sequencing multiple cancer susceptibility genes, surpassing the conventional Sanger testing. Methods : A total of 148 women with TNBC were recruited from December 2015 to November 2018 at Yonsei Cancer Center. Of them, 103 patients received neoadjuvant chemotherapy (NCT). The targeted genes related to hereditary cancers were sequenced using the 65-gene germline NGS (gNGS) panel pathogenic and likely pathogenic variants (P&LPs) were determined by Sanger sequencing. We examined the occurrence of pathologic complete remission (ypCR) in patients with P&LPs. Results : The patients’ median age was 47 years (range, 27–69 years). Twenty (13.7%) of 148 patients had P&LP in six genes, including BARD1 (n=2), BRCA1 (n=9), BRCA2 (n=5), CHEK2 (n=1), RAD51C (n=1), and RAD51D (n=2). Among the 103 patients with NCT, 43 (28.9%) achieved ypCR (P&LPs; 9 individuals vs. non-variants; 34 individuals). Among the 103 patients with NCT, 14 (9.3%) had P&LPs. Nine of 14 patients with P&LPs, including BARD1 (n=2), BRCA1 (n=4), BRCA2 (n=1), RAD51 C (n=1), and RAD51D (n=1), achieved ypCR ( P = .066). Conclusion : Germline P&LP mutations in TNBC patients can be detected by gNGS. This panel test can identify BRCA and BRCAness mutations that may predict ypCR in TNBC.

Purpose As both the number of long-term cancer survivors and cases of late cancer recurrence increased, understanding aspects of lifestyle, such as participation in physical activity (PA), is of importance among long-term cancer survivors. This study aimed to investigate PA levels among long-term cancer survivors compared with age-, sex-, and body mass index (BMI)-matched non-cancer participants. Methods The study analyzed data from 648 cancer survivors (66% female, average age of 60.0 ± 10.9 years) who visited the Cancer Prevention Center, Yonsei Cancer Center, between 2015 and 2022. A total of 3240 participants were randomly selected from the Korea National Health and Nutrition Examination Survey (2000–2020) using propensity score matching based on age, sex, and BMI. Results Among long-term cancer survivors, the majority of participants were survivors of colorectal (43.1%), gastric (23.8%), and breast cancer (24.1%). Long-term cancer survivors participated more in moderate-to-vigorous leisure-time PA (106.2 ± 244.3 vs. 47.3 ± 135.8 min/week, P < 0.05), transportation PA (130.3 ± 231.8 vs. 101.2 ± 174.9 min/week, P < 0.05), and total walking time (311.4 ± 380.1 vs. 227.6 ± 347.7 min/week, P < 0.05) compared with matched non-cancer participants. However, long-term cancer survivors were less likely to participate in less resistance exercises (0.7 ± 1.8 vs. 1.5 ± 1.5 day/week, P < 0.05) compared with the non-cancer matched participants. Although long-term cancer survivors are generally more physically active, a significant proportion of long-term cancer survivors (76.1%) still did not meet the PA guidelines for moderate-intensity aerobic exercise. Conclusion Although long-term cancer survivors participated in more PA compared to the matched non-cancer participants, the majority of long-term cancer survivors still do not meet the PA guidelines, suggesting that further strategies to increase PA are required.

Prenatal environmental exposures may influence the development of atopic dermatitis (AD), yet the underlying mechanisms remain poorly understood. In this study, we conducted an epidemiological analysis of 986 children, revealing that first-trimester PM 2.5 exposure significantly increases AD risk at age three. Integrated epigenomic and transcriptomic profiling of 48 placentae identified PM 2.5 -associated hypomethylation driving FCER1G overexpression, predominantly in fetal macrophages. This "epigenetic scar" persists across developmental stages, as demonstrated through single-cell RNA sequencing of 404,169 cells from placenta, fetal, and adult skin samples. In adult AD skin, FCER1G was progressively overexpressed in M2 macrophages, particularly within AD-specific cell states. Network analysis and in vitro PM exposure experiments identified a mechanistic axis involving FCER1G, CD68, CYBB, MRC1, and MS4A4A, promoting NADPH oxidase–driven reactive oxygen species signaling. Our findings establish a mechanistic link between prenatal PM 2.5 exposure and AD pathogenesis through persistent macrophage FCER1G dysregulation, highlighting placental epigenetic scars as critical mediators of immune pathogenesis and a target for intervention.

Purpose Studies on the longitudinal clinical features of asthma or allergic comorbidities in children are limited. We aimed to examine the trajectories of asthma and allergic comorbidities and determine whether these trajectories differ according to clinical asthma phenotypes from birth to adolescence. Methods We enrolled 958 children with physician-diagnosed asthma from the Korean childhood Asthma Study (KAS) cohort. Children with asthma were classified using hierarchical cluster analysis. Information on the diagnosis and treatment of allergic diseases before cohort entry was collected through linkage with national claims data from the Health Insurance Review and Assessment Service. Results In the KAS cohort, approximately half had a history of atopic dermatitis (AD) before infancy, with its prevalence gradually decreasing during adolescence. The prevalence of allergic rhinitis (AR) increased with age. The prevalence of asthma increased during early childhood and decreased during adolescence. According to the natural progression of asthma, AD, and AR trajectories, 4 distinctive phenotypes were identified using latent class analysis: “almost controlled,” “early-onset asthma with AD and late-onset AR,” “early-onset asthma only,” and “intermediate-onset asthma and late-onset AR.” Four distinct clinical trajectory patterns of asthma, AD, and AR were identified among the 4 cluster phenotypes based on baseline characteristics. Cluster 1 comprised male-dominant, atopic asthma with early-onset AD and late-onset AR. Cluster 2 included early-onset, atopic asthma with AD” persistent into adolescence. Cluster 3 encompassed “puberty-onset, female-dominant atopic asthma” with early-onset and low remission rates. Cluster 4 comprised “early-onset asthma with less atopic features” and the lowest comorbidities of AD and AR. Conclusions The longitudinal trajectories of asthma and allergic comorbidities in Korean children can be classified into distinct clusters. Most phenotypes exhibited early-onset asthma with a varying prevalence of comorbidities. The persistence of AD, rather than its onset age, determines the phenotype.

Purpose The partner and localizer of breast cancer 2 (PALB2) mutation is a predisposition to breast cancer development. However, limited clinical data are available for the Korean population. Therefore, this study aimed to compare the characteristics and oncological outcomes of patients with PALB2-mutated and non-mutated PALB2 in Korea. Methods A total of 1,463 breast cancer (BRCA) 1/2 mutation-negative breast cancer underwent comprehensive multigene sequencing between 2016 and 2019 at Severance Hospital, Seoul, Korea. Clinicopathological data and oncological results of PALB2 mutated patients were prospectively collected and compared with those of the non-mutated group. Results PALB2 mutations were identified in 1.2% (17/1,463) of the patients. The median age at diagnosis was 46 (30–73) years, and the median tumor size was 1.8 (0.05–3.5) cm. All patients with PALB2 mutations had histologic grades II–III, and a triple-negative subtype was found in 23.5% (4/17); however, there were no significant differences in clinicopathological data between the groups. During the median follow-up time of 38.5 months, locoregional recurrence occurred in 4.2% (44/1,043), distant recurrence was reported in 3.0% (31/1,043), and contralateral breast cancer was diagnosed in 0.8% (9/1,043) of patients, with no significant difference observed between the groups. All-cause mortality was observed in 1.8% (19/1,028) of the non-mutated group and none in the PALB2 mutation group. However, survival analyses demonstrated no significant differences in all-cause mortality (p = 0.524) and recurrence-free survival (p = 0.319). Conclusion Clinicopathological features and oncological outcomes of PALB2 mutated breast cancer were not significantly different from those of non-mutated breast cancer in the Korean population.

Background Whether the progression of precursor lesions or the occurrence of cancer is influenced by lifestyle factors in carriers of genetic mutations has not been fully investigated, especially in Asian patients of hereditary colorectal cancer (CRC) syndrome. Methods Patients at a high risk of hereditary CRC were included. For polyposis CRC syndromes, colorectal polyp burden was measured using at least 60 images per colonoscopy in each patient and classified into five stages using the International Society for Gastrointestinal Hereditary Tumours staging system according to the polyp number and size. Increase in tumor burden stage for polyposis CRC syndrome and the occurrence of CRC or any cancer for Lynch syndrome were analyzed according to lifestyle factors. Results Ninety‐six patients with suspected hereditary polyposis CRC syndrome and 106 patients with Lynch syndrome were recruited. For polyposis CRC syndromes, multivariate analysis showed that exposure to smoking and > 100 polyps independently predicted a high risk of increased polyp burden ( p = 0.008 and p = 0.012, respectively). Significant genetic mutations or phenotype of polyposis syndromes were significantly associated with an increased polyp burden. For Lynch syndrome, smokers showed to be diagnosed with CRC in younger age than never‐smokers (42.2 years vs. 49.0 years; p = 0.021), and heavy drinkers had high risk for occurrence of CRC (HR, 2.381, 95% CI, 1.338–4.236; p = 0.003) and any cancer (HR, 2.254; 95% CI, 1.334–3.806; p = 0.002). Conclusions The lifestyle factors (smoking and alcohol consumption) were associated with increasing precursor lesions and occurrence of cancer in patients with hereditary CRC syndrome. Lifestyle modifications may reduce the risk of hereditary CRC in carriers.