Author

S. Youn

View on Semantic Scholar

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Open access Aug 2026

Benchmarking Open-Source Vision-Language Models for Brain Metastasis Assessment on Single-Slice Contrast-Enhanced MRI

Purpose Open-source vision-language models (VLMs) can be locally deployed without external internet access, potentially enhancing data security. This study compared the diagnostic performance of general-purpose and medical-purpose open-source VLMs and evaluated their ability to characterize brain metastases on contrast-enhanced (CE) MRI. Materials and Methods Sixty lesion-positive axial CE T1-weighted images and sixty matched lesion-negative images from 60 patients were analyzed using three general-purpose VLMs-InternVL3-8B, Qwen2.5-VL-7B-Instruct, and MiniCPM-V-4.5-and three medical-purpose VLMs-MedGemma-4B-it, LLaVA-Med v1.5, and HuatuoGPT-Vision-7B. Lesion detection performance was assessed using sensitivity, specificity, and balanced accuracy. On lesion-positive images, accuracy was evaluated for lesion count, laterality, anatomic location, enhancement pattern, necrosis, vasogenic edema, and mass effect. Model differences were assessed using Cochran's Q tests followed by pairwise McNemar tests with Benjamini-Hochberg correction. Results The median age of the study patients was 67 years (IQR, 61.0-70.5 years), and 35 patients were male (58.3%). MiniCPM-V-4.5 showed the most balanced diagnostic performance, with a sensitivity of 78.3% (95% CI, 66.4-86.9%) and a specificity of 85.0% (95% CI, 73.9-91.9%), and significantly higher balanced accuracy than all other models. Significant overall differences were observed for lesion count, laterality, location, enhancement pattern, necrosis, and mass effect, but not for vasogenic edema (FDR-adjusted P = 0.056). HuatuoGPT-Vision-7B and MedGemma-4B-it showed relatively consistent accuracy across multiple image assessment tasks, although their performance remained modest. Conclusion Our study demonstrated substantial heterogeneity in the performance of open-source VLMs in brain metastasis evaluation, and medical-purpose VLMs did not outperform general-purpose VLMs.

J. Kim, B.-S. Kim, J.-S. Ko et al. · medRxiv · 0 citations

Open access Jul 2026

MLHeatmap: an interactive application for transcriptomic marker-panel discovery

Abstract Motivation Transcriptomic biomarker studies often require separate tools for normalization, classification, feature ranking, and visualization. MLHeatmap was developed to integrate these steps into a cross-platform, browser-based workflow and to support compact marker-panel discovery directly from count matrices. Results MLHeatmap accepts a count matrix and sample-group labels and performs gene mapping, normalization, multiclass classification with nested cross-validation, feature attribution, differential expression analysis, and interactive heatmap visualization, with support for multiple classifiers and panel-selection methods in the same workflow. As a case study, we applied MLHeatmap to colorectal cancer consensus molecular subtype (CMS) classification using 511 primary tumors from The Cancer Genome Atlas (TCGA) with published Colorectal Cancer Subtyping Consortium (CRCSC) labels. Random Forest with forward selection achieved 89.8% out-of-fold accuracy and a macro-averaged area under the receiver operating characteristic curve (macro AUC) of 0.973, yielding a 13-gene compact panel with a held-out AUC of 0.960. External validation of the compact panel gave 74.8% accuracy and a macro AUC of 0.917 in GSE39582, an independent Affymetrix GPL570 microarray cohort evaluated in 500 tumors with confident CMScaller-derived CMS assignments, and 69.18% accuracy and a macro AUC of 0.909 in CMCBSN, an independent Korean RNA-seq cohort with 159 confidently labeled tumors. Across the five classifiers and four panel-selection methods evaluated, compact-panel AUCs ranged from 0.895 to 0.970, with tree-ensemble classifiers reaching the highest values. Availability and implementation MLHeatmap is freely available at https://github.com/kangk1204/MLHeatmap and can be installed and run locally on Windows 11, macOS, and Ubuntu.

Eun Young Lee, Jihye Park, S. Youn et al. · Bioinformatics Advances · 0 citations