AI-powered chest X-ray pneumonia detection with 86% accuracy and 96.4% sensitivity, validated on an independent (cross-operator) cohort of 485 pediatric samples. Built with TensorFlow & FastAPI.

Interpretable TCGA-BRCA RNA-seq breast cancer subtype classification with ML, SHAP, pathway enrichment, and METABRIC external validation.

Biomedical abstract sentence classification workflow with calibration and external validation

Benchmark honesto de descriptores de imagen para clasificar lesiones cutaneas dermatoscopicas (ISIC): 390 combinaciones descriptor x algoritmo, control del efecto de lote y validacion externa entre instituciones. AUC 0.741 fuera del archivo de entrenamiento.

Reserved-hospital evaluation of domain shift, calibration, and operating-point transportability in histopathology.

Independent external validation, recalibration & fairness audit of EchoNext-Mini (AI-ECG for structural heart disease) in MIMIC-IV

Sports activity classification using wearable-device features, XGBoost, and external validation on unseen data.

Designed and validated a computational oncology pipeline for hepatocellular carcinoma using RNA expression data, integrating machine learning, survival analysis, SHAP explainability, tumor microenvironment deconvolution, and multi-cohort external validation for precision medicine applications.

Audit whether a medical-ML result measures biology or the lab that produced the data.

Interpretable machine-learning prediction and transportability assessment of late or persistent delirium in older ICU patients using MIMIC-IV and eICU-CRD.

Reproducibility code, processed data, out-of-fold predictions, and machine-readable results for leakage-aware e-commerce product classification using Mercadona and Amazon Reviews 2023.

Developed ML algorisms on transcriptomics landscapes identi-fies SPON1 and ALDH1A2 as candidate prognostic biomarkers in TME and serous ovarian cancer progression

근감소증 예측 멀티모달 ML - 체성분/악력/음성/터치 피처, CatBoost+앙상블, 다기관 외부검증 (부트스트랩 AUC CI, 합성 데이터로 재현)

Auditable CROSS-WELL protocol and reproducible field-level ROP validation evidence

Evidence-aware somatic variant pathogenicity classifier. Benchmarks a Cross-Attention fusion network over ESM-2 protein embeddings against LightGBM, with gene-disjoint external validation and a leakage-audit for Type-1 circularity.