* Conducted statistical analysis of heterogeneous clinical datasets across 37 projects, including preprocessing, hypothesis testing, survival analysis, and regression modeling, producing clinically interpretable results that supported multiple publications and improved reliability of medical conclusions
* Developed a deep learning pipeline (YOLO, U-Net) for vertebral detection and segmentation from X-ray images (
[email protected]: 0.90, DICE: 0.94), enabling automated 3D reconstruction and computation of sagittal balance parameters, and integrated the solution into clinical workflows via 3D Slicer
* Designed and curated annotated medical imaging datasets in collaboration with clinicians, resulting in a registered patent for machine learning applications
* Collaborated with physicians to translate clinical requirements into machine learning solutions, validate model assumptions, and ensure interpretability of results for clinical use
* Developing physics-informed neural networks (PINNs) for biomechanically consistent spine modeling, aiming to replace computationally expensive numerical simulations with data-driven approximations
* Applying PINN-based approaches to surgical planning and rehabilitation outcome prediction