Subject-Specific Musculoskeletal Modeling Based on Medical Images for the Application in Biomechanical Analysis

Abstract

This paper summarizes the achievements in the group A01-3 in Multidisciplinary Computational Anatomy project. We aimed to develop a subject-specific musculoskeletal model for the application in health sciences including medicine, biomechanics, and sports science. We developed multiple methods for generating a musculoskeletal model that reflects subject-specific anatomy using medical images. In this paper, we describe the four major achievements: 1) musculoskeletal segmentation using deep learning, 2) CT-MRI cross modality image synthesis, 3) 2D-3D registration of x-ray projection image and CT image for the analysis of skeletal movement, 4) estimation of muscle fiber arrangement from a clinical CT using a template derived from the high resolution cryo-section images.