A survey paper published in Medical Image Analysis

A survey paper led by Fengming Lin, published in Medical Image Analysis, reviews deep learning approaches for medical image-to-mesh reconstruction of three-dimensional anatomical models. The paper introduces a structured taxonomy of methods, spanning template, statistical shape, generative, and implicit models, and evaluates their strengths, limitations, and applications across diverse anatomical domains.

Beyond reconstruction techniques, the work also provides an extensive analysis of datasets, loss functions, and evaluation metrics, while highlighting key challenges such as topological correctness, geometric accuracy, and multi-modal integration. The paper serves as an important reference for researchers developing digital twins, computational simulations, and in silico trials, and outlines future directions for advancing patient-specific computational medicine.