Related Experiment Video
Updated: Jul 8, 2025
05:33
Three-Dimensional Shape Modeling and Analysis of Brain Structures
Published on: November 14, 2019
7.1K
Construction of Shape Atlas for Abdominal Organs using Three-Dimensional Mesh Variational Autoencoder
Summary
This study introduces a mesh variational autoencoder for accurate organ shape and position modeling. The model achieves high accuracy in reconstructing complex anatomical variations, benefiting applications like surgical guidance.
Area of Science:
- Medical imaging
- Computational anatomy
- Machine learning
Background:
- Accurate organ shape and position modeling is crucial for clinical applications like radiotherapy and surgical guidance.
- Existing linear models struggle with soft organ variability, while nonlinear models risk overfitting with limited data.
Purpose of the Study:
- To develop a highly accurate and generalizable shape atlas for organ modeling.
- To address the limitations of current models in reconstructing complex anatomical variations.
Main Methods:
- Designed a mesh variational autoencoder (MVAE) capable of reconstructing nonlinear organ shape and position.
- Trained and validated the MVAE model using liver meshes from 125 patient cases.
Main Results:
- The MVAE model demonstrated high accuracy in reconstructing complex, nonlinear organ shapes and positions.
- Validation on 19 test cases showed an average reconstruction accuracy of 4.3 mm for liver meshes.
Conclusions:
- The developed MVAE provides a robust method for creating accurate and generalizable organ shape atlases.
- This approach holds significant potential for improving precision in image-guided interventions and radiotherapy planning.
