Related Experiment Video
Updated: Jul 6, 2025
07:53
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
1.5K
Real-time estimation of lung deformation from body surface using a general CoordConv CNN
Mingkang Liu1, Yongtai Zhuo1, Jie Liu2
1School of Biomedical Engineering and Institute of Medical Robotics, Shanghai Jiao Tong University, Shanghai, China.
Computer Methods and Programs in Biomedicine
|January 4, 2024
Summary
This study introduces a new method to estimate lung deformation using only two CT phases, reducing radiation exposure. The technique achieves high accuracy in real-time, improving radiotherapy and surgical navigation.
Area of Science:
- Medical imaging
- Computational anatomy
- Radiotherapy physics
Background:
- Accurate estimation of 3D lung deformation is crucial for radiotherapy and surgical navigation.
- Current methods often require 4D-CT, increasing radiation dose.
- A more efficient method for lung deformation estimation is needed.
Purpose of the Study:
- To develop a novel method for estimating lung tissue deformation using depth maps and two CT phases.
- To reduce the radiation dose associated with lung deformation estimation.
- To improve the accuracy and efficiency of lung deformation analysis for clinical applications.
Main Methods:
- A 3D motion model was developed, representing voxel movement as linear displacement.
- Direction vectors and amplitudes were derived from end-of-exhale (EOE) and end-of-inhale (EOI) CT phase registration.
- A neural network estimated voxel phase, utilizing Coordinate Convolution (CoordConv) for multimodal data fusion and absolute position embedding.
Main Results:
- The method achieved average errors of 2.11 mm on the DIR-Lab dataset and 1.36 mm on the 4D-Lung dataset.
- Real-time performance was demonstrated, with processing times under 7 ms per frame on a high-end GPU.
- The approach successfully integrated depth map and CT data for deformation estimation.
Conclusions:
- The proposed method offers comparable or superior accuracy to existing techniques.
- It achieves this accuracy using fewer CT phases, thereby reducing radiation exposure.
- The real-time capability makes it suitable for integration into clinical workflows for radiotherapy and surgery.
