Related Experiment Video
Updated: Oct 7, 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
1.6K
Two-stage multitask U-Net construction for pulmonary nodule segmentation and malignancy risk prediction.
Yangfan Ni1,2, Zhe Xie1,2, Dezhong Zheng1,2
1Laboratory for Medical Imaging Informatics, Shanghai Institute of Technical Physics, Chinese Academy of Science, Shanghai, China.
Quantitative Imaging in Medicine and Surgery
|January 7, 2022
Summary
This study introduces an accurate algorithm for segmenting pulmonary nodules and predicting malignancy risk. The novel coarse-to-fine 2.5D strategy enhances computer-aided diagnosis systems for better clinical decision-making.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Accurate pulmonary nodule segmentation is crucial for nodule analysis and malignancy risk prediction.
- Manual segmentation suffers from interobserver variability, necessitating robust automated methods.
- Developing an accurate segmentation and malignancy risk prediction algorithm is essential.
Purpose of the Study:
- To construct an accurate algorithm for pulmonary nodule segmentation.
- To develop a reliable method for predicting pulmonary nodule malignancy risk.
- To improve computer-aided diagnosis systems for lung nodule analysis.
Main Methods:
- A coarse-to-fine 2-stage framework utilizing two convolutional neural networks.
- A 3D multiscale U-Net for initial nodule localization.
- A 2.5D multiscale separable U-Net (MSU-Net) for segmentation refinement and multitask malignancy risk prediction.
Main Results:
- The proposed method achieved state-of-the-art results on the LIDC-IDRI dataset.
- Achieved a Dice Similarity Coefficient (DSC) of 83.04% for segmentation.
- Obtained 77.8% accuracy and 84.3% AUC for malignancy risk prediction, outperforming inter-radiologist agreement.
Conclusions:
- The multitask end-to-end framework demonstrated effectiveness.
- The coarse-to-fine 2.5D strategy improved accuracy and efficiency in nodule segmentation and risk prediction.
- The method provides clinicians with accurate quantitative information for treatment planning.
