Related Experiment Video
Updated: Jul 4, 2025
10:26
Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
1.9K
Evaluation of an artificial intelligence U-net algorithm for pulmonary nodule tracking on chest computed tomography
Yuhei Takeshita1, Shiro Onozawa1, Shichiro Katase1
1Department of Radiology, Kyorin University School of Medicine, 6-20-2 Shimorenjaku, Mitaka-shi, Tokyo, Japan 181-8611 Japan.
The Journal of International Medical Research
|February 7, 2024
Summary
Artificial intelligence (AI) demonstrates high feasibility for automatic lung nodule tracking in computed tomography (CT) follow-up. This AI method accurately tracks pulmonary nodules over time, aiding in patient monitoring.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in radiology
- Pulmonary nodule detection
Background:
- Accurate follow-up of lung nodules is crucial for patient management.
- Manual tracking of lung nodules can be time-consuming and subjective.
- Advancements in artificial intelligence offer potential for automated solutions.
Purpose of the Study:
- To implement image registration for lung nodule follow-up.
- To assess the feasibility of automatic lung nodule tracking using an AI method.
- To evaluate the performance of an AI algorithm in predicting nodule position over time.
Main Methods:
- Retrospective observational study of patients with pulmonary nodules (5-30 mm diameter) on CT.
- AI algorithm utilizing a U-net neural network to predict deformation vector fields between scans.
- Radiologist-defined cuboids used to assess the success of AI-driven nodule tracking.
Main Results:
- Analysis of 49 lung nodules across 368 follow-up CT examinations.
- AI achieved a 94% success rate for time evaluations and 78% for nodule-by-nodule evaluations.
- Lower success rates were observed for smaller nodules or those decreasing in size.
Conclusions:
- Automatic tracking of lung nodules using AI is highly feasible.
- The AI method shows promise for efficient and accurate lung nodule follow-up.
- Further refinement may improve tracking for nodules with size variations.
