Related Experiment Video
Updated: Dec 21, 2025
10:26
Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
2.3K
Deep Learning in CT Images: Automated Pulmonary Nodule Detection for Subsequent Management Using Convolutional Neural
Yi-Ming Xu1, Teng Zhang1, Hai Xu1
1Department of Radiology, The First Affiliated Hospital of Nanjing Medical University, Nanjing, People's Republic of China.
Cancer Management and Research
|May 20, 2020
Summary
A new 3D CNN-based computer-aided detection (CAD) model significantly improved pulmonary nodule detection on CT scans compared to radiologists. This AI tool accurately identifies nodules for better patient management.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Pulmonary nodules are common incidental findings on thin-section CT.
- Accurate detection and characterization of pulmonary nodules are crucial for timely diagnosis and management of lung cancer.
- Computer-aided detection (CAD) systems aim to assist radiologists in interpreting medical images.
Purpose of the Study:
- To compare the detection performance of a 3D convolutional neural network (3D CNN)-based CAD model against radiologists with varying experience levels.
- To evaluate the efficacy of a re-trained 3D CNN CAD model in detecting pulmonary nodules on thin-section CT.
Main Methods:
- Retrospective review of 1109 patients who underwent thin-section CT.
- A 3D CNN model for nodule detection was re-trained and enhanced with expert input.
- Detection performance was assessed using free-response receiver operating characteristic (FROC) analysis, comparing the CAD model with three radiologists.
Main Results:
- The re-trained 3D CNN CAD model demonstrated significantly higher sensitivity (93.09%) compared to the pre-trained network (38.44%) and radiologists (average 50.22%).
- The CAD model's performance did not significantly increase false positives per scan (1.64 vs 0.68).
- In the training set, 15 of 101 solid nodules were confirmed as lung cancer, with 922 nodules recommended for follow-up.
Conclusions:
- The expert-augmented, re-trained 3D CNN CAD model is an accurate and efficient tool for identifying incidental pulmonary nodules.
- This AI-powered approach can aid in the subsequent management of patients with pulmonary nodules.
- The study highlights the potential of advanced AI in improving diagnostic accuracy in radiology.
