Related Experiment Video
Updated: Nov 26, 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
1.8K
Deep convolutional neural networks for multiplanar lung nodule detection: Improvement in small nodule identification
Sunyi Zheng1, Ludo J Cornelissen1, Xiaonan Cui2
1Department of Radiation Oncology, University Medical Center Groningen, University of Groningen, 9713 AV, Groningen, The Netherlands.
Medical Physics
|December 10, 2020
Summary
This study introduces a deep learning framework for accurate lung nodule detection using multiple imaging planes. The multiplanar approach improves detection rates, especially for small nodules, enhancing early lung cancer diagnosis.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Deep Learning for Cancer Detection
- Radiology and Diagnostic Imaging
Background:
- Early lung cancer detection significantly improves patient survival rates.
- Current screening methods often rely on single-plane analysis, potentially missing subtle nodules.
- Integrating multiple imaging planes (axial, coronal, sagittal) mirrors clinical practice and can enhance nodule detection accuracy.
Purpose of the Study:
- To develop and evaluate an accurate deep learning framework for lung nodule detection.
- To leverage a multiplanar approach, combining axial, coronal, and sagittal views, for improved nodule identification.
- To enhance the Computer-Aided Detection (CAD) system's performance in identifying both small and large lung nodules.
Main Methods:
- A two-stage deep learning system was designed: multiplanar nodule candidate detection and multiscale false positive reduction.
- The first stage utilized a deeply supervised encoder-decoder network trained on axial, coronal, and sagittal slices.
- A 3D multiscale dense convolutional neural network was employed for refining results by reducing false positives, using the LIDC-IDRI dataset with tenfold cross-validation.
Main Results:
- The proposed system achieved high sensitivity, reaching 94.2% with 1.0 false positive (FP)/scan and 96.0% with 2.0 FPs/scan.
- It demonstrated effectiveness for small nodules (<6 mm), achieving 93.4% sensitivity at 1.0 FP/scan and 95.0% at 2.0 FPs/scan.
- The multiplanar method significantly outperformed single-plane detection in identifying more nodules.
Conclusions:
- The developed deep learning framework effectively detects lung nodules, including small ones and large lesions.
- The multiplanar approach is crucial for enhancing the accuracy of lung nodule detection systems.
- This CAD system shows significant promise for improving early lung cancer diagnosis.
