Related Experiment Video
Updated: Jan 9, 2026
10:26
Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
2.5K
Flexible state space modelling for accurate and efficient 3D lung nodule detection.
Wenjia Song1, Fangfang Tang1, Henry Marshall2,3,4
1School of Electrical Engineering and Computer Science, The University of Queensland, Brisbane, Australia.
Biomedical Physics & Engineering Express
|December 9, 2025
Summary
A new AI model, FCMamba, improves early lung cancer detection in CT scans by enhancing context awareness and reducing processing steps. This flexible connected visual state-space model offers more accurate and efficient 3D lung nodule identification for clinical use.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Early and accurate detection of pulmonary nodules in computed tomography (CT) scans is crucial for reducing lung cancer mortality.
- Current deep learning models like CNNs and Transformers struggle with global context, high complexity, and reliance on post-processing.
- Existing methods often require manual tuning and post-processing steps like non-maximum suppression (NMS), hindering efficiency.
Purpose of the Study:
- To develop a novel 3D lung nodule detection framework, FCMamba, that balances local and global contextual awareness with low computational complexity.
- To minimize reliance on manual threshold tuning and redundant post-processing in lung nodule detection.
- To introduce a flexible connected visual state-space model adapted from the Mamba architecture for enhanced spatial modeling.
Main Methods:
- Proposed FCMamba, a flexible connected visual state-space model, incorporating a flexible path encoding strategy for adaptive 3D feature sequence reordering.
- Integrated a Top Query Matcher, guided by the Hungarian matching algorithm, to replace traditional NMS for end-to-end one-to-one nodule matching.
- Trained and evaluated the model using 10-fold cross-validation on the LIDC-IDRI dataset (888 CT scans).
Main Results:
- FCMamba outperformed state-of-the-art CNN, Transformer, and hybrid models across various false positive per scan (FPs/scan) levels.
- Achieved a sensitivity improvement of 2.6% to 20.3% at 0.125 FPs/scan and delivered superior CPM and FROC-AUC scores.
- Demonstrated balanced performance across nodule sizes, reduced false positives, and improved robustness, especially in high-confidence predictions.
Conclusions:
- FCMamba offers an efficient, scalable, and accurate solution for 3D lung nodule detection.
- The model's flexible spatial modeling and elimination of post-processing make it suitable for clinical applications.
- The proposed framework is adaptable for other medical imaging tasks, highlighting its versatility.
