A Symmetric Encoder-Decoder Network with Enhanced Group-Shuffle Modules for Robust Lung Nodule Detection in CT Scans.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Lung cancer is a leading cause of death globally.
• Early detection of lung nodules via CT scans is crucial for improving patient outcomes.
• Current deep learning segmentation models face challenges like feature diversity, low spatial discrimination, and overfitting.

Purpose of the Study:

• To introduce an enhanced symmetric encoder-decoder segmentation network, the Improved Group-Shuffle Module (IGSM).
• To address limitations in current deep learning models for lung nodule segmentation.
• To improve the accuracy and reliability of automated lung nodule detection.

Main Methods:

• Developed the Improved Group-Shuffle Module (IGSM) inspired by human brain processing.
• Hierarchically divided feature maps into groups, processed them independently, and shuffled channels to enhance inter-group interaction and feature diversity.
• Optimized IGSM configurations including module placement, grouping size, and shuffle strategies.
• Compared the IGSM-enhanced model against U-Net and DeepLab using metrics like mIoU, Dice Score, Accuracy, Sensitivity, and Specificity.

Main Results:

• The IGSM-enhanced model demonstrated superior performance compared to benchmark models.
• Achieved a mean Intersection over Union (mIoU) of 0.7735, Dice Score of 0.9665, and Accuracy of 0.9873.
• The IGSM improved discrimination between nodules and background and enhanced generalization across various nodule morphologies.

Conclusions:

• The IGSM is an effective approach for enhancing deep learning-based lung nodule segmentation.
• The proposed method offers a reliable tool for automated lung cancer detection.
• The IGSM's ability to capture discriminative spatial and contextual patterns is key to its success, especially for complex nodule structures.