Detection of pulmonary ground-glass opacity based on deep learning computer artificial intelligence

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Deep learning artificial intelligence (AI) systems show promise for the early detection of pulmonary nodules and ground glass opacities (GGOs).
• Accurate identification of these lung abnormalities is crucial for timely diagnosis and treatment.

Purpose of the Study:

• To evaluate the efficacy of deep learning models for detecting pulmonary nodules and GGOs.
• To compare the performance of different deep learning architectures (AlexNet, GoogLeNet, ResNet50) in identifying lung abnormalities.

Main Methods:

• Utilized computed tomography (CT) images from the LIDC-IDRI database for pulmonary nodule detection using AlexNet and GoogLeNet.
• Employed 221 GGO images from Xinhua Hospital with ResNet50 for GGO detection.
• Implemented CT image radial reorganization for 3D feature input and deep learning analysis.

Main Results:

• Achieved 88.0% accuracy and an F-score of 0.891 in identifying lung nodules.
• The GGO nodule classification reached a peak F-score of 0.87805.
• A novel RGB superposition preprocessing method enhanced nodule-tissue differentiation, outperforming existing solutions.

Conclusions:

• The proposed deep learning method demonstrates superior sensitivity compared to recent systems for lung abnormality detection.
• The AI system exhibits a lower average false positive rate, enhancing diagnostic reliability.