[Pulmonary nodule detection method based on convolutional neural network].

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Pulmonary nodules are critical indicators for early lung cancer detection.
• Accurate and automated detection of pulmonary nodules in low-dose computed tomography (CT) images remains a challenge.
• Existing methods often require significant manual intervention or lack sufficient accuracy.

Purpose of the Study:

• To develop and evaluate a robust method for detecting pulmonary nodules in low-dose CT images.
• To improve the accuracy and efficiency of automated pulmonary nodule detection using deep learning.
• To provide an auxiliary diagnostic tool for early lung cancer screening.

Main Methods:

• Image preprocessing techniques including clipping and normalization were applied to low-dose CT images.
• Data augmentation was used to balance positive and negative samples for the convolutional neural network (CNN).
• A 2D CNN model was trained and optimized, with performance evaluated using five-fold cross-validation on the LUNA16 dataset.

Main Results:

• The proposed 2D CNN model achieved high performance metrics: 92.3% accuracy, 92.1% sensitivity, and 92.6% specificity.
• The model demonstrated improved performance compared to existing automatic detection and classification methods for pulmonary nodules.
• Perturbation experiments confirmed the model's stability and anti-interference capabilities.

Conclusions:

• The developed method effectively identifies pulmonary nodules in low-dose CT images.
• The model offers a stable and reliable tool for auxiliary diagnosis in early lung cancer screening.
• This approach has the potential to enhance the accuracy and efficiency of lung cancer diagnosis.