Contactless Intelligent Anti-Interference Lung Nodule Detection Method for Early Disease Detection

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Accurate lung nodule detection is crucial for early-stage lung cancer treatment.
• Computed tomography (CT) scanning is vital, but image quality can be compromised by radiation interference and patient movement.
• Existing methods struggle with detecting nodules in noisy or disturbed CT images.

Purpose of the Study:

• To develop an artificial intelligence (AI)-based method for robust lung nodule detection in CT images under interference conditions.
• To improve the accuracy and stability of lung nodule detection despite image noise and artifacts.
• To enhance early lung cancer diagnosis through improved nodule identification.

Main Methods:

• Proposed an AI-based anti-interference lung nodule detection method, Yolov8-AH, integrating Yolov8 with adaptive gating sparse attention (AGSA) and haar wavelet downsampling (HWD).
• AGSA module was designed to stabilize detection by focusing on critical image regions, mitigating the impact of disturbances.
• HWD module was employed to enhance nodule visibility by prioritizing relevant frequency components and reducing noise without blurring nodule edges.

Main Results:

• The Yolov8-AH model demonstrated significant improvements in detecting lung nodules under interference conditions.
• Ablation studies and experiments with varying noise levels confirmed the model's effectiveness.
• Achieved a 24% improvement in mean Average Precision at IoU threshold 0.5 (mAP50) and an 8.2% increase in precision compared to existing models.

Conclusions:

• The Yolov8-AH model offers a superior solution for accurate lung nodule detection in compromised CT images.
• The integration of AGSA and HWD modules effectively addresses interference, enhancing diagnostic capabilities.
• This AI-driven approach holds significant potential for improving early lung cancer diagnosis and patient outcomes.