Deep learning object detection-based early detection of lung cancer.

Area of Science:

• Medical imaging analysis
• Artificial intelligence in oncology
• Radiomics and computational pathology

Background:

• Early lung cancer diagnosis is crucial for effective treatment and improved patient survival.
• Artificial intelligence (AI) has significantly advanced medical image analysis capabilities.
• Object detection models are increasingly utilized for analyzing complex medical scans like CT images.

Purpose of the Study:

• To compare the performance of various You Only Look Once (YOLO) versions for lung cancer detection and classification in CT images.
• To evaluate the efficacy of YOLOv5, YOLOv8, YOLOv9, YOLOv10, and YOLOv11 algorithms on the Lung-PET-CT-Dx dataset.
• To determine the optimal YOLO model for assisting physicians in accurate lung cancer diagnosis and localization.

Main Methods:

• Utilized the public Lung-PET-CT-Dx dataset for model training and validation.
• Implemented and compared multiple versions of the You Only Look Once (YOLO) object detection algorithm, including YOLOv5, YOLOv8, YOLOv9, YOLOv10, and YOLOv11.
• Assessed model performance based on precision and recall rates for lung cancer detection and classification tasks.

Main Results:

• YOLOv8 achieved superior prediction results compared to other evaluated YOLO versions.
• YOLOv8 demonstrated a precision rate of 90.32% and a recall rate of 84.91% in lung cancer detection.
• The findings indicate significant potential for AI-driven tools in enhancing diagnostic accuracy.

Conclusions:

• YOLOv8 effectively assists physicians in the diagnosis of lung cancer.
• The model improves the accuracy of disease localization and identification in medical imaging.
• AI-powered object detection offers a promising approach to augment clinical decision-making in oncology.