Development and Validation of a Deep Learning Model for Histopathological Slide Analysis in Lung Cancer Diagnosis

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Lung cancer is a leading cause of cancer mortality globally.
• Delayed diagnosis and poor prognosis significantly contribute to lung cancer fatalities.
• Deep learning (DL) offers potential for early lung tumor detection and treatment monitoring via medical imaging.

Purpose of the Study:

• To develop and evaluate a DL-based model for efficient lung cancer detection using whole-slide images.
• To assess the model's accuracy, robustness, and speed in distinguishing cancerous from non-cancerous lung cells.
• To compare the model's performance against that of experienced pathologists.

Main Methods:

• Utilized a DL model combining convolutional neural networks (CNNs) and separable CNNs with residual blocks.
• Trained and tested the model on whole-slide images for lung cancer detection.
• Evaluated model performance metrics including accuracy and detection time.

Main Results:

• The DL model achieved 96% to 98% accuracy in distinguishing cancerous from non-cancerous lung cells in under 10 seconds.
• The model demonstrated superior performance compared to pathologists, achieving 100% accuracy versus pathologists' 79%.
• A significant positive correlation was found between pathologists' accuracy and their years of experience (r=0.71, p=0.022).

Conclusions:

• The developed DL model significantly enhances lung cancer detection accuracy and efficiency.
• The model shows potential for assisting in the training of junior pathologists.
• AI-powered image analysis can improve early lung cancer diagnosis and patient outcomes.