Lung cancer histopathology image classification using transfer learning with convolution neural network model.

Area of Science:

• Computational pathology
• Artificial intelligence in oncology
• Medical image analysis

Background:

• Lung cancer (LC) is a significant health threat requiring early detection for effective treatment.
• Accurate histological classification is crucial for determining appropriate LC management strategies.
• Deep learning (DL) offers potential for analyzing complex histopathology images in LC diagnosis.

Purpose of the Study:

• To implement a pretrained EfficientNetB7 model for classifying lung cancer histopathology images.
• To categorize LC into primary malignancy types: adenocarcinoma, squamous cell carcinoma, and large cell carcinoma.
• To evaluate the classification performance using accuracy as the primary metric.

Main Methods:

• Utilized a dataset of 15,000 lung cancer histopathology images.
• Employed EfficientNetB7, a convolutional neural network (CNN) pretrained on ImageNet, for transfer learning.
• Trained the EfficientNetB7 model on the LC dataset and evaluated its performance.

Main Results:

• The EfficientNetB7 model achieved a high accuracy of 99.77% in classifying LC histopathology images.
• This performance surpasses or matches accuracy levels reported in previous studies (90-99%).
• Transfer learning with EfficientNetB7 effectively extracted relevant features for accurate classification.

Conclusions:

• CNN-based EfficientNetB7 model accelerates lung cancer diagnosis from histopathology images.
• This AI tool can alleviate the workload of pathologists, facilitating earlier patient treatment.
• Automated classification of LC aids in timely and precise therapeutic interventions.