Ensemble CNN for colon cancer detection using histopathological image.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Colon cancer is a leading cause of mortality, with delayed detection reducing survival rates.
• Traditional histopathological analysis is time-consuming, contributing to diagnostic delays.
• Artificial intelligence offers potential for faster and more accurate cancer detection.

Purpose of the Study:

• To develop an AI-driven ensemble model for rapid and accurate colon cancer detection.
• To prioritize recall in the model for improved detection of various colon cancer types.
• To assess the model's generalizability beyond its training data.

Main Methods:

• A weighted ensemble model combining RegNet X, RegNet Y, Swin B, and VGG11 convolutional neural networks was developed.
• A novel preprocessing pipeline using histogram equalization and contrast stretching was implemented for image enhancement.
• The model was trained on 10,000 histopathological images of colon tissues (benign and malignant).

Main Results:

• The ensemble model achieved a test accuracy of 99% and a recall of nearly 100%.
• The model demonstrated superior recall metrics compared to traditional machine learning approaches.
• The model showed strong generalizability, performing well when tested on lung cancer images.

Conclusions:

• The proposed AI ensemble model significantly enhances the speed and accuracy of colon cancer detection.
• The model's high recall and generalizability show promise for clinical applications in cancer diagnostics.
• AI-powered histopathological analysis represents a significant advancement in oncology diagnostics.