Advancing Cancer Classification with Hybrid Deep Learning: Image Analysis for Lung and Colon Cancer Detection

Abstract

This paper presents a groundbreaking approach to classifying lung and colon cancer from high- dimensional histopathological images, employing an advanced hybrid deep learning model. Our dataset encompasses 20,000 high-resolution images across five distinct classes, posing significant challenges in terms of computational efficiency and model accuracy. The dual-CNN structure of our model ensures a comprehensive extraction of both fine-grained and global features, crucial for accurate cancer classification. Through innovative feature reduction techniques, we effectively mitigate the curse of dimensionality, enhancing the model’s computational efficiency and robustness. The integration of global average pooling and dense layers with dropout regularization further contributes to the model’s performance, preventing overfitting and ensuring generalizability. Our approach achieves an impressive classification accuracy of 99%, demonstrating the model’s capability to handle high-dimensional datasets with precision. This work marks a significant contribution to medical image analysis, providing a reliable and efficient solution for cancer classification and setting a new standard in the field.