Histopathological Image-Based Classification of Lung and Colon Cancer Using Deep Learning Architectures with Preprocessing Enhancements

Lung and colon cancer detection through histopathological images is a critical application of deep learning in medical diagnostics. In this study, we utilized the LC25000 dataset to classify lung and colon cancer using various deep-learning architectures. We applied EfficientNet-B0 and a custom Convolutional Neural Network (CNN) model, achieving high classification accuracies. For lung cancer, EfficientNet-B0 and CNN yielded accuracies of 99.97% and 99.40%, respectively, while both models achieved 100% accuracy in colon cancer classification in the test images. This outstanding performance is attributed to optimized parameter tuning and an effective image preprocessing pipeline, which includes Contrast Limited Adaptive Histogram Equalization (CLAHE), pseudo-color mapping, and image normalization. Additionally, we compared these results with other pre-trained models, Inception (97.6%, 98.2%), VGG19 (98.5%, 98.1%), and MobileNet (99.1%, 98.8%), demonstrating the superior performance of EfficientNet-B0 and CNN. These findings highlight the importance of optimized deep-learning models and preprocessing techniques for accurate cancer classification.