Explainable AI based Lung and Colon Cancer diagnosis using Enhanced Feature Analysis

Histopathological and CT scan image analysis plays a critical role in the fast-developing area of medical diagnostics by helping to identify and categorize different types of cancer. Considering lung and colon cancers having high incidence and fatality rates, they are especially important. This work presents a strategy for using histopathological image analysis to identify and categorize these tumors. To extract relevant picture characteristics, advanced image processing techniques are employed across several color spaces, especially using Gabor features, Discrete Fourier Transform (DFT), Local Directional Pattern (LDP), Local Binary Pattern (LBP), and Discrete Wavelet Transform (DWT). For classification tasks, these characteristics are subjected to many machine learning classifier implementations. A methodical analysis of classifier performance in several color spaces is carried out, demonstrating how color space selection affects the efficacy of feature extraction techniques. Three datasets containing histopathological images and CT scan images have been used, where Random Forest and Gradient Boosting gave 100% accuracy and 89% accuracy in first and second dataset in greyscale images respectively and Random Forest gave 96% accuracy in third datasets in HSV colorspace. The results emphasize the relative advantages of each technique as well as the significance of color space in machine learning model optimization for cancer diagnosis in histopathology pictures.