Enhancing Diagnostic Precision in Oncology: A Machine Learning Approach for Tissue Sample Classification

Lung cancer significantly affects the respiratory system and is characterized by prodromes such as a prolonged cough, dyspnea, and angina. Smoking remains the leading risk factor for lung cancer. In contrast, colon cancer affects the digestive system. A Computed Tomography (CT) scan can provide valuable insights for diagnosing lung diseases. This study aims to utilize advanced ML and deep learning techniques to classify the histopathological images of tissue samples into various classes, determining whether the cells are cancerous based on distinct morphological patterns. These patterns may include irregular shapes, increased density, and inconsistent sizes compared to normal cells. The goal is to reduce the reliance on pathologists to make manual diagnoses, which can be both time-consuming and subjective. In the proposed approach, a comparative analysis is conducted on image classes that have been preprocessed and trained from scratch using various models namely Support Vector Machine (SVM), Convolutional Neural Networks (CNN), and XGBoost. A CNN based on the EfficientNetB0 architecture is employed, which is the most minor and most efficient version in the EfficientNet family. This model is pre-trained on the ImageNet dataset for feature extraction, omitting the top fully connected layer to enable transfer learning. As a result, this model achieves an overall accuracy of 99.54%, surpassing all other models tested.