Enhancing Breast Cancer Histopathological Image Classification: The Impact of Stain Normalization on Multilevel Feature Extraction

Abstract

Histopathological examination involves the microscopic analysis of breast tissue samples, which is crucial for diagnosing and classifying breast cancer. Recent advancements in artificial intelligence have led to the development of machine learning and deep learning models for classifying histopathological images. Feature extraction depends heavily on the quality and uniformity of the input data. Stain normalization techniques are employed to standardize the visual appearance of images, ensuring a consistent representation of underlying tissue structures. This research explores the impact of stain normalization on various image features used to classify breast cancer histopathological images. We investigate low-, mid-, and high-levels of feature extraction with different classifiers for the multiclass classification of the Breast Cancer Histology dataset (BACH). The findings demonstrated that stain normalization methods are instrumental in enhancing the interpretability and dependability of histopathological image analysis across multiple levels.