Addressing Data Intrinsic Characteristics for Augmentation for Breast Cancer Classification
Armaan Garg · 2023
Breast cancer is the most frequently diagnosed cancer among females worldwide. The task of correctly diagnosing cancer using histopathology in its very earlier stages is a challenging and critical task. Most of the present machine learning techniques require a lot of data to analyze and predict a benign tumour in its early stages, and such data is not available readily. In this paper, we propose the idea of data augmentation of breast cancer tissue images by addressing data intrinsic characteristics. The aim is to detect the micro presence of the tumour cells and highlight it over multiple synthetic images for classifiers to predict benign tumours in very early stages with high accuracy. The initial experimental analysis highlights the proposed technique’s impact and significance in boosting the performance of standard classifier(s).