Breast Cancer Histopathology Image Classification Using Ensemble Feature Set
A.M Arul Raj., Lekha S. Nair · 2023
Breast cancer accounts for 30% of newly diagnosed tumours due to lifestyle and genetic causes. According to WHO data, breast cancer is the second most frequent form of cancer globally. It was even predicted that by 2020 there will be a rise of 2.3 million more instances. Histopathologists manually quantify mitosis, which is time-consuming and subjective. CAD, a secondary reader system, improves diagnosis and therapy. Recent technology advances have focused on mitosis detection utilising hand-crafted or CNN-learned characteristics. CNN-based techniques are computationally intensive, yet hand crafted features are inaccurate. Our technique combines Convolutional Neural Network-learned deep features with handcrafted features and employs an ensemble machine learning model to classify. The model was trained using MITOS-ATYPIA-14 grand challenge mitotic images. The model fed HC and deep features has an F1 score of 96%. Thus, this model is more reliable and computationally efficient than current methods.