performance analysis for breast cancer histopathology image classification using CNN

Abstract—The paper provides an in depth performance on
Conventional CNN analysis using the histopathology image
classification of breast cancer. Breast cancer is currently one
of the most important health problems. in the rest of the
world, and histopathology image are diagnosed while the body
still is in its early stages. This plays an important role in
mortality reduction. The study utilizes the BreakHis dataset of
different benign and malignant, tissue subtypes. The two major
dimensions of analysis are performed: non-technical analysis
of the significance of diabetes mellitus imaging and diagnostic
problems, and technical analysis using hierarchical CNN model
where EfficientNetB4 multi-level classification (benign, malignant
and their subtypes respectively). The model achieves high main
class prediction accuracy of 89.7% validity. subclass accuracy
of 78.9%.Performance evaluation uses accuracy measures, loss
curves, confusion tables, and confidence scores. Also, the Grad-
CAM visualization is used to understand the model decision
regions, showing the demonstration of the attention of CNN to
tissue structures of physiological interest. The results certify the
usefulness of CNN-based solutions in aiding swift and dependable
breast cancer diagnosis by the pathologists.
Keywords:Breast Cancer, Histopathology, CNN, Classification,
Deep Learning, Grad-CAM, Medical imaging, EfficientNet.