Abstract:
Breast cancer is a prevalent cancer type that primarily affects women, although it can also occur
in men. It typically originates in the cells lining the breast's ducts or lobules, but it has the potential
to spread to other areas of the breast tissue and metastasize to other organs. The global incidence
of breast cancer has been increasing, leading to significant social, psychological, and economic
consequences. Researchers have investigated the use of artificial intelligence (AI) in breast cancer
screening and detection, showcasing its potential to enhance accuracy[8], [9], [10].
Various deep learning (DL) models incorporating convolutional neural networks (CNNs), transfer
learning, and hybrid approaches have achieved high accuracy rates in detecting and classifying
malignant breast cancer. However, additional research is necessary to explore alternative network
architectures, apply vital preprocessing tasks like noise removal& segmentation, investigate the
integration of multi-modal data, evaluate the model performance using various metrics and gather
larger datasets to enhance performance and enable early-stage cancer detection.
In study, utilized a dataset from the Kaggle repository consisting of 780 with 3 classes (benign,
malignant, or healthy) later increased to 6,280 images in PNG format using augmentation
techniques, each resized to 64x64 pixels, to determine the type of breast cancer. To eliminate noise,
we employed techniques such as adaptive median filtering (AMF), Gaussian filtering, and adaptive
histogram equalization (AHE) to extract the most important part of each image. Furthermore,
image segmentation algorithms including k-means clustering, waters
