Automated Cancer Detection in Histopathology Images using Improved Convolutional Neural Networks

Cancer causes so many deaths across the globe, due to the lack of screening and awareness precautions along with the medical facilities shortage. The histopathological images in manually grading faced difficulties such as time-consuming and accurate diagnosis. An existing approached struggled to handle high diagnostic cancer detection accuracy across varying image modifications and diverse histopathological environments, specially in preserving discriminative features in high-resolution data. To address these limitations, this research proposes an improved deep learning model namely Improved Convolutional Neural Networks (CNNs) adeptly capture spatial hierarchies within the images, enhancing the model's robustness against domain shifts and varying magnifications. Utilizing the BreakHis dataset, then the pre-processing technique namely histogram equalization used to balance the low and high contrast of the image thereby enhancing the image quality. The feature extraction technique like Principal Component Analysis (PCA) utilized to extract the features to improve the detection accuracy. This improved approach enhances the preservation of critical diagnostic features, leading to improved detection efficiency. The proposed improved CNN model attains an accuracy of 96.93%, outperforming existing methods such as XGBoost and Support Vector Machine (SVM), thereby demonstrating its efficacy in breast cancer subtype classification.