Advances in Image Processing Methods for Breast Cancer Diagnosis and Classification Using A Hybrid Machine Learning – A Systematic Review

Abstract

Breast Cancer (BC) is the second leading cause of death among many women worldwide. Detecting and treating BC can be challenging for radiologists. Basic care plays a crucial role in preventing diseases and reducing mortality. This study aims to improve treatment options and increase survival rates through early identification. Deep Learning (DL) and Machine Learning (ML) methods are powerful tools in identifying BC, offering enhanced precision and effectiveness. However, current techniques face scalability and performance limitations, prompting further investigation. This work introduces an Integrated Image Processing method for BC Diagnosis and Classification (IIP-BCDC). The approach combines DL using the AlexNet model for extracting Deep Features (DF) and ensemble-based ML algorithms for classification. AlexNet is implemented with five Convolutional Levels (ConvL) and three Completely Connected Levels (CCL) to preserve multidimensional DF while ensuring optimal performance. The unique DF extracted from the AlexNet DL models are consolidated into a single feature set, used as input for a Support Vector Machine with a Radial Basis Function (SVM-RBF) for a two-level categorization task. Comprehensive tests were conducted using a publicly accessible database of Invasive Ductal Carcinoma (IDC) images from breast biopsy. The results from thorough studies provide compelling evidence of the resilience and outstanding performance of the proposed hybrid approach. The proposed AlexNet model surpassed contemporary models with a higher accuracy rate of 98.7%, precision rate of 98.6%, recall rate of 99%, and an F1-score of 98.65%.