Analysis of Feature Extraction and Classification Methods on Histopathological Images for Diagnosing Invasive Ductal Carcinoma
Elvira Sukma Wahyuni, Vera Giyaning Tiyas, Suatmi Murnani · 2022
Invasive Ductal Carcinoma (IDC) is one of the most common types of breast cancer, accounting for almost 70-80% of all breast cancer diagnoses. Early detection of breast cancer is an effort to prevent or control the occurrence of breast cancer. The first step in treating breast cancer is a proper examination, typically conducted using biomedical images such as mammograms and histopathology images. In this study, we conducted breast cancer detection using histopathological images in several steps, namely histopathological data input, feature extraction, and classification. As feature extraction is the most important step in the classification process, we combined several feature extraction techniques, including Gray-Level Co-occurrence Matrix (GLCM), Red Green Blue (RGB), Hue Saturation Value (HSV), and Histogram. The feature extraction results were subsequently classified using the backpropagation and Support Vector Machine (SVM) classifiers. We used a public dataset from Kaggle that has 1,080 breast histopathology images consisting of 540 negative IDC and 540 positive IDC. Our experimental results show that combining the aforementioned feature extraction techniques and using a backpropagation classifier obtained the best prediction accuracy of 96.94%. The result implies that IDC type breast cancer based on histopathological images can be detected accurately using suitable methods.