Image Analysis For Breast Cancer Classification Using Learning Vector Quantization (LVQ) Method

Abstract

Breast cancer is a common disease in women, making early detection crucial to improve treatment effectiveness. This study aims to create a breast cancer classification system using MATLAB and the Learning Vector Quantization (LVQ) algorithm through mammography image analysis. The data used was taken from the public platform Kaggle. The process includes preprocessing (conversion to grayscale and normalization), texture feature extraction with Gray Level Co-occurrence Matrix (GLCM), LVQ model training, and performance evaluation using accuracy, precision, recall, and F1-score. Test results show that the LVQ model can achieve an accuracy of 80.45%, precision of 78.92%, recall of 100%, and an F1-score of 88.30%. The system is equipped with a MATLAB-based user interface (GUI) that allows for direct image classification. Although the results are positive in detecting cancer images, errors in classifying normal images are still present. Future improvements will focus on data balancing and improving model performance. This system is expected to be a tool for rapid and accurate early screening of breast cancer in clinical settings.