Klasifikasi Nodul Kanker Payudara pada Citra Ultrasonografi (USG) berdasarkan Fitur Posterior

Abstract

Breast cancer is a deadly disease, especially in women. Based on Estimated New Female Breast Cancer Cases and Deaths by Age, US, 2017 the number of Carcinoma In Situ cases for all ages reached 63,410 cases, for Carcinoma Invasive cases it reached 252,710 cases and for the death rate reached 40,610. Ultrasonography is one of the most commonly used methods for screening for breast cancer, this ultrasound method is cheaper and safer. But the results of screening using ultrasonography are very subjective. Therefore, we need a system that is able to diagnose the severity of breast cancer with various parameters, one of which is the posterior feature. The posterior characteristics used are divided into three classes namely posterior acoustic enhancement, posterior acoustic shadow and no features. This final project proposes a simple method, namely by extracting areas that are suspected of having posterior characteristics. The initial stage of pre-processing starts with manually cropping the nodule image then filtering the image using an adaptive median filter. Then the segmentation process is done using the active contour method. The characteristics of segmentation results in feature extraction by taking the average intensity value from the area under the nodules using the method of mean comparison. Finally the classification is done using Multilayer Perceptron (MLP The final results after the selection of features showed 73.25% accuracy performance, 80.3% specification, 51.3% sensitivity. This indication shows that the classification process can work well.