Ekstraksi Fitur Citra Radiografi Thorax Menggunakan DWT dan Moment Invariant

Feature Extraction of Thorax Radiography Image Using DWT and Moment Invariant

Abstract

COVID-19 is an infectious disease caused by the coronavirus family, namely severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The fastest method to identify the presence of this virus is a rapid antibody or antigen test, but confirming the positive status of a COVID-19 patient requires further examination. Lung examination using chest X-ray images taken through X-rays of COVID-19 patients can be one way to confirm the patient's condition before/after the rapid test. This paper proposes a feature extraction model to detect COVID-19 through chest radiography using a combination of Discrete Wavelet Transform (DWT) and Moment Invariant features. In this case, haar wavelet transform and seven Hu moments were used to extract image features in order to find unique features that represent chest radiographic images as suspected COVID-19, pneumonia, or normal. To find out the uniqueness of the proposed features, it is coupled with the kNN and generic ANN classification techniques. Based on the performance parameters assessed, it turns out that the wavelet-based and moment invariant thorax radiographic image feature model can be used as a unique feature associated with three categories: Normal, Pneumonia, and Covid-19. This is indicated by the accuracy value of 82.7% in the kNN classification technique and the accuracy, precision, and recall of 86%, 87%, and 86% respectively with the ANN classification technique.