Chest X-Ray Outlier Detection Model Using Dimension Reduction and Edge Detection

With the advancement of Artificial Intelligence technology, development various applied software and studies are actively conducted on detection, classification, prediction through interdisciplinary convergence integration. Among them, medical AI has been drawing huge interest popularity in Computer-Aided Diagnosis, which collects human body signals to predict abnormal symptoms health, diagnoses diseases images such as X-ray CT. Since CT medicine use high-resolution images, they require high specification equipment energy consumption due computation learning recognition, incurring costs create an environment for operation. Thus, this paper proposes a chest outlier detection model using dimension reduction edge solve these issues. The proposed method scans image window certain size, conducts difference imaging adjacent segment-images, extracts information binary format AND To convert extracted edge, is visual information, into series lines, it computed convolution with filter that coefficient 2 n lines divided 16 types. By counting converted data, one-dimensional 16-size array per one segment-image produced, reduced data used input RNN-based model. In addition, study experiments based COVID-chest dataset evaluate performance According experiment results, LFA-RNN showed highest accuracy at 97.5% calculated learning, followed by CRNN 96.1%, VGG 96.6%, AlexNet 94.1%, Conv1D 79.4%, DNN 78.9%. lowest loss about 0.0357.