Tongue Detection For Identification Of Syndrome Diagnosis In Heart Disease Using Convolutional Neural Network

Abstract

Convolutional Neural Network (CNN) which is one of the Deep Learning methods for Image identification and CNN models can identify images well but in this case it requires higher accuracy because the case is very crucial to determine the risk of heart disease. The initial stage in this study was the collection of tongue image data, 4836 training data and 1209 testing data. The image data used were the front, right side, left side of the tongue and under the tongue. The dataset was obtained from taking pictures using a smartphone camera centimeters above the object. The distribution of data in each class is shown in the following figure. The model from the two CNN algorithm experiments has accuracy performance. Based on the training results the model from the algorithm gets an accuracy value and Testing by identifying 20% of the total dataset as test data. The identification results are formed in a Confusion Matrix to then be poured into a classification report and obtain: train loss 0.301446, train accuracy 0.862696, test loss 0.314132 and test accuracy 0.850290 so that from the results of the tongue data test it can be concluded that the accuracy value is quite good, above 80%.