Lung X-Ray Image Enhancement to Identify Pneumonia with CNN
Nur Nafi’iyah, Endang Setyati · 2021
Chest x-rays have various values of intensity and high contrast. Chest X-rays require a contrast stretching process so that the image can be analyzed and diagnosed correctly. Image contrast improvement can be based on the histogram value of the image intensity. Pneumonia can be diagnosed by taking chest X-rays. Diagnosis of Pneumonia based on chest x-rays can be done automatically by a computer. Computer-based Pneumonia diagnosis requires a reliable and accurate algorithm. A reliable and accurate algorithm, namely Convolutional Neural Network. This research aimed to prove whether the chest X-ray image that was performed by contrast improvement had a significant effect in diagnosing Pneumonia. The algorithm proposed in diagnosing Pneumonia is a Convolutional Neural Network. The CLAHE repaired chest X-ray image was trained with 8 CNN architectural models. The training results of the eight CNN architectural models respectively have a loss function value of 0.0057, 0.028, 0.0964, 0.0446, 0.0473, 0.0573, 0.0979, 0.1407. The results of diagnostic testing for Pneumonia in the eight CNN architectural models were 79.65%, 79.01%, 80.29%, 76.92%, 82.53%, 80.45%, 79.81%, 78.04%, respectively. The highest accuracy result when testing is 82.53% with the CNN 35 Layers architectural model, with a description of the input image is grayscale with a size of 224x224.