Pneumonia Detection through Adaptive Deep Learning Models of Convolutional Neural Networks

• Published in: 2020 11th IEEE Control and System Graduate Research Colloquium (ICSGRC) pp. 88 - 93
• Main Authors: Militante, Sammy V., Dionisio, Nanette V., Sibbaluca, Brandon G.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2020
• Subjects: Adaptive Deep Learning; Deep Convolutional Neural Network Architecture; Pneumonia Detection
• DOI: 10.1109/ICSGRC49013.2020.9232613

An infectious illness known as Pneumonia is often caused by infection due to a bacterium in the alveoli of lungs. When an infected tissue of the lungs has inflammation, it builds up pus in it. To find out if the patient has Pneumonia, experts conduct physical exams and diagnose their patients through Chest X-ray, ultrasound, or biopsy of lungs. Misdiagnosis, inaccurate treatment, and if the disease is ignored, it may lead to the death of a patient. The progression of Deep Learning contributes to the aid in the decision-making process of the experts to diagnose patients with pneumonia. The study employs flexible and efficient approaches of deep learning applying six models of CNN in predicting and recognizing a patient unaffected and affected with the disease employing a chest X- ray image. GoogLeNet, LeNet, VGG-16, AlexNet, StridedNet, and ResNet-50 models with a dataset of 28,000 images and using a 224x224 resolution with 32 and 64 batch sizes are applied to verify the performance of each models being trained. The study likewise implements Adam as an optimizer that maintains an adjusted 1e-4 learning rate and an epoch of 500 employed to all the models. Both GoogLeNet and LeNet obtained a 98% rate, VGGNet-16 earned an accuracy rate of 97%, AlexNet and StridedNet model obtained a 96% while the ResNet-50 model obtained 80% during the training of models. GoogleNet and LeNet models achieved the highest accuracy rate for performance training. The six models identified were capable to detect and predict a pneumonia disease including a healthy chest X-ray.