Download PDFOpen PDF in browserDeep belief Convolutional Neural Network with Artificial Image Creation by GANs Based Diagnosis of Pneumonia in Radiological Samples of the Pectoralis MajorEasyChair Preprint no. 4696, version 216 pages•Date: May 13, 2021AbstractIn this paper, we delineate a comparative classification of Pneumonia using Machine Learning and Deep Learning models. The data used was the dataset of Chest X-Ray Images for Classification made available by (Kermany, 2018) with a total of 5863 images, with 2 classes: normal and pneumonia. For the purpose of correcting the class imbalance between normal and pneumonia images, we use General Adversarial Networks to generate pneumonia ridden images. The comparative classification is built on a final dataset of 19784 images. Logistic regression, SVC, KNN, Random Forest and other machine learning models such as XgBoost and CatBoost are compared with deep learning models such as MobileNet and VGG-16. The machine learning model is based on blob segmentation and detecting the difference between blobs of pneumonia ridden and normal individuals. Finally, a new deep learning model with convolutional and artificial neural networks has been proposed for the classification purpose which increases the accuracy significantly and has a classification report which is best suitable for medical analysis. Keyphrases: activation function, ANN, Artificial Intelligence, Artificial Neural Network, Batch Normalization, BioImage sensing, Blob Segmentation, Chest Radiograph, Chest X-ray, class imbalance, CNN, computational power, Confusion Matrix, Convolutional Layer, Convolutional Neural Network, deep learning, deep learning model, F-CNN, GAN, logistic regression, machine learning, Machine Learning Algorithm, Max Pooling, neural network, neural networks, phoenix model, pneumonia image, pneumonia ridden, Pre-processing, structural co occurrence matrix, Transfer Learning, Validation Set
|