A New Convolution Neural Layer Based on Weights Constraints

With the rapid development of artificial intelligence techniques, emerging deep neural networks (DNN) is one of the most effective ways to solve many challenges. Convolution neural networks (CNNs) are considered one of the most popular AI techniques used to extract and analyze meaningful features for image datasets, especially in the medical diagnosis field. In this paper, a proposed constrained convolution layer (COCL) for the CNN model is proposed. The new layer uses a constrained number of weights in each kernel trained in the phase of learning and excludes the others weights with zero values. The proposed method is introduced to extract a special type of feature considering the local shape of a sub-image (window) and topological relations between group pixels. The features extract according to a random distribution of weights in kernels that are determined considering a particular desired percentage. Furthermore, this paper proposed a CNN model architecture that uses COCL rather than the traditional CNN layer (TCL). The efficiency of the method is evaluated using three types of medical image datasets compared with the traditional convolution layer, pre-trained deep neural networks (pre-DNNs), and state-of-art methods. The proposed model outperforms other methods in terms of accuracy and F1 score metrics and exceeds more than 98%, 89%, and 93% for the three datasets used in the evaluation, respectively. © 2022 IEEE.

GLCMs Based multi-inputs 1D CNN Deep Learning Neural Network for COVID-19 Texture Feature Extraction and Classification

Coronavirus disease 2019 epidemic (COVID-19) is an infectious disease that appeared because of the newest version of discovered coronavirus. The advent and rapid spread of this disease over the world necessitated a concerted effort to contain and eradicate it. Computer Tomography (CT) imaging and X-Ray images are considered as one of the important medical examinations used for disease diagnosis. To speed up and confirm the correctness of the medical diagnosis, many artificial intelligence techniques and machine learning methods are proposed. In this paper, a new and efficient proposed system is introduced to extract appropriate and meaningful features for CT scans and X-Ray COVID-19 images. The proposed method depends on extracting statistical texture features of the images using the GLCM method. The GLCMs matrices are extracted from different three quantized versions of the original image in different distances and directions. New multi-inputs 1D CNN architecture of the deep neural network is implemented to extract the effective features directly from GLCMs matrices after reducing its dimensions using the PCA technique. Three datasets are used to evaluate our method that includes SARS-CoV-2 CT-scan, COVID-CT, and DLAI3 Hackathon COVID-19 Chest X-Ray datasets. The proposed system achieved a classification improvement in terms of accuracy, F1 score, and AUC metrics compared with other methods and exceeds 98%, 89%, and 93% for three datasets, respectively. © 2022 University of Kerbala.