No abstract

The novel coronavirus termed as covid-19 has taken the world by its crutches affecting innumerable lives with devastating impact on the global economy and public health. One of the major ways to control the spread of this disease is identification in the initial stage, so that isolation and treatment could be initiated. Due to the lack of automated auxiliary diagnostic medical tools, availability of lesser sensitivity testing kits, and limited availability of healthcare professionals, the pandemic has spread like wildfire across the world. Certain recent findings state that chest X-ray scans contain salient information regarding the onset of the virus, the information can be analyzed so that the diagnosis and treatment can be initiated at an earlier stage. This is where artificial intelligence meets the diagnostic capabilities of experienced clinicians. The objective of the proposed research is to contribute towards fighting the global pandemic by developing an automated image analysis module for identifying covid-19 affected chest X-ray scans by employing an optimized Convolution Neural Network (CNN) model. The aforementioned objective is achieved in the following manner by developing three classification models, (i) ensemble of ResNet 50-Error Correcting Output Code (ECOC) model, (ii) CNN optimized using Grey Wolf Optimizer (GWO) and, (iii) CNN optimized using Whale Optimization + BAT algorithm. The novelty of the proposed method lies in the automatic tuning of hyper parameters considering a hierarchy of MultiLayer Perceptron (MLP), feature extraction, and optimization ensemble. A 100% classification accuracy was obtained in classifying covid-19 images. Classification accuracy of 98.8% and 96% were obtained for dataset 1 and dataset 2 respectively for classification into covid-19, normal, and viral pneumonia cases. The proposed method can be adopted in a clinical setting for assisting radiologists and it can also be employed in remote areas to facilitate the faster screening of affected patients.

An efficient detection of Optic disc in colour retinal images is the fundamental step in an automated retinal image analysis system. This paper presents a new approach for the automatic localization and accurate boundary detection of the optic disc. Iterative thresholding method followed by connected component analysis is employed to locate the approximate center of the optic disc. Then geometric model based implicit active contour model is applied to find the exact boundary of the optic disc. The method is evaluated against a carefully selected database of 148 retinal images and compared with the human expert. The optic disc is localized with an accuracy of 99.3%. The sensitivity and specificity of boundary detection achieved in terms of Mean±SD are 90.67±5 and 94.06±5 respectively.

The world is significantly affected by infectious coronavirus disease (covid-19). Timely prognosis and treatment are important to control the spread of this infection. Unreliable screening systems and limited number of clinical facilities are the major hurdles in controlling the spread of covid-19. Nowadays, many automated detection systems based on deep learning techniques using computed tomography (CT) images have been proposed to detect covid-19. However, these systems have the following drawbacks: (i) limited data problem poses a major hindrance to train the deep neural network model to provide accurate diagnosis, (ii) random choice of hyperparameters of Convolutional Neural Network (CNN) significantly affects the classification performance, since the hyperparameters have to be application dependent and, (iii) the generalization ability using CNN classification is usually not validated. To address the aforementioned issues, we propose two models: (i) based on a transfer learning approach, and (ii) using novel strategy to optimize the CNN hyperparameters using Whale optimization-based BAT algorithm + AdaBoost classifier built using dynamic ensemble selection techniques. According to our second method depending on the characteristics of test sample, the classifier is chosen, thereby reducing the risk of overfitting and simultaneously produced promising results. Our proposed methodologies are developed using 746 CT images. Our method obtained a sensitivity, specificity, accuracy, F-1 score, and precision of 0.98, 0.97, 0.98, 0.98, and 0.98, respectively with five-fold cross-validation strategy. Our developed prototype is ready to be tested with huge chest CT images database before its real-world application.

No abstract