Search

The main aim of this study is to identify the shape variation of lungs which are affected by COVID-19 with the help of Hough Transform and Radon transform features. The dataset is taken from the kaggle open source. A total of 200 image samples are obtained using G-Power, by fixing the parameters such as effect size (d) is given 0.57, standard error rate is given as 0.05 and power is given as 0.80 respectively. These segmented images are used in this analysis for differentiation of normal and COVID subjects using the Radon transform for the shape features extraction. By using SPSS software the significance of the best algorithm is obtained. Shape features which are extracted by Hough transform and Radon transform algorithms are classified by using KNN and SVM classifiers. From the obtained results, the feature values of Radon moments are observed to be statistically significant for M15,7 (p=0.03) compared to Hough transformed moments M10,8 (p=0.06). The mean and standard values of M15,7 for inter (normal and COVID) subjects are (0.51±0.24, 0.10±0.05) for Radon normalised features. The mean and standard deviation values of M10,8 for inter subjects are (0.21±0.04, 0.18±0.03) for Hough normalised features. From the above mentioned values, it is clear that the COVID subjects have loss in shape of lungs due to abnormality. The accuracy and F1 score values of Radon transform are 100 and 100 using KNN and SVM classifiers respectively. Conclusion: Therefore, from the above analysis it is clearly observed that Radon transform provides better classification between normal and COVID subjects when compared to Hough transform. [ABSTRACT FROM AUTHOR]

The aim of this analysis is to analyze the quantity of textural variations of lung CT scan images and extraction of features despite incidents of COVID-19. Essential input images for this analysis are acquired from a public domain database called Github.org. In this analysis, the comparison between Local Directional Texture Pattern (LDTP) and Local Directional Number Pattern (LDNP) to analyze the texture deformation of CT lung scans is carried out. Classification of normal and COVID subjects is carried out using Logistic Regression classifier. In COVID subjects textural variation is more despite tissue annihilation and in normal subjects textural variation is less due to the gentle surface of the lung. Feature values acquired using LDN are observed to be statistically significant (p<0.05) compared to LDTP values. LDN has obtained precision (0.891), F1-score (0.871), Recall (0.876), AUC (0.933), Classification Accuracy (0.876) are calculated using logistic regression classifier. Higher values of LDN of normal indicates that in COVID subjects there will be structural and intensity information loss despite the presence of COVID. LDN is tough against illumination, noise, time lapse variations. The extracted textural features are able to confine the significant difference in the normal and COVID CT scan images. [ABSTRACT FROM AUTHOR]

The aim of this analysis is to estimate the texture deformation in lung CT images caused by COVID 19. For this analysis, a total of 176 sample sizes are obtained based on the given parameters. The input images are gathered from Github.org. This analysis comes in for the comparison between Local Binary Pattern and Gradient Descriptor Pattern and Gradient Local Ternary Pattern to analyze the textural changes in Computed Tomography (CT) lung Scans. To classify acquired characteristics, a K-NN and NN classifier are used. The feature values obtained using Gradient local ternary pattern are found to be statistically significant(p<0.05) compared to Local binary pattern and Gradient directional pattern. The significant values of Gradient local ternary pattern are GLTP86, GLTP70, GLTP82with mean values of normal (0.26, 0.19, 0.16) and COVID (0.12, 0.21, 0.13). GLTP acquired Area under curve(0.960), classification accuracy (0.838), F1-score(0.831), precision(0.848), and recall(0.838) are obtained using K-NN classifier. The most significant feature value of LBP and GDP are L11andG94which shows significant ability in differentiating normal(9.46) and COVID(0.38) of LBP and normal(10.80) and COVID(5.86) of GDP. As a result of the low LBP and GDP values, information in the CT images would be lost. The extracted textural features are consistent in demonstrating a substantial difference between normal and COVID CT scan images. [ABSTRACT FROM AUTHOR]

To estimate the texture deformation in lung CT images caused by COVID 19 using local binary pattern and Weber local descriptor Pattern in comparison with local gradient increasing pattern features. Materials and Methods: G-power tool is used to calculate the total number of samples required for the analysis. A total of 176 samples per group is obtained by fixing the standard error rate, effect size, and predefined power parameters in G power tool are set to 0.05, 0.3, and 0.80, respectively. Results: The feature values obtained using Local Gradient Increasing Pattern are found to exhibit the significance of 0.0001 (p<0.05) compared to local binary pattern and Weber Local Descriptor. Significant Local Gradient Increasing Patterns are LGIP11, LGIP19, LGIP23, with mean values of normal as 0.42, 0.48, and 0.45 respectively and COVID as 0.35, 0.39, and 0.41 respectively. Neural network classifier is found to achieve a high accuracy of 0.98 using LGIP features in differentiating normal and COVID subjects compared to logistic regression and K-NN classifier. Conclusion: Local gradient increasing patterns perform significantly better than the local binary pattern and Weber local descriptor in differentiating normal and COVID subjects. [ABSTRACT FROM AUTHOR]