Evaluating the Performance of Hough based Moments to Classify Healthy and COVID Subjects in Comparison with Traditional Shape Measures

The aim of this analysis is to measure and analyse the shape changes in Lung CT scans using orthogonal Zernike moments in comparison with traditional shape measures. Materials and Methods: A total sample size of 176 scans are acquired for this analysis, by assigning parameters such as the effect size = 0.3, standard error rate = 0.05 and algorithm power = 0.80 are predefined in Gpower software. In this analysis, the comparison between traditional shape measures and Hough Transform algorithms in classifying normal and COVID-19 is performed. Results: It is observed that there is no shape change in the lungs of the normal subjects and in COVID subjects the shape of the lungs reduces due to tissue loss. The feature values obtained from Hough Transform are found to be statistically important (p<0.05). The statistical values (Mean ± standard deviation) of normal and COVID subjects are 0.18 ± 0.13 and 0.10 ± 0.13. The significant features for the Zernike moment were M13,9, M10,8. The extracted values from the Computed Tomography images are consistent in displaying a considerable difference between healthy subject and COVID CT- scan images. The proposed Hough Transform based Zernike Moments algorithm has significantly better accuracy (97
^%
) than the Traditional shape measures with accuracy (78
^%
). Conclusion: The Hough transform based Zernike moments algorithm gives a significantly better result oriented to extraction of shape changes and manifestation of a significant difference in the healthy subject and COVID subject CT scan images than Traditional shape measures algorithm.