Analysis and Comparison of Kidney Stone Detection using Minimum Distance to Mean and Gaussian Maximum Likelihood Classifier with Improved Classification Accuracy

Abstract

Aim: The goal of this research is to use Gaussian maximum Likelihood classifier and Minimum distance to mean Classifier to predict and detect kidney stones. Materials and Methods:This investigation made use of a collection of data from Kaggle website. Samples were collected (N=10) for normal kidney images and (N=10) for kidney with stone images. Total sample size was calculated using clinical.com. As a result the total number of samples 20 was considered for analysis. Using Matlab software and a standard data set collected from Kaggle website, the classification accuracy was obtained.Pretest G power 80, sample size calculation can be done through clinicalc.com. Results: The accuracy (%) of both classification techniques are compared using SPSS software by independent sample t-tests. There is a statistical significant difference between Gaussian maximum Likelihood classifier andMinimum distance to mean Classifier. Comparison results show that innovative Gaussian maximum Likelihood classifiers give better classification with an accuracy of (81.34%) than Minimum distance to mean Classifiers (78.85%).There is a statistical significant difference between Gaussian maximum Likelihood classifier and Minimum distance to mean Classifiers. The Gaussian maximum Likelihood classifier with p=0.022, p<0.05 hence significant and showed better results in comparison to Minimum distance to mean classifiers. Conclusion: The Gaussian maximum likelihood classifier appears to give better classification accuracy than the Minimum distance to mean Classifier.

Imprint

Kishore U, Ramadevi R. Analysis and Comparison of Kidney Stone Detection using Minimum Distance to Mean and Gaussian Maximum Likelihood Classifier with Improved Classification Accuracy. Cardiometry; Issue 25; December 2022; p.812-817; DOI: 10.18137/cardiometry.2022.25.812817; Available from: https://www.cardiometry.net/issues/no25-december-2022/classifier-improved-classification-accuracy