Enhancing Accuracy in Classification and Detection of White Blood Cancer Cells using Wavelet Transform over Morphological Segmentation

Abstract

Aim: The ultimate aim of this research work is to improvize the accuracy & specificity of automatic counting of Cancer cells, detection, classification of cancer cells using innovative white blood cancer detection methodology. Materials and Methods: Determined sample size using GPower is 10 for each group (Power of 0.80 and alpha value of 0.05) and groups are categorized as Morphological segmentation classifier (Group 1) and Wavelet transform classifier (Group 2). 70 % of the images are utilized for training and 30 % are used for verification and validation in performance analysis. Result: Morphological segmentation algorithm achieved improved accuracy (97.77%) compared with the Wavelet transform algorithm with an accuracy of (77.77%). Independent sample T-test has been analyzed and achieved a significance of 0.0427 ($\mathrm{p} < 0.05$) for accuracy and 0.006 ($\mathrm{p} < 0.05$) for specificity. Conclusion: Morphological segmentation algorithm provides better accuracy compared with the wavelet algorithm.