Texture feature based classification on microscopic blood smear for acute lymphoblastic leukemia detection. (January 2019)

• Record Type: Journal Article
• Authors: Mishra, Sonali; Majhi, Banshidhar; Sa, Pankaj Kumar
• Is Part Of: Biomedical signal processing and control. Volume 47(2019)
• Page Start: 303
• Page End: 311
• Publication Date: 2019-01
• Subjects: Acute lymphoblastic leukemia (ALL) -- Discrete orthonormal S-transform (DOST) -- AdaBoost with RF (ADBRF) -- Computer-aided diagnosis (CAD); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.08.012
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A 2-dimensional discrete orthonormal S-transform feature extraction scheme is employed to extract texture features followed by an appropriate combination of PCA and LDA feature reduction scheme to get the most relevant features. Performance evaluation of the proposed strategy by using Adaboost RF classifier is noted down for the classification of the disease. The research work can be extended towards the sub-classification of acute lymphoblastic leukemia. Abstract: This paper presents an effective scheme for classification of the normal white blood cells from the affected cells in a microscopic image. The proposed method initially pre-processes the input images using Y component of the CMYK image and a triangle method of thresholding. Subsequently, it utilizes discrete orthonormal S -transform (DOST) to extract the texture features, and its dimensionality is reduced using linear discriminant analysis. The reduced features are then supplied to the proposed Adaboost algorithm with RF (ADBRF) classifier where the random forest is used as the base classifier. A publicly available dataset, ALL-IDB1 is used to validate the proposed scheme. The simulation results based on the five runs of k -fold stratified cross-validation indicate that the proposed method yields superior accuracy (99.66%) as compared to existing schemes.