Adaptive blood cell segmentation and hybrid Learning-based blood cell classification: A Meta-heuristic-based model. (May 2022)

• Record Type: Journal Article
• Authors: Anita Davamani, K.; Rene Robin, C.R.; Doreen Robin, D.; Jani Anbarasi, L.
• Is Part Of: Biomedical signal processing and control. Volume 75(2022)
• Publication Date: 2022-05
• Subjects: Adaptive Blood Cell Segmentation -- Adaptive Fuzzy C-Means clustering -- Hybrid Learning-based Blood Cell Classification -- Neural Network -- Long Short-Term Memory -- Best search-based Moth-Flame Optimization; 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.2022.103570
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Present a blood cell classification model to help medical practitioners. Performed accurate cell segmentation through A-FCM clustering with BS-MFO Using BS-MFO by solving multi-objective functions concerning variance and entropy. Implemented a novel BS-MFO algorithm for maximizing the accuracy rate of diagnosis. Suggested a new BS-MFO algorithm for optimizing parameters of LSTM and NN. Abstract: The fundamental goal of this paper is to develop a novel blood cell classification using a hybrid learning model. The proposed model encompasses different processing steps like "(a) pre-processing, (b) cell segmentation, (c) Feature extraction, and (d) classification". In the initial step, the blood smear images are pre-processed using Red Green Blue (RGB) scale to gray scale conversion and contrast enhancement. Then, the Adaptive Fuzzy C-Means (A-FCM) clustering with heuristic improvement is developed for blood cell classification. During testing, the feature extraction from the segmented cell image is performed by the Gray Level Co-occurrence Matrix1 (GLCM), Local Binary Pattern (LBP), geometric features, and color features. These features are subjected to the hybrid learning model with Neural Network (NN) and Long Short-Term Memory (LSTM) termed NLSTM. The modification of the A-FCM-based cell segmentation and hybrid learning-based cell classification is performed by a Best search-based Moth-Flame Optimization (BS-MFO) algorithm. The experimental analysis specifies that the suggested model has shown better efficiency on the identification of blood cell images, and attains high accuracy when compared over the competitive methods.