VGGNet-Cnn based classification of white blood cell leukemia with efficient salp swarm optimization algorithm

Article type: Research Article

Authors: Amutha, S.^{; *}

Affiliations: India

Correspondence: [*] Corresponding author. S. Amutha, E-mail: [email protected].

Abstract: White blood cell (WBC) leukemia is caused by an excess of leukocytes in the bone marrow, and image-based identification of malignant WBCs is important for its detection. This research describes a new hybrid technique for accurate classification of WBC leukemia. To increase the image quality, the preprocessing is done using Contrast Limited Adaptive Histogram Equalization (CLAHE). The images are then segmented using Hidden Markov Random Fields (HMRF). To extract features from WBC images, Visual Geometry Group Network (VGGNet), a powerful Convolutional Neural Network (CNN) architecture, is used After that, an Efficient Salp Swarm Algorithm (ESSA) is used to optimize the extracted features. The proposed method is tested on two Acute Lymphoblastic Leukemia Image Databases, yielding good accuracy of 98.1% for dataset 1 and 98.8% for dataset 2. While enhancing accuracy, the ESSA optimization picked just 1K out of 25K features retrieved with VGGNet. The combination of CNN feature extraction with ESSA feature optimization could be effective for a variety of additional image classification tasks.

Keywords: WBC leukemia, VGGNet-CNN, ALLIDB, efficient scalp swarm algorithm

DOI: 10.3233/JIFS-221302

Journal: Journal of Intelligent & Fuzzy Systems, vol. 44, no. 4, pp. 6973-6989, 2023