An improved multiclass classification of acute lymphocytic leukemia using enhanced glowworm swarm optimization

• Published in: Scientific reports Vol. 15; no. 1; pp. 13985 - 20
• Main Authors: N, Saranya, M, Kalamani
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 22.04.2025; Nature Portfolio
• Subjects: 631/67; 639/166; Acute lymphocytic leukemia; Algorithms; Bayes Theorem; Enhanced glowworm swarm optimization; Humanities and Social Sciences; Humans; Multiclass classification; multidisciplinary; Multilevel hierarchical marker-based watershed algorithm; Precursor Cell Lymphoblastic Leukemia-Lymphoma - classification; Precursor Cell Lymphoblastic Leukemia-Lymphoma - diagnosis; Precursor Cell Lymphoblastic Leukemia-Lymphoma - pathology; Science; Science (multidisciplinary); Support Vector Machine; Acute lymphocytic leukemia; Enhanced glowworm swarm optimization; Multiclass classification; Multilevel hierarchical marker-based watershed algorithm
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-98823-1

Acute Lymphoblastic Leukemia (ALL), a kind of blood cancer, more frequently observed in the pediatric population, causes rapid production of immature White Blood Cells. Most of the diagnostic techniques like bone marrow aspiration, imaging techniques, etc. are time consuming, error-prone, costly and depend on the skill set of experts. The ultimate goal of this work is to develop a computer aided automatic classification system to classify Benign, Early, Pro-B and Pre-B classes of ALL. Images from the publicly available dataset were subjected to pre-processing and Region of Interest is obtained by adapting the proposed Multilevel Hierarchical Marker-Based Watershed Algorithm (MHMW). A subset of most vital features were selected by utilizing nature inspired metaheuristic Enhanced Glowworm Swarm Optimization (EGSO) algorithm. Popular classifiers -Decision tree, Random Forest, Multi-Layer Perceptron, Naive Bayes and Linear, Polynomial, Radial basis function, sigmoid kernels of Support Vector Machine were used for multiclass classification. Performance of the proposed system has been compared with three other popular optimization algorithms- Particle Swarm Optimization, Artificial Bee Colony Optimization and Elephant Herd Optimization. Random Forest fed with the optimized features obtained from the proposed integration of MHMW and EGSO algorithms outperformed other classifiers with 98.23%, 98.25%, 98.23% of accuracy, precision and F1 score respectively.