Integrating Multi-scale Feature Extraction into EfficientNet for Acute Lymphoblastic Leukemia Classification
Fallah H. Najjar · Journal of Image and Graphics · 2025
Acute lymphoblastic leukemia is a cancer of the white blood cell.It originates in the bone marrow, the spongy tissue inside bones responsible for the production of blood.Despite being the most common cancer in children, Acute Lymphoblastic Leukemia (ALL) has remained an enormous health concern.The results of traditional diagnosis, including morphological examination, immunophenotyping, and genetic marker analysis, are relatively slow, subjective, and depend considerably on the ability of a hematopathologist, hence restricting the classification result's consistency.These disadvantages underline the pressing need for automatic diagnostic systems that are fair and satisfactory.This work presents the Multi-Scale Enhanced EfficientNet, which, through several innovative architectures, can increase sensitivity and specificity in accurately identifying subtle ALL variations.We assess the MSEENet's performance using a dataset of numerous ALL phenotypes.We achieve excellent performance in multiple metrics, like an overall accuracy of 98.77%, an accuracy of 98.99%, a recall of 98.49%, a Matthews Correlation Coefficient of 98.34%, and an F1-Score of 98.72%.This research shows the potential for MSEENet as a feasible, precise, and dependable ALL diagnostic tool, further strengthening patient-specific cancer treatment advancements.