A Methodology-Driven Hybrid Inception V3–ResNet152 Framework for Multiclass Leukemia Subtype Classification
Contributors
silambarasi P
Keywords
Proceeding
Track
General Track
License
Copyright (c) 2026 Sustainable Global Societies Initiative

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Abstract
Automated analysis of peripheral blood smear images can support the classification of leukemia subtypes, but limited annotated data and the visual similarity among subtypes make robust multiclass learning challenging. This paper presents a methodology-focused hybrid deep learning framework that combines pretrained Inception V3 and ResNet152 networks for feature extraction and multiclass classification of leukemia. The methodology emphasizes transfer learning, controlled data augmentation, preprocessing standardization, hybrid feature learning, and complexity-aware fine-tuning. The target task covers Acute Lymphoblastic Leukemia (ALL) subtypes L1, L2, and L3 and Acute Myeloid Leukemia (AML) subtypes M2, M3, and M5, using peripheral blood smear imagery described in the source study. Rather than reporting experimental performance, this paper formalizes the complete processing pipeline, model configuration, training strategy, and evaluation protocol required for reproducible experimentation.