A Novel Advanced Transfer Learning Based Approach for Leukemia Detection
Navpreet Kaur, Amar Jeet Singh · 2024
Transfer learning has been adopted more widely to automate detection in recent years. It enables trained models to work well on new data with great transfer learning benefits. In this study, novel advanced transfer learning technique is proposed to identify leukemia. An integration of ResNet50, Principal Component Analysis (PCA) and Big Bang Big Crunch (BBBC) optimization algorithm is proposed. The early, accurate detection of leukemia is no small feat in medical imaging-especially when it comes to huge amounts of complex image data. A transfer learning model called ResNet50 which extracts important features of microscopic images of leukemia cells is used. As there are too many features which made it complex to understand and interpret, PCA is used to reduce some dimensionality and selected only the best few parameters for the further analysis. In addition, the BBBC algorithm inspired by theories concerning the expansion and contraction of the universe is finally utilized for the optimization to take place efficiently and acutely. Experiments on a CNMC_2019 dataset of leukemia demonstrated the superiority and accuracy of the proposed method when compared with other classification methods, especially in terms of reliability. The findings indicate that such an approach may enhance the early diagnosis of leukemia, enabling more reliable diagnosis and individualized treatment planning.