Hybrid CNN-LSTM with Clahe-Enhanced Segmentation for Automated Leukemia Detection

Early and accurate diagnosis of Acute Lymphoblastic Leukemia (ALL) disease plays a vital role. This work introduces an optimal deep learning architecture that includes CLAHE-enhanced thresholding for a better segmentation process. Then the Convolutional Neural Network with Long Short Time Memory (CNN-LSTM) architecture ensures the efficient feature extraction for better prediction. To optimize performance and gather the best parameters, hyperparameter optimization using Cockroach-Based Optimization (CBO) is introduced for adapting optimality. Bitwise AND operations on the segmented images are performed to highlight the regions affected by enhancing feature discrimination. Experimental results obtained on ALL datasets show that the proposed approach is attain a high accuracy of 98.1% with a loss of 0.05 with reduced computational overhead The integration of CLAHE-based segmentation, CNN-LSTM and CBO optimization provides a robust and efficient method for leukemia diagnosis without overfitting and local minima.