Modeling of Transfer Learning with Convolution Long Short Term Memory based Leukemia Detection and Classification

Abstract

Leukemia is a kind of blood cancer that produces malignant white blood cells (WBCs) in the human body. These abnormal blood cells cause the bone marrow and blood that affect the immune system to be weak. Automatic and precise analysis of Acute Lymphoblastic Leukemia (ALL), is a difficult task. Microscopic blood tests are measured as the foremost process for the analysis of leukemia. Analysis of blood smears is the most general method of determining leukemia, but not the only method. Interventional radiology is a substitute model for the analysis of leukemia. In the past years, deep learning (DL) using Convolutional Neural Networks (CNN) has delivered advanced techniques for image classification issues; but still, there is a gap to increase their efficiency, learning process, and performance. In this view, this study presents a Transfer Learning with Convolution Long Short Term Memory based Leukemia Detection and Classification (TLCLSTM-LDC) technique. The TLCLSTM-LDC technique aims to inspect the microscopic blood images for the identification of leukemia. In the TLCLSTM-LDC technique, the primary stage of image pre-processing using Wiener Filter (WF) based noise eradication and Dynamic Histogram Equalization (DHE) is involved. At the same time, the Watershed Segmentation approach is applied to segment the pre-processed images. For feature extraction, the TLCLSTM-LDC technique exploits the residual network (ResNet50) model. Finally, the classification of leukemia can be accomplished using the CLSTM model and its hyperparameters can be chosen by root mean square propagation (RMSProp) optimizer. A microscopic blood image dataset is used to validate the performance of the TLCLSTM-LDC method. The experimental values stated the better performance of the TLCLSTM-LDC technique compared to existing approaches.