Exploring and Improving Deep Learning-Based Image Filtering and Segmentation Techniques for Enhancing Leukemia Images

Pathologists often diagnose leukemia by examining a blood smear under a microscope, which reveals an unusual proliferation of leukocytes in the bone marrow and blood. Pathologists diagnose and classify leukemia based on the numbers and shapes of specific cell types. However, thorough knowledge and patience are required for the morphological inspection of bone marrow cells due to their varied shape. In order to lessen the workload, lower the chance of making mistakes, and increase productivity, an autonomous diagnosis system that makes extensive use of image analysis and pattern recognition technologies is urgently required. Traditional diagnostic techniques take a long time and may be influenced by the doctors’ experience and training. These conventional approaches are time-consuming and may be impacted by the knowledge and ability of the medical experts performing the diagnostic procedures. Methods based on image processing can be used to examine microscopic smear pictures automatically and rapidly in order to diagnose leukemia. In the proposed study, the combined techniques namely filtering and segmentation attempt is made to examine the available works in the field of medical image processing of blood smear pictures, with an emphasis on automated leukemia identification. The existing publications in the relevant field are evaluated in light of the segmentation technique employed. Even if there are numerous studies for the diagnosis of acute leukemia, there is just a handful for the detection of chronic leukemia. There are a limited number of relevant literature reviews, but none of them categorize earlier investigations based on the segmentation method employed.