Morphological Analysis and Diagnosis of Human Leukemic Cells Using Image Processing: A Review

Leukemia is a type of blood cancer in which abnormal and malfunctioning white blood cells (WBC) proliferate rapidly and on a large scale and hence, it must be detected and diagnosed quickly. Manual detection of leukemia is a time-consuming, labor-intensive procedure that requires the assistance of a hematologist and considerably depends on factors such as the hematologist's experience and tiredness causing the results subjected to error and untrustworthy. These issues are addressed by automating the morphological analysis of blood cells, which reduces time, cost, and human error. It does not necessitate costly lab experiments and testing, and hence images are inexpensive, it will be a better solution for hospitals and laboratories. This automation process uses microscopic images obtained from a dataset, the internet, or a lab or hospital, and it may employ many image processing techniques, and so on Automated detection and diagnosis of leukemia emerge as a challenging area which lay the groundwork to many research opportunities. This paper to discuss on morphological analysis in terms of detection, counting and diagnosis of leukemia, a comparative review and identified research gaps on image processing techniques used in state-of-the-art research. In the review, it was revealed that, Morphological/shape-based segmentation has verified as the best method for image segmentation in ALL diagnosis and K-Nearest Neighbor, Support Vector Machines (SVM) and Ensemble of classifiers have resulted better classification results. One major finding is the dependency of datasets in the accuracy of the solutions and hence there are no standard methods and metrics to compare accuracy among various researches. This issue with the dataset throws into question whether the researchers achieved acceptable high accuracy and expected outcomes. Another dilemma is that when they are employed in real-world settings, they still require final verification from medical staff, suggesting that current developments must be fine-tuned for trustworthiness and reliability. Standardized and scrutinized further research into leukemia diagnosis would provide solid remedies to the present challenges with current image processing-based diagnosis methods.