Review on Leukemia Detection and Classification Frameworks

Abstract

Technology in this era is moving exponentially high. Adoption of technology in almost every aspect of our lives at all times can be found easily. From a decade, we have been studying the various approaches for the automated detection and classification of Acute Lymphoblastic Leukaemia (ALL) using the different technologies. However none of the techniques seems to be perfect by considering the robustness and reliability of prediction. While working with the leukaemia detection and its classification into L1, L2, and L3 classes the key steps under considerations are the accurate segmentation of overlapping blood cells and appropriate features extraction technique. Recently the deep learning methods gains the significant attentions in which the automated features extracted, however such methods most based on automated features extraction and learning which leads the high dimensional features. The high dimensional features extracted from the raw input image not only lead the higher computation complexity but also may degrade the performance in case of large datasets. The recent deep learning based method is introduced just at initial level in which the key steps are missing such as pre-processing and segmentation and evaluated only on small scale dataset. The performance of such methods affected due to lack of appropriate noise reduction and segmentation techniques.