Leukemia Prediction from Microscopic Images of Human Blood Cell Using HOG Feature Descriptor and Logistic Regression

Abstract

Leukemia is a cancer of blood which originates in bone marrow causing disruption in the production of human blood cells. Earlier detection of leukemia is crucial due to its fatality. Detection of leukemia involves microscopic observation of human blood cells. Application of various image processing methods along with existing machine learning algorithm is a fast and convenient way to detect leukemia. These methods require extracting features from microscopic images of blood cells and applies machine learning algorithm to train and test a classifier based model which can predict leukemia with an acceptable accuracy. However, scarcity of publicly available image dataset and the inconsistency of the information provided from them make it more challenging while developing a model which can predict leukemia accurately. Besides the size of small datasets and the computational cost, memory evaluation and the required accuracy are also concerning issue. Keeping these drawbacks in mind, we proposed an efficient classifier based model which extract features from image dataset with and classify them accordingly. In this book, we have introduced an edge feature with HOG feature descriptor and Logistic Regression Classifier based model which can detect leukemia with an accuracy of almost 96%.