BLOOD CANCER DETECTION USING IMAGE PROCESSING

Abstract [English]

The conventional technique of diagnosing blood disorders through microscopic visual examination of blood smears is susceptible to error, time-consuming, and based on the physical acuity of the haematologist. Thus, to enable clinical decisions to be made, an optical image processing system has to be automated. One of the features of leukemia is an abnormal growth of immature, faulty white blood cells (WBC), or "blasts." A disease that occurs in blood and/or bone marrow related to white blood cells (WBCs) is referred to as leukemia. Early diagnosis of leukemia that is timely, safe, and accurate is essential for treating and saving patient lives. Typically, WBCs are analyzed using blood smear under microscope. Various machine learning (ML) algorithms have been trained to provide a high misclassification error rate and to detect various diseases, such as leukemia. Therefore, for detecting the microscope images for WBC count study, we can utilize deep learning (DL) approach. Identification module as well as classification module formed 2 modules of WBC differential count system. To differentiate between all the WBCs, red blood cells, colouring impurities, and blood platelets, the identification module first scanned the raw bone marrow smear images. Subsequently, the identified cells were fed into categorization module. Categorization module comprised 2 steps. In initial step, we identified many cells that are not used to diagnose leukemia crushed, degraded, etc. Countable WBCs were forwarded for multi-class differentiation in second step using a convolutional neural network approach.