Automation of Malarial Cell Count and Stage Classification Using Morphological Operations and Variable Optimization Using Hyper-Parameter Tuning

Abstract

Malaria, caused by the parasite plasmodium, infects the cells of the host and multiplies using their resources. One of the best ways to test malaria involves forming blood smears and examining the slides. Malarial cell count detection is a crucial process in classifying the extent of malarial infection. We have devised and tested an algorithm that automates the process of malarial cell count detection and classifies the images according to stage of malaria present. Images from various patients were obtained and tested. We used morphological operations to demarcate malaria infected cells, and hyper-parameter tuning was employed to optimize the algorithm for better accuracy. The proposed algorithm for malarial cell count detection is capable of automating the cell count process and produces accuracy levels greater than 90% in finding the correct count of parasite infected cells in the sample and greater than 90% accuracy in correctly classifying the stage of malaria.