Machine Counting of Malaria Infected Blood Cells Using RGB Images

Abstract

The most widely used laboratory confirmation technique for malaria is visual inspection of Giemsa stained blood smears on microscope. A detection and counting method for malaria infected blood cells in a colour (RGB) microscopic image was developed with th e help of machine vision and artificial neural networks (ANN). The developed system is capable of detecting individual blood cells in the image and recognized them as malaria infected or non - infected. The system is capable of producing the number of blood cells in each category, which can be use d as an indicator of severity of infection. The system was trained for 40 blood cells (from seven images) manually marking them as infected or non - infected , and 120 blood cells (from 15 images) were used to test the system. The sensitivity and the specificity of the system for that data set was found to be 90.0 % and 95.7 % respectively for the images of blood cells of malaria infected and uninfected by Plasmodium falciparum parasites.