Computer Aided Malarial Diagnosis for JSB Stained White Light Images Using Neural Networks
R. Sriram, Meenalochani Chandar, Kota Srinivas · 2013
This paper focuses on development of sensitive malarial detection system for images of (JSB) stained thick blood slides acquired from conventional light microscopes. Malaria is a life-threatening disease caused by parasites that are transmitted to people through the bites of infected mosquitoes. Light microscopy enables the visualization of malarial parasites in a thick or thin smear of the patient’s blood. Automation of the evaluation process in the diagnosis of malaria is of high importance. The proposed system describes the computerized method of image analysis involving three main phases: pre-processing, where the images are corrected for luminance and transformed to a constant color space. A histogram based image segmentation processing where the maximum artefacts and over stained objects are avoided. Finally, Feature extraction along with a multi-layer, feedforward, backpropagation neural network was employed for classifying the objects as parasite/wbc. The proposed method achieves the 91% of sensitivity, 85% of specificity with positive prediction rate 88%.