Malaria Parasite Detection using a Polarization Sensitive Microscope and Machine Learning

Amit Narawane      Latifah Maasarani

amn58@duke.edu     lam139@duke.edu

Paper PDF

Malaria is a significant global health burden that disproportionately affects populations in low-resource settings. There is therefore great interest in optimizing microscopy techniques for automated diagnosis of malaria that maintains high classification accuracy while reducing cost and the need for trained personnel. Polarization sensitive microscopy presents an effective and low-cost method to enhance contrast in thin blood smears. In this study, a learned sensing approach was used to build a machine learning model that can both determine the optimal polarization configuration for improving contrast and successfully classify infected versus uninfected red blood cells (RBCs). The results demonstrate the potential of this method for joint physical hardware optimization and automated malaria diagnosis.

Paper: Paper PDF

Code and Data: Zip file of project code: Download Code Here