A Neural Network Architecture for Automated Recognition of Intracellular Malaria Parasites in Stained Blood Films

The global burden of malaria is enormous and the development of better laboratory diagnostic tools is a key step in malaria control recommended by the WHO. Our objective was to develop an automated tool for the recognition of intracellular malaria parasites in stained blood films. We used digital images of oil immersion views from microscopic slides captured though a capture card. They were preprocessed by segmentation and grayscale conversion to reduce their dimensionality and later fed into a feed forward backpropagation neural network (NN) for training it. Then a user interface was developed incorporating this trained NN. In the final product, the tool allows a user to view the slide in a graphical user interface. When the user gives a command to analyze, a still image is captured and sent to the NN for recognition after preprocessing. Preliminary results show that the NN can identify carefully selected test data.