Towards an automated medical diagnosis system for intestinal parasitosis

International audience Human parasites are a real public health problem in tropical countries, especially in underdeveloped countries. Usually, the medical diagnosis of intestinal parasites is carried out in the laboratory by visual analysis of stools samples using the optical microscope. The parasite recognition is realized by comparing its shape with the known forms. We offer a solution to automate the diagnosis of intestinal parasites through their images obtained from a microscope connected directly to a computer. Our approach exploits the contour detection based on the multi-scale wavelet transform for detecting the parasite. Active contours are combined with the Hough transform to perform image segmentation and extraction of the parasite. We used the principal component analysis for the extraction and reduction of features obtained directly from pixels of the extracted parasite image. Our classification tool is based on the probabilistic neural network. The obtained algorithms were tested on 900 samples of microscopic images of 15 different species of intestinal parasites. The result shows a 100% recognition rate of success.