A novel technique for segmenting platelets by k-means clustering

Abstract

Platelet is a major component of various blood cells present in blood that helps in clotting of blood. Platelet count often becomes a crucial diagnostic parameter to identify several diseases like dengue, yellow fever, etc. The traditional process of counting platelets by examining blood slides under a conventional optical microscope is subjected to human errors due to manual inspection. In addition, the overhead on pathologist increases manifold when huge numbers of blood samples are to be tested. In this work, we have developed an Android-based mobile app, which takes as input the microscopic image of blood smear and gives as output the total platelet count present in the image. This system reduces the dependency on expert pathologists and avoids manual errors. A comparative study between platelet counts obtained from expert lab technicians and the one given by our developed app have shown it to be robust and efficient for automated platelet counting.