Counting and Characterization of Platelets Using Mathematical Morphology

Platelets, medically also termed as “Thrombocytes” are diminutive, achromatic fragments of cell in our vital fluid, blood, that form clots and prevents or stops bleeding. Various techniques and methodologies have been introduced and implemented apart from the conventional microscopic techniques to detect and count them, as platelets are one of the important cells in the bloodstream. The previously introduced techniques cannot determine the irregular and overlapped cells. This project proposes the Otsu Threshold Algorithm as an efficient algorithm that returns an exclusive intensity threshold that separates the pixels into two classes which are named as foreground and background. Mathematical Morphologies like dilation and erosion will help to smoothen the edges of the binary image from the otsu threshold output. For detecting circles or improper shapes, an algorithm named Circular Hough Transformation or CHT using various extraction techniques in digital image processing will be implemented. As a result, this project will classify and enumerate the platelet cells according to their shape and size (circular, non-circular, size, overlapped and negligible) and determine the count of the cells using Circular Hough Transformation (CHT) in raw microscopic images with a classification accuracy of 97%.