Computer Vision Approach for the detection of Thrombocytopenia from Microscopic Blood Smear Images

The counting and analysis of blood cells is the first step in detecting diseases such as COVID-19, dengue, and leukemia. In primary health care centers, platelet counting is usually performed manually from peripheral blood smears, which is labor-intensive and requires an experienced laboratory technician, which invites error. The complete blood count test is carried out using the automatic hematology analyzer with the development of technology. However, this is an expensive method to count the blood cells. This paper introduces an efficient and cost-effective computer vision technique known as blob detection for automatic platelet counting. The technology uses digitized microscopic blood smear images to identify platelets as blobs used to quantify clinically relevant features, i.e., detection of thrombocytopenia, leading to further medical analysis. Detection of thrombocytopenia is imperative as the reduced platelet count may cause small gum bleeds to severe conditions like hemorrhage or even death. The algorithm is developed using a dataset of 50 digitized microscopic blood smear images. The clinically estimated manual platelet count is considered as a reference. The developed algorithm gives an accuracy of about 96.4% for the detection of platelets. The automated count will ensure greater accuracy and allow hematologists to perform faster analysis. This system can be deployed in remote areas as a supporting aid for telemedicine technology.