Several procedures in medicine require the counting of blood cells. The number of red blood cell (RBC) can be very important to detect and follow the treatment of many diseases like anaemia and leukaemia. The old conventional method for RBC counting under microscope is considered inaccurate and depends on the clinical laboratory technician skill. Currently, the automatic haematology analyser is vastly used for this purpose, but unfortunately is a very expensive machine. On the other hand, image processing approaches appears nowadays as a cost effective alternative to count blood cells. Most of the computer vision approaches proposed to date are based on high-quality microscopy images, so this work explored the analysis of microscope images acquired via mobile devices for blood cell counting. A partnership with Instituto Nacional de Saúde Dr. Ricardo Jorge allowed the image dataset acquisition, more specifically in the blood smears preparation and generation of the ground truth data (manual annotation).
An image processing and analysis module was developed for the recognition and counting of blood cells, more specifically for counting and recognition of red and white blood cells (RBC and WBC, respectively). The images were acquired with a generic microscope using the Skylight smartphone-to-microscope adapter, as well as with the MalariaScope prototype. The developed image processing module is able to count RBC, as well as correctly differentiate and count 2 different groups of WBC (mononuclear and polymorphonuclear).
For any additional information regarding this project, please contact us using the inquiries form.