Red blood cells classification with sharpening segmentation and mask R-CNN

• Main Authors: Arianti, Nunik Destria, Muda, Azah Kamilah, Ahmad, Norashikin
• Format: Conference or Workshop Item
• Language: English
• Published: 2023
• Online Access: http://eprints.utem.edu.my/id/eprint/27878/1/Red%20blood%20cells%20classification%20with%20sharpening%20segmentation%20and%20mask%20R-CNN.pdf; http://eprints.utem.edu.my/id/eprint/27878/; https://link.springer.com/chapter/10.1007/978-3-031-27524-1_63

Identifying and measuring red blood cells (RBC) before prescribing treatment for blood-related disorders is crucial for appropriate diagnosis. A pathologist manually performs such a process under a light microscope, as is customary. Nevertheless, manual visual inspection is arduous and subjective, leading to variance in RBC identification and counting. This paper proposes classification using sharpening segmentation combined with the algorithm mask R-CNN to increase the accuracy of calculating the number of RBCs. Eventually, the RBCs were classified as overlapping or single RBCs using the mask R-CNN classifier algorithm. In this study, a combination of 3 preprocessing image methods, including sharpening with the wand library, clahe, and the Otsu threshold, was used. The classification with the sharpening method proposed in this study gives an average accuracy of 17.4% better than without sharpening. The suggested approach has been evaluated on images of RBC and exhibits a reliable and effective methodology for classifying single and overlapping RBC.