- Gazi Mühendislik Bilimleri Dergisi
- Vol: 9 Issue: 1
- Mask R-CNN Based Segmentation and Classification of Blood Smear Images
Mask R-CNN Based Segmentation and Classification of Blood Smear Images
Authors : Hilal Atici, H. Erdinç Kocer
Pages : 128-143
View : 7 | Download : 6
Publication Date : 2023-04-30
Article Type : Research
Abstract :Analysis of microscopic images is a reliable laboratory method that provides useful information in the diagnosis of disease in the health field. Although advanced technology devices provide important information in the diagnosis of blood diseases, microscopic blood smear examination is needed for definitive diagnosis. Today, the microscope is used by technicians in many laboratories and anomalies in cells (defects in the cell, parasites, low or excess cell count, etc.) are detected. The anomalies detected by the experts provide important information in the diagnosis of diseases. Analysis of microscopic images is a time-consuming and error-prone procedure for the expert. Therefore, in this study, a method that accelerates the examination performed by the expert and that can detect cells automatically is proposed. Segmentation and classification of basic blood cells are emphasized. PBC (Peripheral Blood Cell) dataset blood smear images were used as data set. Mask R-CNN architecture, which is a region-based convolutional neural network, was used in the development of the system. Different backbone structures were used and evaluated for Mask R-CNN. The segmentation of blood cells obtained from the images was determined by different colorings thanks to the sample segmentation feature in the Mask R-CNN algorithm, and the error rates were minimized as a result of the tests. The study focused on detecting eight classes, but the study could be improved by enriching it with more classes and using blood cell images from different angles and better segmentation.Keywords : Derin Öğrenme, Segmentasyon, Sınıflandırma, Kan Hücresi