Sickle Cell Disease Identification by Using Region with Convolutional Neural Networks (R-CNN) and Digital Image Processing

Abstract

In Sri Lanka, medical laboratory technicians manually diagnose sickle cell disease by examining microscopic images of strained blood samples on glass slides and calculating the infected blood cells. However, due to the several processes required in manual assessment, the method of analysis is much time consuming and required skilled medical laboratory technicians because there is a possibility of human error. This research was conducted to provide an Artificial Intelligence integrated computerized solution to prevent the above difficulties in the identification process. Deep Learning is the most powerful approach for object detection and classification in many areas, and many researchers proven the deep learning approach is better than the other learning approaches. Deep learning concepts such as Convolutional Neural Networks (CNN) plays a major role in medical image processing diagnostics. In this research, we develop a region classification Convolutional Neural Network (R-CNN) algorithm for the sickle blood cell identification process, which is one of the most demanding processes in blood classification. The proposed identification algorithm was developed and trained by using a dataset of 150 blood smear sample images and validate by using 20 sample images. According to the algorithm validation results, the prosed method achieved an accuracy of more than 90%.