Recognition and Categorization of Blood Groups by Machine Learning and Image Processing Method

Background. Red blood cells are one of the components of blood. Blood is an important fluid in the human body. Knowing the blood groups is essential in blood transfusion operations, which depend on fixed conditions to avoid fatal errors. The method that is used to determine the blood groups is a traditional method that relies on medical laboratory technicians, as it is subject to human errors. Objective. This paper aims to design and implement a prototype to detect and classify blood groups to avoid human error in blood group detection. The proposed system employs image processing and machine learning algorithms for blood group detection and classification. Methods. The system consists of three stages. First, samples were collected from volunteers. Second, images of the samples were captured using a camera. Third, the images were analyzed using two methods: image proces­sing via MATLAB and machine learning algorithms via Orange, for blood group detection and classification. Results. The accuracy in processing images using the MATLAB program reached 100%, with processing time ranged from 1.5 to 1.6 seconds. Additionally, using machine learning with neural networks in the Orange program, the accuracy was 99.7%, with training times of 13.7 seconds and testing times of 1.2 seconds. Neural networks outperformed other models, as shown in the experimental results. The study concluded that automated blood type detection using image processing and machine learning methods is effective and feasible compared to manual methods. The proposed system outperformed previous studies in terms of accuracy, processing time, training time, and testing time using both methods. Conclusions. This study underscores the urgent need for precise blood type determination before emergency blood transfusions, which currently relies on manual inspection and is susceptible to human errors. These errors have the potential to endanger lives during blood transfusions. The main goal of the research was to develop an approach that combines image processing and machine learning to accurately classify blood groups.