A Non-Invasive Approach to Blood Group Prediction using Fingerprint Biometrics and Swin Transformer

Identification of blood groups is essential in the hospital, emergency and forensic science. Traditional methods involve taking blood samples and laboratory testing, which can be painful, time-consuming and can pose infection risks. Due to these issues, researchers are looking for alternative solutions, such as using biometrics like fingerprints. Studies suggest that fingerprint patterns may be related to human blood groups. This paper presents a non-invasive blood group detection system based on the fingerprint pattern and a deep learning model called Swin Transformer. This system is based on an Android application developed with the Flutter framework, a SecuGen Hamster Pro 20 fingerprint scanner, and a FastAPI server to build the real-time prediction system. A fine-tuned Swin Tiny Transformer model is used to capture and preprocess the fingerprints and analyze with the help of a USB OTG connection. The system can detect the blood types of a person A+, A-, B+, B-, AB+, AB-, O+, and O-. The Swin Transformer is more capable of capturing the details of fingerprint ridges and texture information, which makes the prediction more accurate and reliable than that of traditional CNN models. The experimental results reveals that the proposed system is accurate and efficient for blood group detection and has capability of scalability as a non blood based identification solution.