Blood group classification plays a vital role in medical emergencies, transfusion management, and personalized healthcare. Traditional methods rely on serological testing, which requires invasive sampling and laboratory infrastructure. This study proposes a novel, non-invasive system for predicting human blood groups using fingerprint images. Our solution leverages a hybrid deep learning model combining Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs) to capture both local texture patterns and high-level fingerprint topology. The proposed system consists of a robust preprocessing pipeline that converts grayscale fingerprint images to a normalized format suitable for hybrid feature extraction. The CNN component extracts spatial texture features, while the GNN component analyzes relational structures and ridge connectivity. The final classification is performed through a dense layer that outputs the predicted blood group among the eight major categories (A+, A-, B+, B-, AB+, AB-, O+, O-). To facilitate real-world deployment and ease of use, the system includes optional IoT integration via the R307S fingerprint scanner. Users can either upload fingerprint images through a web interface or scan their fingerprints directly using the R307S scanner connected to a microcontroller (e.g., Arduino). A unified Python backend (predict.py) handles both data paths seamlessly, supporting flexible deployment in both standalone and server-based applications. The dataset comprises over 6000 fingerprint samples across all blood group classes. With 100 training epochs and a batch size of 32, the model achieved a training accuracy of ~70% and demonstrated early signs of generalizability, despite the inherent biological variability in fingerprint patterns across individuals. This project serves as a proof-of-concept that fingerprint morphology can be correlated to blood groups using deep learning, offering a low-cost, non-invasive alternative to traditional blood group testing. It also opens up possibilities for use in rural healthcare, forensic identification, and embedded biometric devices.
Chelli Dileep (Wed,) studied this question.