Automated Blood Group Detection using Fingerprint

Biometric traits such as fingerprints have long served as reliable indicators for personal identification due to their uniqueness and lifelong stability. However, emerging inter disciplinary research has started exploring the potential of fingerprints beyond identity verification—particularly their correlation with biological attributes like blood groups. This research proposes a novel, non-invasive methodology for predicting an individual's blood group based solely on fingerprint images, eliminating the need for conventional, invasive blood sampling techniques. The approach begins by enhancing fingerprint im-ages using Gabor filtering to extract significant features, including minutiae points, ridge flow, and texture characteristics. These features are then processed using Convolutional Neural Networks (CNNs), leveraging four deep learning architectures: ResNet, VGG16, Alex Net, and LeNet. Each model is trained and evaluated on a labeled dataset of fingerprint images corresponding to known blood groups. The performance of these models is analyzed using key metrics such as accuracy, precision, recall, and F1-score. Among all models, ResNet exhibited superior performance in terms of both accuracy and consistency. This technique offers a cost-effective, painless, and rapid alternative to traditional blood typing, particularly valuable in emergency situations or remote regions with limited access to laboratory facilities.