An Advanced Non-Invasive Blood Group Detection Using Fingerprint Biometrics: A Deep Learning Approach

Abstract

The correct blood group is important in healthcare, forensics and emergency care, but traditional approaches are invasive and can only be applied in laboratory facilities with trained personnel, which is not applicable in emergency or resource-constrained cases. In an effort to overcome these difficulties, this research proposes a superior non-invasive blood group identification system that utilizes the fingerprint biometrics and deep learning. On the basis of the genetic association between fingerprint patterns and blood groups, fingerprint ridges and texture characteristics were examined with Convolutional Neural Networks on eight blood group types (A+, A, B+, B, O+, O, AB+, AB). Different deep learning models such as a custom CNN, MobileNetV2, EfficientNetB0, ResNet50, and a hybrid CNN-GNN were trained and tested, the custom CNN showed the highest performance with a 87% training accuracy and 95% testing accuracy, indicating a high level of generalization and little overfitting. These findings confirm the usefulness of automatic feature extraction that is based on deep learning and can be used to identify blood group reliably on a non-invasive basis with no dependence on hand-crafted biometric features.