Enhanced Blood Group Prediction with Fingerprint Images using Deep Learning

Determining a person’s blood group is essential in medical diagnostics, emergency care, and forensic investigations. Traditional serological methods for blood group detection are invasive, time-consuming, and require laboratory infrastructure. This project proposes a non-invasive and efficient approach for predicting human blood groups using fingerprint images through deep learning techniques. By leveraging Convolutional Neural Networks (CNNs) such as LeNet, VGG16, AlexNet, and ResNet-34, the system learns complex ridge and minutiae patterns from fingerprint images to classify blood groups (A, B, AB, O) along with Rh factors. The proposed system processes fingerprint inputs through stages of image preprocessing, feature extraction, and classification to deliver accurate predictions. Experimental results indicate that the ResNet-34 architecture achieves superior accuracy, with 95.54% training accuracy and 81.42% validation accuracy, demonstrating the feasibility of fingerprint-based blood group prediction. This work presents a cost-effective, real-time, and non-invasive alternative to conventional methods, with potential applications in healthcare, blood donation management, and forensic science.