Blood group detection using fingerprint patterns – a deep learning approach

Accurate and timely determination of blood group in medical treatments, transfusion medicine, and forensic science. Traditional blood typing methods rely on laboratory-based serological testing, which is accurate but invasive, time-consuming, and resource-intensive. This research proposes a non-invasive approach to blood group classification using fingerprint patterns and deep learning techniques. By leveraging the unique ridge patterns in fingerprints, multiple Convolutional Neural Network (CNN) architectures were evaluated including shallow, custom, and deep variants to identify the most effective approach. The dataset which was sourced from Kaggle was preprocessed thoroughly through class balancing, image normalization, and augmentation to enhance the model robustness. Performance was assessed using standard classification metrics, including accuracy, precision, recall, and F1-score. Interestingly, the Shallow CNN outperformed more complex models, achieving the highest accuracy of 89.49\% with significantly lower training time. These findings highlight the potential of well-optimized shallow networks for biometric-based blood typing offering a non-invasive, cost-effective, and scalable alternative to conventional methods. Future work should focus on increasing dataset diversity (e.g., different age groups), transfer learning, and validating the models in real-world clinical settings to improve robustness and generalizability.