Deep Learning Based Blood Group Detection Using Fingerprint Image

Blood group determination is a critical requirement in healthcare systems, particularly in emergency medical care, blood transfusion services, and patient management. Conventional blood grouping techniques are invasive, time-consuming, and dependent on laboratory facilities, skilled technicians, and chemical reagents, which limits their applicability in urgent or resource-constrained scenarios. This paper proposes a non-invasive, intelligent system for predicting human blood groups using fingerprint images and deep learning methods. Fingerprints are stable biometric traits influenced by genetic factors, similar to blood group inheritance, making them suitable for exploratory physiological analysis.The proposed system employs a multi-stage framework consisting of fingerprint image acquisition, quality validation, preprocessing, feature enhancement, and deep learning–based classification. Image validation is carried out using Laplacian variance to assess clarity, Canny edge detection to measure ridge density, and frequency-domain analysis using the Fast Fourier Transform to verify periodic ridge patterns. Only validated images are forwarded to the classification stage. A residual neural network architecture with transfer learning is used to classify fingerprints into eight major blood group categories. The system is implemented as a web-based application providing real-time predictions and confidence scores. Experimental results indicate encouraging accuracy and reliability. While not a replacement for laboratory testing, the approach demonstrates the potential of biometric-based, non-invasive blood group prediction for preliminary screening and research applications.