Paper Contents
Abstract
Blood group identification is a critical requirement in medical diagnostics, transfusion safety, and emergency healthcare. Conventional blood typing methods are invasive, time-consuming, and require laboratory facilities. This work proposes a non-invasive approach for predicting human blood groups using fingerprint patterns combined with machine learning(ML) and gray-level(GL) texture analysis. Fingerprint images are preprocessed through normalization and contrast enhancement to improve feature clarity. From these images, texture features such as contrast, correlation, energy, and homogeneity are extracted using Gray Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP). These features are then used to train ML models including Support Vector Machines (SVM), Random Forest, and Logistic Regression. Additionally, a lightweight Convolutional Neural Network (CNN) is employed to automatically learn deep features from fingerprint ridge patterns. A late-fusion strategy integrates the outputs of both GL-based ML models and the CNN to enhance prediction performance. Experiments with fingerprint datasets and stratified cross-validation demonstrate that the hybrid approach achieves higher accuracy compared to individual methods, indicating strong correlations between dermatoglyphic features and blood group categories. This method shows promise as a rapid, cost-effective, and non-invasive alternative for preliminary blood group detection, while highlighting the need for larger, diverse datasets for clinical application. The models were trained on 80% of the dataset, with the remaining 20% reserved for testing and validation health. Blood group detection using fingerprint features achieves an accuracy of 89% with traditional ML models and up to 90% using deep learning approaches.This shows strong potential for a non-invasive, reliable, and rapid identification method
Copyright
Copyright © 2025 Priyanka L, Dr. Kumar Siddamallappa.U, Anusha Jajur.J. This is an open access article distributed under the Creative Commons Attribution License.
