Timely and accurate blood grouping is essential for emergency care applications, transfusions, and organ transplants. Traditional serology testing is time-consuming, facility-intensive, and obtrusive, notwithstanding its accuracy. This study offers a novel, non-invasive method for identifying blood types by utilizing fingerprint pictures and Convolutional Neural Networks (CNNs). This method looks at morphological fingerprint patterns, such as loops, whorls, and arches, which can be linked to blood groups based on genetic characteristics. The technique makes use of the known relationship between dermatoglyphic patterns and genetic markers that affect blood group determination and fingerprint formation during embryonic development. In order to extract and evaluate minute ridge properties, minutiae points, and pattern classifications that function as biomarkers for blood type identification, our unique CNN architecture was created. Training on an extended fingerprint data set of 6000 photos, which included a variety of preprocessing approaches like noise reduction, contrast enhancement, and geometric changes to mimic real-world imaging settings, significantly improved the model's robustness. To avoid overfitting and preserve feature extraction efficiency, the deep learning system used multiple convolutional layers with batch normalization and dropout regularization. The generalizability of the model across a range of demographic groups was guaranteed via cross- validation procedures. With a 99.65% medical classification accuracy, the system demonstrated its potential as a rapid and precise diagnostic tool in the medical field, providing notable benefits in emergency situations, resource-constrained settings, and point-of-care testing environments where conventional laboratory infrastructure might be unfeasible or impractical.
Nandhini et al. (Wed,) studied this question.