Can Fingerprint Patterns Predict ABO and Rh Blood Groups? A Critical Review of Dermatoglyphic Evidence and Deep Learning Evaluation

Abstract

Fingerprints can be captured in seconds and without a needle, so the idea of reading someone\'s ABO and Rh blood group from them is understandably attractive. The harder question is whether it actually works. A pattern that turns up a little more often in one blood group than another, across a whole population, is not the same thing as a reliable prediction for the person sitting in front of a scanner. In this narrative review we look at the observational dermatoglyphic studies, the newer deep learning work, and what the medical machine learning literature says about trustworthy evaluation. Some studies report significant associations; others, of similar size, find nothing useful. We consider why populations might disagree, why testing for an association is a different task from eight-class image classification, and where the usual weak points lie: images from the same person ending up in both training and test sets, demographic confounding, uneven class prevalence, unclear image sources, no independent external testing, poor calibration, and the clinical cost of a wrong label. Our own earlier CNN experiment is treated here as published exploratory evidence and is not re-reported as a new result. We then set out a practical evaluation framework built on verified blood-group labels, subject-disjoint data splits, transparent reporting and prospective validation, all of which should come before any medical use is considered. On the evidence so far, the question deserves more careful, hypothesis-driven work, but fingerprints have not been shown to be a clinically reliable way of determining ABO/Rh status.

Introduction

This review examines whether fingerprint patterns (dermatoglyphics) can be used to predict human blood groups (ABO and Rh) using deep learning techniques, particularly Convolutional Neural Networks (CNNs). Although both fingerprint patterns and blood groups are influenced by genetics, there is currently insufficient evidence to establish a reliable biological relationship between them. The paper critically evaluates existing research, highlights methodological limitations and proposes guidelines for future studies.

1. Review of Existing Studies

Previous observational studies have reported mixed findings:

• Some studies found statistically significant associations between fingerprint patterns and ABO blood groups.

• Other studies found no significant relationship between fingerprint patterns and ABO or Rh blood groups.

• Some research suggests a possible association with ABO groups, while evidence for Rh classification remains weak.

Differences in population characteristics, sample sizes, fingerprint collection methods and data analysis may explain these inconsistent findings. Counting multiple fingerprints from the same person as independent samples can also produce misleading statistical results.

2. Deep Learning and CNN-Based Prediction

CNN models can automatically identify fingerprint features such as ridge patterns, orientation and texture. An earlier study by the authors used 6,000 labelled fingerprint images and reported approximately 76% validation accuracy on 1,197 images for eight blood-group classes (A+, A−, B+, B−, AB+, AB−, O+ and O−).

However, this accuracy does not establish clinical reliability. The model may learn unrelated features such as image backgrounds, scanner differences or repeated images rather than genuine biological patterns. Independent testing on different participants and devices is necessary to establish generalizability.

3. Major Methodological Challenges

The review identifies several important issues:

• Participant overlap: Multiple fingerprints from one individual must remain in the same dataset partition.

• Label reliability: Blood groups should be confirmed through laboratory serological testing.

• Class imbalance: Rare blood groups, particularly Rh-negative groups, require sufficient representation.

• Data leakage: Training and testing datasets must contain different participants.

• External validation: Models should be tested using independent datasets from different locations and scanners.

• Model reliability: Precision, recall, F1-score, balanced accuracy, confusion matrices and confidence intervals should be reported.

The study also emphasizes that explainability techniques, such as saliency maps, cannot independently prove that a CNN has learned biologically meaningful fingerprint features.

4. Proposed Evaluation Framework

The authors recommend a three-stage research approach:

1. Association testing: Examine relationships between fingerprint characteristics and laboratory-confirmed blood groups.

2. Prediction testing: Train CNN models and evaluate them using participant-independent datasets.

3. External validation: Test the final model on a new population using different scanners and clinical settings.

Separate evaluations should be conducted for ABO and RhD classification before combining them into an eight-class prediction system. Negative-control experiments should also be performed to identify whether models are learning background patterns or other irrelevant information.

5. Ethical and Practical Considerations

Fingerprints are sensitive biometric identifiers. Therefore, research must ensure informed consent, privacy protection, secure data storage and controlled access. The system should be described as an experimental fingerprint-based blood-group prediction model rather than a blood-group detection system.

Most importantly, fingerprint-based predictions must never replace laboratory blood grouping or compatibility testing for blood transfusion, transplantation or emergency medical treatment.

Conclusion

The published evidence on fingerprint morphology and ABO/Rh status is mixed. Some observational studies report significant relationships, while others, just as carefully documented, find none. Deep learning makes the hypothesis easier to test with richer image representations, but apparent predictive success may come from subject overlap, site-specific artefacts or population structure rather than from blood-group information. We therefore recommend person-level inference, external replication, transparent baselines and a clear line between experimental prediction and clinical testing. For now, fingerprints are not an established substitute for laboratory blood grouping, and whether they relate to ABO/Rh phenotypes at all remains a question for well-controlled research.

References

[1] World Health Organization, Guidelines and Principles for Safe Blood Transfusion Practice, Module 3: Blood Grouping and Compatibility Testing. Geneva, Switzerland: WHO. [Online]. Available: https://cdn.who.int/media/docs/default-source/blood-transfusion-safety/guidelines-and-principles-for-safe-blood-transfusion-practice-module-3-.pdf [2] S. Chaudhary, S. Deuja, M. Alam, P. Karmacharya, and M. Mondal, “Fingerprints as an alternative method to determine ABO and Rh blood groups,” J. Nepal Med. Assoc., vol. 56, no. 208, pp. 426–431, 2017, doi: 10.31729/jnma.3374. [3] S. Manikandan et al., “Dermatoglyphics and their relationship with blood group: An exploration,” J. Pharm. Bioallied Sci., vol. 11, suppl. 2, pp. S285–S288, 2019, doi: 10.4103/JPBS.JPBS_13_19. [4] A. Thakur, J. Yadav, and G. Tiwari, “Fingerprint patterns in relation to gender and blood group among residents of central Indian district,” Indian J. Forensic Med. Toxicol., vol. 13, no. 3, pp. 73–77, 2019, doi: 10.5958/0973-9130.2019.00170.1. [5] A. Rastogi, M. A. Bashar, and N. A. Sheikh, “Relation of primary fingerprint patterns with gender and blood group: A dermatoglyphic study from a tertiary care institute in eastern India,” Cureus, vol. 15, no. 5, e38459, 2023, doi: 10.7759/cureus.38459. [6] A. Joseph et al., “Relationship of fingerprints with blood group and sex—A comparative study,” J. Pharm. Bioallied Sci., vol. 17, no. 1, pp. 14–18, 2025, doi: 10.4103/jpbs.jpbs_1120_24. [7] S. Bhan, T. S. Singh, S. Sandhu, S. S. Rahman, and M. P. Sarma, “Forensic examination to determine the correlation between fingerprint patterns and blood groups in the population of Assam,” Sci. Rep., vol. 16, art. 10845, 2026, doi: 10.1038/s41598-026-42044-7. [8] P. Rewatkar and M. Chandankhede, “Blood group detection using fingerprint,” Int. J. Adv. Res. Sci. Commun. Technol. (IJARSCT), vol. 6, no. 3, pp. 483–492, Apr. 2026, doi: 10.48175/IJARSCT-32571. [9] A. K. Jain, A. Ross, and S. Prabhakar, “An introduction to biometric recognition,” IEEE Trans. Circuits Syst. Video Technol., vol. 14, no. 1, pp. 4–20, 2004, doi: 10.1109/TCSVT.2003.818349. [10] C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” J. Big Data, vol. 6, art. 60, 2019, doi: 10.1186/s40537-019-0197-0. [11] I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” in Proc. Int. Conf. Learn. Represent. (ICLR), 2019. [Online]. Available: https://openreview.net/forum?id=Bkg6RiCqY7 [12] Rajumavinmar, “Finger Print Based Blood Group Dataset,” Kaggle. [Online]. Available: https://www.kaggle.com/datasets/rajumavinmar/finger-print-based-blood-group-dataset/data

Copyright

Copyright © 2026 Pranjali Rewatkar, Dr. Minakshi Chandankhede. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.