Authors: Nalla Bhargavi, Mangiri Devika, K Harika, O Rupa Tejasree, Mrs. Archana S, Mr. Pandreti Praveen, Dr. R. Karunia Krishnapriya, Mr. V. Shaik Mohammad Shahil
The “Blood Group Detection using Vision Transformer” project focuses on creating a novel system based on thumb impressions to detect blood groups. The system increases accuracy and efficiency in detecting blood groups by utilizing deep learning methods, i.e., Convolutional Neural network (CNN) variants like ResNet integrated with Recurrent Neural Network (RNN) and MobileNet and Vision Transformer. The user Interface has made this project more interactive by using Flask in the backend. In the evaluation of model performance across different architectures, the results reveal distinct levels of accuracy. The MobileNet achieved an accuracy of 75.04% on the training set and 77.56% on the validation set, demonstrating a solid performance with a loss of 0.6431 and 0.5700, respectively. The ResNet combined with an RNN exhibited lower accuracy, achieving 61.45% on the training data and 75.57% on validation, with corresponding losses of 1.0035 and 0.6754, The Vision Transformer outperformed all models, reaching an impressive accuracy of 97.84% on the training set and 92.52% on validation, accompanied by a loss of 0.0673 and 0.2618.
Introduction
1. Introduction
Traditional blood typing methods are invasive, time-consuming, and unsuitable for emergencies. This study proposes a non-invasive, fingerprint-based blood group detection system using deep learning models (CNNs, ResNet, RNN, MobileNet, Vision Transformer) to improve speed, accuracy, and scalability in healthcare.
2. Fingerprint Patterns
Fingerprint types—loops, whorls, and arches—are analyzed, as these dermal ridge patterns show correlations with blood groups. Biometric systems leverage these unique patterns for individual identification.
3. Literature Review
Prior work shows:
CNNs and hybrid models (CNN + LSTM/RNN) effectively predict blood groups from fingerprint images with accuracies up to 95%.
Traditional ML or image processing methods have lower accuracy and scalability.
Studies confirm moderate correlation (~70-75%) between fingerprint types and blood groups, especially for O+ and A+ (loops) and B/AB (arches).
4. Methodology
The proposed system uses a hybrid of ResNet and RNN, MobileNet, and Vision Transformer (ViT) for classifying blood groups from fingerprint images.
Steps Involved:
Data Collection: Grayscale, high-resolution fingerprint images linked with 8 major blood groups.
Preprocessing: Image enhancement and deep feature extraction.
Model Training:
ResNet + RNN: Combines spatial and sequential pattern learning.
MobileNet: Lightweight, efficient for real-time clinical use.
Vision Transformer: Captures global context from image patches.
5. Results & Model Comparison
All models were evaluated with a 70/30 train-test split. Results include:
Model
Training Accuracy
Validation Accuracy
Training Loss
Validation Loss
CNN
96.98%
89.25%
0.1119
0.2825
MobileNet
75.04%
77.56%
0.6431
0.5700
ResNet + RNN
61.54%
75.57%
1.0035
0.6754
Vision Transformer
97.84%
92.52%
0.0673
0.2618
ViT showed the best overall performance and generalization.
MobileNet was efficient and suitable for low-resource environments.
ResNet+RNN demonstrated strong generalization despite lower training accuracy.
CNN offered high accuracy but required careful regularization to avoid overfitting.
Conclusion
This research presents a non-surgical methodblood group categorization based on fingerprint patterns that does not require surgical interventions or invasive tools. Utilizing sophisticated deep learning algorithms such as MobileNet, ResNet with RNN, and vision transformer, the approach guarantees high-quality analysis. It solves the shortcomings of conventional blood typing by minimizing invasiveness and inaccuracies due to human error, thus improving reliability. Moreover, the method allows for quick blood group identification, which makes it extremely relevant for emergency health conditions. Apart from immediate uses, this innovation creates new frontiers for study, promoting the incorporation of biometric systems into healthcare and eventually enhancing patient care.
References
[1] PAT?L, V., & Ingle, D. R. (2022). A novel approach for ABO blood group prediction using fingerprint through optimized convolutional neural network. International Journal of Intelligent Systems and Applications in Engineering, 10(1), 60-68.
[2] Nihar, T., Yeswanth, K., & Prabhakar, K. (2024). Blood group determination using fingerprint. In MATEC Web of Conferences (Vol. 392, p. 01069). EDP Sciences.
[3] Mahmood, M. F. (2024). Recognition and categorization of blood groups by machine learning and image processing method.
[4] Wang, Y., & Cao, Y. (2020). Human peripheral blood leukocyte classification method based on convolutional neural network and data augmentation. Medical physics, 47(1), 142-151.
[5] Pradhan, A. K., Das, K., Mishra, D., & Chithaluru, P. (2023). Optimizing CNN?LSTM hybrid classifier using HCA for biomedical image classification. Expert Systems, 40(5), e13235.
[6] AMRANE, L., & CHEBOUAT, D. A. BLOOD GROUP PREDICTION USING DEEP LEARNING (Doctoral dissertation, UNIVERSITY OF KASDI MERBAH OUARGLA).
[7] Dargan, S., & Kumar, M. (2020). A comprehensive survey on the biometric recognition systems based on physiological and behavioral modalities. Expert Systems with Applications, 143, 113114.
[8] Das, R., Piciucco, E., Maiorana, E., & Campisi, P. (2018). Convolutional neural network for finger-vein-based biometric identification. IEEE Transactions on Information Forensics and Security, 14(2), 360-37
[9] Kirola, M., Memoria, M., Dumka, A., & Joshi, K. (2022). A comprehensive review study on: optimized data mining, machine learning and deep learning techniques for breast cancer prediction in big data context. Biomedical and Pharmacology Journal, 15(1), 13-2
[10] Nihar, T., Yeswanth, K., & Prabhakar, K. (2024). Blood group determination using fingerprint. In MATEC Web of Conferences (Vol. 392, p. 01069). EDP Sciences
[11] Fayrouz, I. N. E., Farida, N., & Irshad, A. H. (2012). Relation between fingerprints and different blood groups. Journal of forensic and legal medicine, 19(1), 18-21
[12] Patil, V., & Ingle, D. R. (2021). An association between fingerprint patterns with blood group and lifestyle based diseases: a review. Artificial intelligence review, 54(3), 1803-1839.
[13] Ahmad, Z., & Karmakar, R. N. (2014). A medicolegal study to evaluate relationship between fingerprint pattern and ABO blood groups. Indian Journal of Forensic Medicine & Toxicology, 8(2), 208-211
[14] Al Habsi, T., Al Khabori, H., Al Qasmi, S., Al Habsi, T., Al Mushaiqri, M., Das, S., & Sirasanagandla, S. R. (2023). The association between fingerprint patterns and blood groups in the Omani population. Arab Gulf Journal of Scientific Research, 41(3), 283-292
[15] Vijaykumar, P. N., & Ingle, D. R. (2021, April). A novel approach to predict blood group using fingerprint map reading. In 2021 6th International Conference for Convergence in Technology (I2CT) (pp. 1-7). IEEE.