Fingergen Biopredictor: Deep Neural Model for Blood Group Classification

Fingergen Biopredictor: Deep Neural Model for Blood Group Classification

Jyothirmai (1)

Dr. M. Veera Kumari , M.Tech, Ph.D (2)

M-Tech Student, Dept.of CSE,UNIVESAL COLLEGE OF ENGINEERING AND TECHNOLOGY,Andhra pradhesh, India (1)

Associate Professor,HOD Dept.of CSE,UNIVERSAL COLLEGE OF ENGINEERING AND TECHNOLOGY,Andhra pradhesh, India (2)

Abstract – Blood group identification is an essential procedure performed before blood transfusion during emergency situations and blood donation processes. It helps ensure that compatible blood is provided to patients during surgeries, accidents, or critical medical conditions. Transfusing incompatible blood can lead to severe complications and may even become life-threatening. Therefore, accurate blood group determination is a crucial step before any transfusion procedure.Traditional blood group testing methods using microscopy are often time-consuming and require skilled laboratory experts for accurate analysis. In many cases, the results may also be difficult to reproduce consistently, especially during emergency situations where rapid decision-making is necessary. Due to these limitations, there is a growing need for an automated and reliable blood group detection system.To address this issue, a blood group detection system based on digital image processing and neural networks is developed. The system analyzes images obtained from slide-based blood testing procedures and automatically identifies the blood group by examining blood agglutination patterns. The acquired images are processed and reused for accurate detection, reducing human error and improving efficiency.By integrating image processing techniques with neural network algorithms, the proposed automated system can provide fast, accurate, and reliable blood group detection. This approach is highly beneficial in emergency medical situations, where quick and error-free blood identification is critical for patient safety and effective treatment.

Keywords – Blood Group Determination, Blood Transfusion, Incompatible Blood Detection, Microscopy, Automation, Image Processing, Neural Network, Blood Agglutination Analysis, Medical Diagnosis, Intelligent Healthcare System.

1. Introduction

   The primary objective of this project is to analyze and classify different types of blood cells, including Red Blood Cells (RBCs), White Blood Cells (WBCs), and platelets, using deep learning and image processing techniques. Red blood cells are responsible for transporting oxygen from the lungs to various parts of the body, while white blood cells help protect the body against infections. Platelets play a major role in blood clotting, and all blood cells are produced in the bone marrow. The proposed system focuses on automatic blood cell classification and counting using neural network-based deep learning models. By analyzing blood sample images, the system can assist in identifying abnormalities associated with diseases such as dengue, malaria, cholera, and other blood-related infections. The neural network model helps improve the accuracy and efficiency of disease detection through automated analysis. Several image processing techniques are incorporated in the system, including preprocessing, Discrete Wavelet Transform (DWT), and Gray Level Co-occurrence Matrix (GLCM) feature

   extraction. These methods help enhance image quality, extract texture information, and improve classification performance. In addition, blob detection techniques are used to identify and classify blood cell structures from microscopic images. The project also considers the analysis of haemoglobin, an essential protein present in red blood cells that carries oxygen throughout the body. By combining deep learning with advanced image processing methods, the system provides an efficient and intelligent approach for blood cell analysis and disease prediction.

   1. Problem Statement:

      Accurate blood group identification plays a vital role in medical emergencies, surgical procedures, and blood transfusion services. Conventional blood group testing methods mainly depend on laboratory-based serological analysis, which requires specialized medical equipment, trained professionals, and significant processing time. In remote or resource-constrained regions, access to such facilities may be limited, resulting in delays in diagnosis and treatment that can negatively affect patient survival and healthcare

      efficiency.Recent developments in Deep Learning and biometric technologies have created new opportunities for fast, intelligent, and non-invasive healthcare solutions. Fingerprints, which are unique to every individual, have been studied for their potential relationship with genetic and physiological characteristics, including blood group patterns. These advancements have encouraged researchers to explore automated systems capable of predicting blood groups using fingerprint images.Despite ongoing research, there is still no universally accepted and highly reliable automated framework that utilizes fingerprint biometrics for blood group classification. Therefore, developing a deep learning-based fingerprint blood group prediction system can provide a rapid, cost- effective, and accessible alternative for preliminary blood group identification, especially in emergency situations and healthcare environments with limited laboratory resources.

   2. Challenges in Blood groups:

      The major challenge in this research is the development of an intelligent and reliable system that can accurately extract and learn complex fingerprint patterns for effective blood group classification. Fingerprint images often contain variations in quality due to factors such as noise, improper finger placement, lighting conditions, low contrast, and sensor limitations, which can affect feature extraction and prediction accuracy. Another significant challenge is the limited availability of large-scale labeled datasets linking fingerprint patterns with blood group information. Insufficient training data can reduce the learning capability and generalization performance of deep learning models. Furthermore, designing a model that consistently achieves high accuracy across diverse fingerprint samples is essential for ensuring practical usability in real-world healthcare applications.Therefore, building a robust system capable of handling image variations, improving feature learning, and maintaining reliable prediction performance remains a critical research challenge in fingerprint-based blood group classification.

   3. Proposed Work Aim:

      The concept of blood groups was first discovered in 1901 by Karl Landsteiner, an Austrian scientist whose work significantly advanced modern blood transfusion practices. Among the various blood classification methods, the ABO blood group system and the Rh factor system are the most widely used for identifying an individuals blood type. The process used to determine blood type is known as blood typing. Blood groups are identified based on the presence or absence of specific antigens located on the surface of red blood cells (RBCs). In the ABO blood group system, there are

      four major blood types: A, B, AB, and O. These blood groups are determined by the combination of antigens present on red blood cells and antibodies found in the blood plasma. Individuals with blood group A possess A antigens on the surface of RBCs and anti- B antibodies in the plasma. Similarly, individuals with blood group B have B antigens and anti-A antibodies. Blood group AB contains both A and B antigens and does not contan antibodies against either antigen, while blood group O lacks both A and B antigens but contains both anti-A and anti-B antibodies in the blood. In addition to the ABO system, the Rh blood group system plays an important role in blood classification. This system is determined by the presence or absence of the Rh D antigen on red blood cells. If the Rh D antigen is present, the individual is classified as Rh-positive; if the antigen is absent, the person is considered Rh-negative. Accurate identification of ABO and Rh blood groups is essential for safe blood transfusions and medical treatments.

   4. Objectives:

      Therefore, this project focuses on the design and implementation of FingerGen BioPredictor, a deep neural network-based system developed for blood group prediction using fingerprint images. The proposed system aims to analyze fingerprint patterns and automatically classify corresponding blood groups through advanced deep learning techniques.The objective of this approach is to provide a rapid, cost-effective, and non-invasive alternative to conventional blood group testing methods. By eliminating the dependency on complex laboratory procedures and specialized equipment, the system can improve accessibility to blood group identification, particularly in emergency situations and resource- limited healthcare environments.Through the integration of biometric analysis and deep learning, the proposed framework enhances diagnostic efficiency, reduces processing time, and supports intelligent healthcare applications with improved accuracy and automation.

   5. Overview of the Paper:

   Deep Neural Networks (DNNs) have emerged as state-of-the-art machine learning models in various domains, including image analysis, computer vision, and natural language processing. Due to their ability to automatically learn complex patterns and representations from large datasets, these models are widely adopted in both academic research and industrial applications.Recent advancements in deep learning have created significant opportunities in the field of healthcare and medical imaging. Deep neural network models are increasingly being applied to medical image processing, disease diagnosis, clinical

   decision support systems, and healthcare data analysis. Their capability to identify subtle patterns in medical images has improved the accuracy and efficiency of diagnostic procedures.In medical image analysis, deep learning techniques are used for tasks such as image classification, segmentation, feature extraction, and abnormality detection. These methods help automate complex medical processes, reduce manual effort, and support healthcare professionals in making faster and more reliable decisions.despite these advancements, several challenges remain in applying machine learning to healthcare applications. Issues such as limited labeled medical datasets, variations in image quality, high computational requirements, data privacy concerns, and the need for interpretable models continue to be important research areas. Nevertheless, deep learning continues to play a transformative role in advancing intelligent medical imaging and healthcare technologies.

2. LITERATURE SURVEY

   The literature survey is one of the most important phases in the software development process. Before developing a system, it is essential to analyze factors such as development time, cost, available resources, and organizational capabilities. After evaluating these factors, suitable technologies, programming languages, and operating systems are selected for implementing the proposed system. During the development process, programmers often require technical guidance and external support, which can be obtained from experienced developers, research papers, books, and online resources. Considering these aspects before implementation helps in building an efficient and reliable system.

   A literature review is a structured analysis of existing research, theoretical concepts, methodologies, and significant findings related to a particular topic. It provides a comprehensive understanding of previous work carried out in the selected research area and helps identify research gaps, challenges, and opportunities for improvement. Literature reviews are considered secondary sources because they summarize and evaluate already published studies rather than presenting new experimental results.

   In academic research, literature reviews are commonly included before the methodology and results sections of a project or thesis. The primary purpose of a literature review is to establish the background of the study, provide context for the research problem, and demonstrate how the proposed work differs from or improves upon existing approaches. It also helps researchers select

   appropriate techniques, tools, and methodologies for developing effective solutions.

3. PROPOSED METHODOLOGY

   The proposed system, FINGERGEN BioPredictor, presents an intelligent blood group classification framework based on Deep Neural Network (DNN) and Convolutional Neural Network (CNN) techniques. The system is designed to predict human blood groups by analyzing fingerprint images through advanced deep learning algorithms. Unlike conventional blood group testing methods that require blood samples and laboratory procedures, the proposed approach utilizes biometric fingerprint data as the primary input for prediction. The fingerprint images are processed, analyzed, and classified automatically by the deep learning model to determine the corresponding blood group. By integrating biometric analysis with artificial intelligence, the system provides a non-invasive, fast, and automated solution for blood group identification. This approach can improve healthcare accessibility, reduce dependency on traditional laboratory testing, and support rapid diagnosis in emergency and resource-limited environments.

   Input: The first step in the proposed system involves capturing or uploading the fingerprint image of the individual. The acquired fingerprint image serves as the input to the system for further preprocessing, feature extraction, and blood group prediction using deep learning techniques.

   Preprocessing:

   The captured or uploaded fingerprint image undergoes a preprocessing stage to enhance its quality and make it suitable for further analysis. This includes image resizing to ensure uniform dimensions, noise removal to eliminate unwanted distortions, and grayscale conversion to simplify the image by focusing on intensity values rather than color information. These preprocessing steps improve feature extraction efficiency and enhance the overall accuracy of the deep learning model.

   Feature Extraction:

   After preprocessing, important fingerprint characteristics are extracted for analysis. These include ridge patterns, which represent the overall flow and structure of fingerprint lines; minutiae points, such as ridge endings and bifurcations that provide unique identity features; and texture features, which capture the local variations and surface patterns within the fingerprint. These extracted features play a crucial role in improving the accuracy of the deep learning model for blood group classification.

   Deep Learning Model:

   A Convolutional Neural Network (CNN) is then applied to learn and extract complex patterns from the fingerprint images. The CNN automatically identifies hierarchical features such as edges, ridges, textures, and fine-grained fingerprint structures without requiring manual feature engineering. By learning these deep representations, the model effectively captures the distinctive characteristics of each fingerprint, which helps in improving the accuracy of blood group classification.

   Classification:

   Finally, the trained deep learning model predicts the blood group based on the extracted fingerprint features. The system classifies the input into one of the ABO blood groups (A, B, AB, or O) along with the Rh factor (Rh positive or Rh negative). This automated prediction enables fast, non-invasive, and efficient blood group identification, making it suitable for real-time healthcare and emergency applications.

   Output:

   The final stage of the system involves displaying the predicted blood group to the user. Once the model completes classification, the output resultsuch as A, B, AB, O along with Rh positive or Rh negativeis shown on the interface. This provides a clear and user- friendly visualization of the prediction, enabling quick interpretation and practical use in healthcare and emergency situations.

   A. Algrotihm:

   Input: Fingerprint image

   Format: 150×150×3150 \times 150 \times 3150×150×3 , Preprocessed and normalized

   Output:Set of finger print images y^

   {A+,A,AB+,AB,B+,B,O+,O},

   P=[0.01,0.02,0.90,…]

   Prediction = AB+

   CONVOLUTIONAL NEURAL NETWORKS ALGORITHM SUDO CODE:

   BEGIN

   1. LOAD DATASET

      • Load fingerprint images from dataset folder

      • Assign labels (A, B, AB, O)

        1. BUILD CNN MODEL

           INITIALIZE model

           ADD Convolution Layer (Conv2D)

           • Filters: 32

           • Kernel size: 3×3

           • Activation: ReLU

             ADD MaxPooling Layer

           • Pool size: 2×2

             ADD Convolution Layer (Conv2D)

           • Filters: 64

           • Kernel size: 3×3

           • Activation: ReLU

             ADD MaxPooling Layer

           • Pool size: 2×2

             ADD Convolution Layer (Conv2D)

           • Filters: 128

           • Kernel size: 3×3

           • Activation: ReLU

             ADD MaxPooling Layer

           • Pool size: 2×2

        FLATTEN feature maps into 1D vector END

        1: Local Training on Local Server 2: For each client k:

        3: Initialize local model W0kW_0^kW0k 4: For each round t=1to T:

        5: For each batch (, ) :

        6: Forward pass through CNN

        7: Convolutional Feature Extraction

        8: Conv(64, 3×3) + ReLU

        9: Conv(64, 3×3) + ReLU

        10: MaxPool(2×2)

        11: Conv(128, 3×3) + ReLU

        12: Conv(128, 3×3) + ReLU

        13: MaxPool(2×2)

        14: Conv(256, 3×3) + ReLU

        15: Conv(256, 3×3) + ReLU

        16: MaxPool(2×2)

        17: Fully Connected Layers

        18: Fully Connected (4096) + ReLU

        19: Fully Connected (Forgery/Authentic) + Softmax 20: Compute loss: () = (, ^)

        21: Compute gradients: ()()

        22: Update local weights: = ()

   2. PREPROCESS DATA

      FOR each image IN dataset:

      • Resize image to fixed size (e.g., 128×128)

      • Convert to grayscale (if needed)

      • Normalize pixel values (0 to 1) END FOR

   3. SPLIT DATASET

      • Divide dataset into Training set and Testing set

      • (e.g., 80% training, 20% testing)

   End For

   System Architecture:

   +1

   Fig 01:system architecture

4. RESULTS AND DISCUSSIONS

A. Data Collection and Preprocessing

1. Data Collection

Data collection is a fundamental step in the development of the FingerGen BioPredictor system, as the performance of the deep learning model largely depends on the quality, diversity, and accuracy of the dataset. A well-structured dataset ensures better learning, generalization, and prediction capability of the model.

Sources of Data:

Fingerprint images can be obtained from publicly available biometric databases such as FVC (Fingerprint Verification Competition) datasets. In addition, medical or research-oriented datasets may be utilized if available. For improved accuracy and real-world applicability, custom data collection can also be performed using fingerprint scanners or mobile-based biometric sensors. The collected fingerprint images are then manually labeled with their corresponding blood groups for supervised learning.

Data Requirements:

The dataset should consist of high-resolution fingerprint images with clear ridge details. It is important to maintain a balanced distribution across all blood group categories, including A+, A, B+, B, AB+, AB, O+, and O. Adequate samples for each class are necessary to prevent bias and improve model performance.

Challenges in Data Collection:

One of the major challenges is the limited availability of datasets that directly link fingerprint patterns with blood group information. Additionally, variations in fingerprint quality due to noise, scanning conditions, and user input can affect consistency. Ethical and privacy concerns related to biometric data collection must also be carefully

addressed to ensure secure and responsible handling of sensitive information.

Prepossessing

Preprocessing plays a crucial role in ensuring that the input fingerprint data is clean, consistent, and suitable for training the CNN model. It improves the quality of the dataset and enhances the overall performance of the system.

Step 1: Image Resizing

In this step, all fingerprint images are resized to a fixed dimension of 150 × 150 pixels (I 150 × 150). Standardizing the image size ensures uniformity across the dataset, which is essential for efficient CNN processing. It also helps in reducing computational complexity and improves the speed and stability of model training.

Data Augmentation

To enhance the models generalization capability, data augmentation techniques are applied to the fingerprint images. These transformations help simulate real-world variations and improve the robustness of the CNN model.

The augmentation process includes rotation within a range of ±10° to

±30°, horizontal and vertical flipping, as well as zooming and shifting of images. These modifications create diverse versions of the same fingerprint samples, effectively expanding the dataset size.

As a result, data augmentation helps in reducing overfitting, improving model stability, and enabling the system to perform more accurately on unseen data.

C. Model Architecture CNN

D. Model Evaluation and Deployment

Table 1. Performance comparison of various classification models with proposed Model for blood group using fingerprint.

.V. CONCLUSION

This paper focuses on the classification and counting of blood cells using deep learning techniques implemented on the MATLAB platform. The proposed approach leverages wavelet transform-based image analysis to extract meaningful features from microscopic blood cell images. Due to its multiresolution capability, wavelet transform allows efficient decomposition of images at different scales, making it suitable for identifying detailed structures in biological images.The method utilizes different wavelet functions such as Haar and Daubechies wavelets, which are selected based on the specific characteristics of the problem. These wavelets help in detecting image segment features and analyzing boundary signals for effective classification. In addition, two- dimensional wavelet transforms can be applied to capture texture patterns of individual image segments, improving the accuracy of feature extraction and classification.Although the approach was initially applied to analyze microscopic crystal structures, it demonstrates strong applicability in broader domains. These include biomedical image processing, satellite image analysis, communication systems, and remote sensing applications. The study highlights the versatility of wavelet-based techniques in solving complex texture analysis problems across multiple interdisciplinary fields.

VI SCOPE FOR FUTURE WORK:

The proposed work on blood cell classification and counting using deep learning and wavelet-based image processing opens several opportunities for further enhancement and research. In future

developments, the system can be improved by integrating advanced deep learning architectures such as CNN-LSTM or attention-based models to achieve higher accuracy and better feature learning from complex microscopic images. The methodology can also be extended by combining wavelet transform techniques with modern deep learning frameworks to create hybrid models that improve robustness in feature extraction and classification. Additionally, implementing real-time blood cell analysis using optimized algorithms can support faster medical diagnosis in clinical environments. Future research can focus on expanding the dataset with more diverse and high-quality microscopic images to improve model generalization across different conditions. The system can also be enhanced to detect a wider range of blood-related disorders beyond basic classification and counting, such as anemia severity, infection levels, and other hematological abnormalities. Moreover, deploying the model on cloud-based platforms or edge devices can enable remote healthcare monitoring and support hospitals with limited computational resources. The approach can also be adapted for other interdisciplinary applications such as medical imaging, satellite image analysis, and texture-based pattern recognition, demonstrating its wide scope in scientific and industrial domains.

VI REFERENCES

1. G. C. C. Lim, Overview of Cancer in Malaysia, Japanese Journal of Clinical Oncology, Department of Radiotherapy and Oncology, Hospital Kuala Lumpur, 2022.

2. T. R. Golub, D. K. Slonim, P. Tamayo, et al., Molecular Classification of Cancer: Class Discovery and Class Prediction by Gene Expression Monitoring, Science, vol. 286, no. 5439, pp. 531 537, 1999.

3. F. Scotti, Automatic Morphological Analysis for Acute Leukemia Identification in Peripheral Blood Microscope Images, in Proceedings of CIMSA, 2005, pp. 96101.

4. Q. Liao and Y. Deng, An Accurate Segmentation Method for White Blood Cell Images, in IEEE International Symposium on Biomedical Imaging, Atlanta, USA, 2022, pp. 245248.

5. P. Bamford and B. Lovell, Method for Accurate Unsupervised Cell Nucleus Segmentation, in IEEE Engineering in Medicine and Biology Society Conference, 2022, vol. 3, pp. 27042708.

6. N. Sinha and A. G. Ramakrishnan, Blood Cell Segmentation Using EM Algorithm, in 3rd Indian Conference on Computer Vision, Graphics and Image Processing, 2022, pp. 445450.

7. R. D. Labati, V. Piuri, and F. Scotti, ALL-IDB: The Acute Lymphoblastic Leukemia Image Database for Image Processing, in Proceedings of IEEE International Conference on Image Processing (ICIP), Brussels, Belgium, 2021, pp. 20452048.

8. M. Sezgin and B. Sankur, Survey over Image Thresholding Techniques and Quantitative Performance Evaluation, Journal of Electronic Imaging, vol. 13, no. 1, pp. 146165, 2025.

9. K. Nallaperumal and K. Krishnaveni, Watershed Segmentation of Cervical Images Using Multiscale Morphological Gradient and HSI Color Space, International Journal of Imaging Science and Engineering, vol. 2, no. 2, pp. 212216, 2023.

10. R. Adollah, M. Mashor, N. Nasir, H. Rosline, H. Mahsin, and H. Adilah, Blood Cell Image Segmentation: A Review, in IFMBE Proceedings, Springer-Verlag, Berlin, 2023.

11. F. Scotti, Robust Segmentation and Measurement Techniques of White Cells in Blood Microscope Images, in IEEE Conference on Instrumentation and Measurement Technology, 2006, pp. 4348.

12. C. C. Chang and C. J. Lin, LIBSVM: A Library for Support Vector Machines, ACM Transactions on Intelligent Systems and Technology, vol. 2, no. 3, p. 27, 2024.

13. S. Mohapatra, D. Patra, and S. Satpathi, Image Analysis of Blood Microscopic Images for Acute Leukemia Detection, in IECR Proceedings, 2024, pp. 215219.

14. S. Mohapatra, S. Samanta, D. Patra, and S. Satpathi, Fuzzy-Based Blood Image Segmentation for Automated Leukemia Detection, in ICDeCom Proceedings, 2024, pp. 15.

15. S. Mohapatra, D. Patra, and S. Satpathi, Automated Cell Nucleus Segmentation and Acute Leukemia Detection in Blood Microscopic Images, in ICSMB Proceedings, 2024, pp. 4954.

16. R. Rangayyan, Biomedical Image Analysis, Biomedical Engineering Series, CRC Press, Boca Raton, 2025.

17. R. Walvick, K. Patel, S. Patwardhan, and A. Dhawan, Classification of Melanoma Using Wavelet-Transform-Based Optimal Feature Set, in SPIE Medical Imaging: Image Processing, vol. 5370, pp. 944951, 2025.

FIRST AUTHORS:

JYOTHIRMAI Completed B-Tech In Computer Science And Engineering and pursuing her M.Tech in Computer Science And Engineering in Universal College Of Engineering And Technology.

SECOND AUTHOR:

Dr. M. VEERA KUMARI,M.Tech, Ph.D and B.Tech received her all degrees in Computer Science And Engineering. She is currently working as Associate Professor and HOD of CSE in Universal College Of Engineering And Technology.