2023 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES) | 979-8-3503-1920-0/23/$31.

00 ©2023 IEEE | DOI: 10.1109/ICSES60034.2023.10465448

A Systematic Procedure to Identify Human Blood

Groups by using Image Processing Assisted
Learning Principle
1
Anita Titus, 2K.Mekala Devi, 3G. Divya, 4Padmaja Nimmagadda, 5S. Vijay Shankar
1
Professor, Department of ECE, Jeppiaar University, Chennai.
2
Assistant Professor, Department of ECE,S.A. Engineering College, Chennai.
3
Assistant Professor, Department of Networking and Communications, Faculty of Engineering and Technology, SRM Institute of
Science and Technology, Kattankulathur, Chennai
4
Professor, Department of Electronics and Communication Engineering,Mohan Babu University, (Erstwhile Sree Vidyanikethan
Engineering College), Tirupati, Andhra Pradesh
5
Associate Professor, Department of Electrical and Electronics Engineering,Sona College of Technology,Salem

E-mail : [email protected], [email protected], [email protected], [email protected],

[email protected]

Abstract- In the occurrence of a life-threatening situation count of platelets, to diagnose a variety of diseases,
needing the transfer of blood, it is crucial to first determine including anemia, leukemia, immune system problems,
the patient's blood type. Currently, technicians execute these parasite and bacterial infections, as well as a great deal of
checks manually, which might introduce room for error. other conditions [2]. There are a few tests that must be run
Rapid and accurate blood type determination, free of human
error, is crucial. If we use AI methods like artificial
before the blood transfusion. Blood type analysis is one of
intelligence and deep learning to determine a person's blood these examinations [3]. There are several emergencies
type, we can minimize the little margin of error introduced where giving blood right away is required to avoid
by human calculations and outcomes. As the field of endangering the patient's life. The currently available tests
computer vision continues to advance at a rapid pace, we are call for changing the laboratory, may not provide enough
able to provide the highest quality service to our clients. time to identify the blood type, and deliver blood type O
Schematic Learning based Human Blood Group Analyzer negative, which is considered a universal donor and hence
(SLHBGA) is a unique deep learning based image has a lower chance of incompatibility [4]. The following
processing technology created to determine the human blood figure Fig.1 illustrates the flow diagram of human blood
types. Methods from the field of image processing, such as
filtering and structural processes, are employed. The blood
group identification
group may be determined rapidly and precisely with this
method. The proposed approach of SLHBGA provides
better efficiency in results, which will be shown clearly with
proper comparisons of conventional learning assisted model
called Convolutional Neural Network (CNN) Model. The
following resultssection then offers the necessary evidence to
demonstrate the effectiveness of the suggested method in
terms of following metrics such as: blood group
identification accuracy, blood group classification timeline,
reduced loss-ratio, sensitivity and specificity.

Index Terms—Classification, Deep Learning, Schematic

Learning, Human Blood Group, SLHBGA, Image Processing,
CNN.

I. INTRODUCTION

Blood tests are commonly used to evaluate an individual's

health [1]. The distinction of platelet types is one of the
most crucial elements of direct blood study. A whole
blood calculation is essentially a compilation of those tiny
molecule components, and information clinicians most
frequently use it as part of a blood examination.
Clinicians can use full blood counts, especially white Fig.1 Flow Diagram of Human Blood Group Identification

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:17:55 UTC from IEEE Xplore. Restrictions apply.
 Immunophenotyping [5] is the process of classifying lowers the danger of transfusion responses, and stores the
blood according to whether or not it contains inherited results without human error [12].
antigenic substances on the surface of the blood-cells,
frequently referred to as blood kinds or blood groups. The assaying of particular red blood cell agglutination in
Antigens such as these could consist of sugars, proteins, response to the addition of of monoclonal antibodies onto
glycoproteins, or glycolipids, depending on the blood whole blood samples for testing is now possible
group system. The antigens and antibodies already present automatically [13]. The technique may be applied to
in the blood are what are utilized to determine the blood identify blood types. The technique is based on Tsoliclone
type. An antigen is any substance that stimulates the monoclonal antibodies applied to blood samples and
immune system to respond by producing antibodies. laying drop photometry. Better cell agglutination as well
Antigens include infectious microorganisms like bacteria, as redistribution across the volume of the sample are
viruses, or fungi. Antibodies, also known as encouraged by mechanical shaking of drop samples.
immunoglobulin, are a kind of protein produced by the Experimental research showed that samples with as well
immune system to fight against antigens. There is a wide as without agglutination had significantly different
variety of blood groups. However, the two most frequent photometric values. This technique might serve as the
types of blood groups are the ABO blood system and the foundation for the creation of a portable automated blood-
Rhesus blood system [6][7]. typing system [13].

II.RELATED STUDY The primary purpose of this study [14] is to employ image
processing to identify different blood types. In the past,
Prior to blood donations or transfusions, especially in we often had blood tests performed to determine our
emergency cases, this is a must [11]. Medical blood type and hemoglobin level. The initial step in
technologists now carry out the tests manually in the labs. treating an illness and the first step in a doctor's
Manual blood typing can occasionally be subject to prescription are both blood type determination. Any
human mistake, which can lead to improper blood emergency scenario calls for it. On the basis of image
grouping and incorrect type in the report, which can cause processing technology, we propose the target as well as
deadly transfusion responses. The purpose of this research identification of blood types. Most of the time, the human
[11] was to develop a method for quickly and accurately eye misses a few crucial details and doesn't provide an
identifying a person's blood type using a combination of accurate result, but with the help of this technology, even
image processing and machine learning techniques. In little mistakes or inaccurate results may be computed. in
order to detect eight different blood types [11] entailed order to lessen human error and outcomes from humans.
the construction of a capture box linked with a web Blood type determination must be completed quickly and
camera system. In this study, we used colour and texture without human mistake. On the basis of processing the
attribute feature extraction, image segmentation, and a photos captured during the slide test, a technique is
number of machine learning techniques [11]. Following devised. The morphological procedures and thresholding
training, the findings revealed that coarse tree DT, schemes are used as image processing techniques. The
employing 70:30 holdout validation, had the greatest pathology laboratory provides the slide test pictures,
performance accuracy score with a score of 97.77%. which are then processed and the presence of
According to testing findings verified by a certified agglutination eliminated [14]. As a result, the automated
medical technologist, the system is completely accurate system that has been created uses image processing to
[11]. identify the blood type. To diagnose the blood type
without human mistake in an emergency, the established
Determine the patient's blood type before administering a approach is helpful [14].
blood transplant is necessary over a critical situation. [12].
These tests are currently carried out manually by Developing countries, in particular, have a pressing need
laboratory professionals, which are tedious and can result for a low-cost Blood Group measurement technique [15].
in mistakes [12] when dealing with a lot of samples. The Image Processing, the most ubiquitous tool in both
suggested system attempts to provide an embedded abundant in resources and limited by resources
system that uses an algorithm for image processing to circumstances, might be put to excellent use in the
conduct blood tests using Rh and ABO blood type construction of this solution. This study proposes a non-
systems. The suggested method reduces the need for invasive technique for determining a person's blood type.
human interaction and conducts the whole test Data collection settings, bio-signal processing techniques,
automatically, from the addition of antigens through the theoretical foundations, the image frequency signal and
creation of the results. The suggested system seeks to feature extraction approach, computational algorithms and
provide results as quickly and accurately as feasible while detection models are all compared and contrasted for their
also storing the data for future use as references. As a impact on the final blood group computation. The
consequence, the technology eliminates traditional research findings were then used to provide realistic
transfusions depending on the idea of a universal donor, strategies for creating image-processing-based safe point-

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:17:55 UTC from IEEE Xplore. Restrictions apply.
 of-care equipment for blood type diagnosis. With the goal
of providing data gathering methods, signal extraction
procedures, feature calculation procedures, and
algorithms for image processing for constructing the
noninvasive Blood Group estimate utilizing an image
[15], ways for blood group measurement are proposed in
this research.

III.METHODOLOGY
The Schematic Learning based Human Blood Group
Analyzer (SLHBGA) is addressed in this study employing
image-processing techniques that lab personnel or a first- Fig.3 RGB to HSV Image Conversion
time user with no knowledge with blood type
identification procedures can utilize. All that is required is (ii) Binary Conversion: The HSV picture is then
to appropriately mix anti-serum with the blood onto the transformed into a binary image. When converting a
white plate before taking a photograph. The technology grayscale image to a binary image, we calculate a global
will be enabled to evaluate the image and provide the threshold level.
precise blood group as the final outcome in no time. The (iii) Structural Opening Operation: To fix any flaws
following are the primary phases in this application: and improve the binary picture for future processing, we
performed some simple operations on it.
A. Image Acquisition
The accompanying figure, Fig.2, which was created using C. Segmentation and Blood Group Detection
test slide images captured with a digital camera. We can simply crop the region of interest using this
Antiserum is combined with a blood sample on slides. knowledge about the region's starting and ending
coordinates, and we may segment our processed picture
for further processing. Two characteristics from the
segmented picture are used for blood group detection and
are as follows:
(i) By first determining the white pixel density in each
divided section.
(ii) The entire number of objects within each segmented
picture is then determined.

IV.RESULTS AND DISCUSSIONS

These days, microscopic examination is used to determine
a person's blood type. Expert counsel is necessary;
Fig.2 Digital Image Sample however mistakes made by humans may be mitigated by
doing the same procedure using image processing at high
B. Image Pre-Processing speed with no room for misunderstanding. Currently,
The image is resized as part of the preprocessing stage to antigens are added to three blood samples taken from a
put it in a certain format. It entails changing an image's patient. Agglutination is possible in the future. After the
RGB into HSV, HSV to grayscale, grayscale to a binary agglutination process is complete, the slide image is
picture, and finally segmenting the processed image into photographed and subjected to additional processing
areas for further processing. The few modifications that using image analysis software. Errors caused by humans
are taken into consideration are as follows: can be minimized with this system. Pre-processing,
feature extraction, and classification are some of the
(i) HSV Conversion: Based on color attributes like image processing techniques employed in this blood type
purity, family, as well as intensity like tint, shade, and detection. Blood group detection is one of several
tone, HSV conversion is used. The conversion process is procedures that must be done before a blood transfusion
shown in the following figure Fig.3, which provides a may take place to ensure the safety of the procedure by
more accurate picture of segmentation and moves us ensuring that the recipient's blood type is suitable with the
closer to our objective. blood type of the donor. Blood transfusions are life-
saving in medical emergencies, but the time-consuming
process of determining a patient's blood type in a
laboratory can slow things down. The danger of
incompatibility is minimal in blood group O negative,

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:17:55 UTC from IEEE Xplore. Restrictions apply.
 which is considered universal donor. Second, because the
pre-transfusion testing are conducted manually by
technicians, human mistake can occasionally occur, with
potentially catastrophic consequences for the patient. We
have thus developed an automated approach for detecting
the blood group, which not only does so in a more time-
efficient manner but also allows for the detection of the
rare Bombay-blood-group (which is also referred to as the
HH group).

This may be easily detected when antigen H is introduced

to the blood sample. Schematic Learning based Human
Blood Group Analyzer (SLHBGA) has been shown to be
an efficient and successful approach for determining a
patient's blood type. The patient's blood type may be
determined quickly and accurately with this technology Fig.5 Image Pre-Processing
because it makes use of image processing methods that
allow for automated detection of agglutination. The Fig. 6 shows the results of applying the suggested method
suggested system's ultimate goal is to provide a platform to images in terms of segmentation and classification.
that employs an Image processing technique to conduct
blood tests according to various blood type systems. This
system accepts a blood sample as input and processes it
by capturing photographs of it using an image processing
algorithm, which then utilizes an algorithm to classify the
images and an image pattern recognition approach to
match the images.

Fig. 4 depicts the blood sample picture that will serve as

the starting point for the preprocessing and subsequent
image processing steps in the suggested technique.

Fig.6 Image Segmentation

All the following figures are produced based on the real-

world evaluations of the proposed scheme, in which the
approach is placed over the hospital scenario for testing
the efficiency of the proposed scheme for continuous ten
days to evaluate the integrity of the scheme. To
demonstrate the efficacy of the suggested technique, we
cross-validate it with a standard learning model termed
CNN, as shown in Fig. 7. This figure depicts the accuracy
of blood type identification using the proposed approach.
Table-1 below provides an excellent illustration of this
Fig.4 Input Image
point.
Table-1: Accuracy Estimation
The result of the proposed method's image pre-processing Samples
is depicted in Fig. 5, which can then be used as input for Days SLHBGA (%) CNN (%)
Tested
other stages of image processing including classification. 1 89 96.07 94.43
2 64 96.18 94.19
3 52 96.08 94.07
4 79 97.02 94.54
5 106 96.15 93.69
6 54 96.64 93.14
7 73 95.75 94.89

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:17:55 UTC from IEEE Xplore. Restrictions apply.
 8 81 96.24 93.27 To demonstrate the effectiveness of the suggested
9 59 96.18 93.16 technique, we cross-validated it with a standard learning
10 63 97.61 94.43 model termed CNN, as shown in Fig. 9. This resulted in a
significantly lower loss-ratio, as shown in the figure.
Table-3 below provides an excellent illustration of this
point.

Table-3: Reduced Loss-Ratio

No. of
Samples Tested SLHBGA CNN
Days
1 89 1.64 2.31
2 64 1.31 2.18
3 52 1.02 1.97
4 79 1.96 2.84
5 106 2.34 3.76
6 54 1.09 1.96
7 73 2.05 2.77
8 81 1.72 2.05
9 59 0.86 1.39
10 63 1.26 2.54
Fig.7 Accuracy Estimation

To demonstrate the effectiveness of the proposed method,

we cross-validate it with a standard learning model called
CNN, as shown in Fig. 8. This figure depicts the timeline
for blood group categorization using the method; Table-2
below provides an excellent illustration of this point.

Table-2: Classification Timeline

CNN (Avg.
SLHBGA (Avg.
Samples Time per
Days Time per image
Tested image in
in seconds)
seconds)
1 89 15 19.31
2 64 11 15.14
3 52 7 11.09
4 79 14 18.16
5 106 17 19.24
6 54 7.5 10.5 Fig.9 Reduced Loss-Ratio
7 73 12 16.21
8 81 14.3 18.63 To demonstrate the effectiveness of the suggested
9 59 8.2 10.54 technique, we cross-validated it with a standard learning
10 63 10.6 14.36 model called CNN, as shown in Fig. 10. This figure
indicates the sensitivity and specificity of the proposed
approach. The same is shown in a straightforward fashion
in Table-4, which follows.

Table-4: Sensitivity and Specificity

SLHBG SLHBG
CNN CNN
Day Sampl A A
Sensitivi Specifici
s es Sensitivi Specifici
ty (%) ty (%)
ty (%) ty (%)
1 89 93.14 90.02 87.29 84.17
2 64 92.06 91.72 86.16 84.54
3 52 93.64 90.57 85.27 84.76
4 79 91.27 91.64 86.19 84.62
5 106 92.86 92.24 86.15 83.47
6 54 93.75 91.16 86.67 83.51
7 73 93.24 90.14 85.45 84.08
8 81 92.46 91.28 86.24 83.21
Fig.8 Classification Timeline 9 59 91.12 90.75 86.52 83.36
10 63 93.45 92.68 87.01 84.14

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:17:55 UTC from IEEE Xplore. Restrictions apply.
 [2] KhKhaleduzzaman, MdHasan Mahmud, et al., "Detection and
Implementation of Blood Group and Hb Level by Image
Processing Techniques", https://www.researchgate.net,
DOI:10.34104/ajeit.021.073081, 2021.
[3] R. T. Prabu, et al, "COVID-19 Identification and Detection from
CT-Images using AI Based Ensemble Model," (ICSES),
Chennai, India, 2022, pp. 1-10, doi:
10.1109/ICSES55317.2022.9914227.
[4] Shashidhar S, Harshitha G U, et al., "Automatic Detection of
Human Blood Group with Special Case in Low Resolution
Using Image Processing", International Journal of Progressive
Research in Science and Engineering, 2020.
[5] G. Anitha, et al "Efficient Internet of Things Enabled Smart
Healthcare Monitoring System Using RFID Security Scheme"
Intelligent Technologies for Sensors, 1st Edition, 2023, Apple
Academic Press, ISBN: 9781003314851.
[6] Priyadarshini, R.,et al, “Medchain for Securing Data in
Decentralized Healthcare System Using Dynamic Smart
Contracts”, ICIPCN 2022. Lecture Notes in Networks and
Fig.10 Sensitivity and Specificity Systems, vol 514, pp. 574–586 2022, Springer, Cham.
https://doi.org/10.1007/978-3-031-12413-6_45.
V.CONCLUSION AND FUTURE SCOPE [7] Nuha Odeh, Anas Toma, et al., "An Efficient System for
Automatic Blood Type Determination Based on Image
Matching Techniques", https://www.mdpi.com,
Pervasive techniques like injections are become https://doi.org/10.3390/app11115225, 2021.
unnecessary thanks to this initiative. In-house needs may [8] Sayali Neve, Pallavi Kumawat, et al., "Blood Group
Identification Using Machine Learning- A Review of
simply include this. It's a cheaper technique that can help Literature", http://ijariie.com, 2021.
speed up and improve the accuracy of the process of [9] AksheenMalhotra, Chethana.N, et al., "Automated Blood Group
identifying blood. This work can be put to use in times of Determination Using Image Processing Techniques With
crisis where quick blood detection is essential. Our work Integration Of Raspberry PI", https://www.jetir.org, 2021.
[10] D. Jayaram and J Shiva Sai, "A Novel Blood Group Detection
represents a major step toward this goal as the profession Using Deep Learning", Turkish Online Journal of Qualitative
moves toward generation in every area. There are three Inquiry, 2021.
stages to the research: gathering the images, processing [11] Marife A. Rosales† and Robert G. de Luna, "Computer-Based
them, and then converting them to binary. The amount of Blood Type Identification Using Image Processing and Machine
Learning Algorithm", Journal of Advanced Computational
preprocessing necessary converts the picture to infrared Intelligence and Intelligent Informatics, 2022.
and eliminates any distracting noise. In the preprocessing [12] Raj Bhagat, Snehal Marge, et al., "Automatic Detection of
stage, the typical image processing method is utilized, Human Blood Group System using Deep Learning and Image
which is an unpredictable artificial screening method. Processing", International Research Journal of Engineering and
Technology, 2021.
After that, we used a structural approach and rendered the [13] A. A. Aristoy, Yu. A. Rozenbaum, et al., "An Automated
image, which led to the creation of a histogram; finally, Method for Blood Type Determination by Red Blood Cell
we took the derivative and zeroed in on the blood in the Agglutination Assay", https://link.springer.com, 2022.
image. The digitized image was then used to determine [14] Gururaj, N., Vinod, V. &Vijayakumar, K. Deep grading of
mangoes using Convolutional Neural Network and Computer
the type of blood of a subject. In the future, a handheld Vision. Multimed Tools Appl (2022).
gadget similar to the diabetes testing machines we use in https://doi.org/10.1007/s11042-021-11616-2.
our everyday lives might be manufactured and used by
non-specialists in locations with limited access to medical
professionals.

REFERENCES
[1] S. A. Shaban* and D. L. Elsheweikh, "Blood Group
Classification System Based on Image Processing Techniques",
https://www.techscience.com, DOI:10.32604/iasc.2022.019500,
2021.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:17:55 UTC from IEEE Xplore. Restrictions apply.