AUTOMATIC DETERMINATION OF HUMAN BLOOD TYPES

USING IMAGE PROCESSING TECHNIQUES

Ana Ferraz, Vítor Carvalho and Patrícia Brandão

School of Technology, Polytechnic Institute of Cávado and Ave, Urbanização Quinta da Formiga Ed. Panorâmico II

Lote 3 4750-117 Arcozelo, Barcelos, Portugal

Keywords: Blood Types, Image Processing, IMAQ Vision, Plate Method.

Abstract: This paper presents a methodology to automatically determine human blood types using image processing

techniques. As a reference, in the experimental analysis it was used the plate method, being registered the

results with a digital camera. The obtained images were analyzed and processed with a custom application

developed with IMAQ Vision from National Instruments, allowing the automatic blood type classification

of the sample under test. The implementation in health units of a system based in the presented approach

will enable between others, the risk reduction of fatal transfusions associated with wrong human blood type

interpretation.

1 INTRODUCTION

Currently, the determination of human blood type is

carried out manually using the plate or cards

methods (Datasheet of Diamed, 2008) (Rod, Tate

and Trent, 2005) (Diamed (a) 2009) (Diamed (b),

2009). The first one (plate) is based on the

immunological reactions that occur when there are

mixed certain serums (anti-A, anti-B, anti-AB and

anti-D) in the sample of blood in test (Figure 1)

(Datasheet of Diamed, 2008) (Rod, Tate and Trent,

2005). Then, the interpretation of the agglutination

reactions allow to determine the antigens present in

the red globules of the sample of blood, allowing the

classification of the blood type (Rod, Tate and

Trent, 2005). The second one (cards, Card-ID for the

ABO/Rh system) uses the Diamed system (Rod,

Tate and Trent, 2005) (Diamed (a), 2009) (Diamed

(b), 2009). This approach requires mixing the

sample of blood with the content of the microtubes

available in the Card-ID, followed by a

centrifugation and the results interpretation

(Datasheet of Diamed-ID, 2008). It is an accurate

approach, although it has the disadvantage of

requiring 30 minutes, which is excessive especially

in emergency situations. Thus, considering that the

plate method allows getting results immediately and

efficiently (in the worse case it takes 2 minutes), it

was used as a reference in this work.

Figure 1: Serums in plate [a) – anti-A, b) –anti-B, c) –

anti-AB, d) – anti-D].

As referred previously, the results obtained by the

plate or cards methods require human interpretation,

being susceptible to failures. In this scope, it is

known that the risk of a fatal reaction due to

administration of a wrong sanguineous type is 1 in

each 600,000-800,000 transfusions (Alexander,

2007) (Muller and Girard November 1983). An

automatic methodology of human blood type

determination will contribute to minimize or even

eliminate these statistical results. During the

bibliographical research it was possible to identify

two approaches based on optical sensors (Alexander,

2007) (Nano2Life, 2009). However, the innovative

method presented in this work, using image

processing techniques, is characterized by an inferior

cost of implementation and production, lower

Ferraz A., Carvalho V. and Bandão P. (2010).

AUTOMATIC DETERMINATION OF HUMAN BLOOD TYPES USING IMAGE PROCESSING TECHNIQUES.

In Proceedings of the Third International Conference on Biomedical Electronics and Devices, pages 69-74

DOI: 10.5220/0002698600690074

 SciTePress

complexity and greater portability, being an added

value commercial solution.

2 APPLIED IMAGE ANALYSIS

TECHNIQUE VALIDATION

The methodology applied in this work was

validated, as referred previously, using the plate

method as a reference. The obtained reactions were

registered in real size, using a CCD camera (Kristian

and Blouke, October 1982) (Sony Cyber-shot DSC-

S600) with 6.0 megapixels of resolution.

Subsequently, the analysis of the obtained images

was performed using an image processing tool

(IMAQ Vision from National Instruments (IMAQ,

2004)).

Figure 2: [a) - Application of the function Line Profile in

the image edges section, b) - Magnification of the image

a) pointing out the different zones of intensity (A, B and

C)].

Observing Figure 2 and Figure 3 it is verified that in

zone A a reduced level of pixels intensity oscillation

occurs, corresponding to a low level agglutination

zone. On the contrary, in zone B significant

oscillations are verified, corresponding to a region

where the agglutination is highly notorious. Finally,

in zone C t is observed a significant descending of

Figure 3: Results of the application of the function Line

Profile in Figure 2.

the pixels intensity amplitude, corresponding to the

used serum location.

Figure 4 illustrates the Line Profile function

application (Klinger, 2003) (Dougherty, 2009)

(Bernd, 2009) (Burger and Burge, 2009) (Tinku and

Ajoy, 2005) to an image without agglutination.

Figure 5 presents the results.

Figure 4: [a) - Application of the function Line Profile in

the edges section, b) - Magnification of the image a)

pointing out the different zones of intensity (A and B)].

Observing Figure 5 it is verified that in zone A

oscillations practically do not occur. This fact results

from the non occurrence of agglutination in Figure

4, leading to an almost constant level of pixels

intensity. In zone B, corresponding to the zone of the

used serum, a descending of the pixels intensity is

verified.

Line Profile

Zone A Zone C

Zone B

Line Profile

Zone A

Zone B

BIODEVICES 2010 - International Conference on Biomedical Electronics and Devices

Figure 5: Results of the application of the function Line

Profile in Figure 4.

Thus, through the obtained results, it is possible to

conclude that when agglutination occurs, there are

observed zones in the image analysis that present

high levels of oscillation of pixels intensity. These

preliminary results allow to validate the

methodology used, enabling a secure determination

of the agglutination occurrence or not. Then, using

statistical analysis it is possible to quantify

mathematically the obtained.

Figure 6 presents a schematic which resumes the

different image processing steps applied (Figure 2

and Figure 4) as well as their transformations, in the

initial images.

Figure 6: Application of the function Line Profile (Figure

2 and Figure 4).

3 ALGORITHM OF AUTOMATIC

DETERMINATION OF

OCCURRENCE OF

AGGLUTINATION

To automatically determine the sanguineous type of

a sample of blood, there were applied sequentially,

the following image processing techniques available

in the IMAQ Vision tool:

1) Function Extract Color Planes, specifically in the

option Extract RGB Green (Klinger, 2003)

(Dougherty, 2009) (Bernd, 2009) (Burger and

Burge, 2009) (Tinku and Ajoy, 2005). This function

allows extracting the green color plan of an RGB

image (IMAQ, 2004). After this action, it is possible

to identify with superior emphasis, the zones with

occurrence of agglutination in the analyzed images.

Figure 7 presents the application of this function in

the images of Figure 2a) – with agglutination

occurrence and Figure 4a) – without agglutination

occurrence.

Figure 7: Application of the function Extract Color Planes

(Extract RGB Green) to the images of the Figure 2a) and

Figure 4b), respectively identified as a) and the b).

Observing Figure 7a) it is possible to identify

variations in the color intensity between black and

white (agglutination occurrence). Figure 7b)

assumes a homogeneous tonality of black color,

without having significant oscillations in the color

intensity (without agglutination occurrence). As the

transition level between contrasts black and white, is

superior to the transition levels between red

contrasts and the colors used in the serums, the

application of this function makes possible a

mathematical quantification of the agglutination

occurrence.

2) Quantify function (Klinger, 2003) (Dougherty,

2009) (Bernd, 2009) (Burger and Burge, 2009)

(Tinku and Ajoy, 2005). This function allows

quantifying statistically, through the levels of pixels

intensity, selected areas of an image. This

quantification includes the average, the standard

deviation, the minimum and the maximum

amplitude of the analyzed pixels (Bisquerra,

Martínez and Sarriera, 2004).

Figure 8 presents the application of the Quantify

function in the images of Figure 7.

AUTOMATIC DETERMINATION OF HUMAN BLOOD TYPES USING IMAGE PROCESSING TECHNIQUES

Figure 8: [a) Application of the Quantify function to the

image of Figure 7a), b) application of the Quantify

function to the image of Figure 7b)].

Table 1 presents the results of the statistical

parameters, determined with the Quantify function,

in the selected areas of Figure 8 images.

Analyzing Table 1, it is verified that the standard

deviation of the image where agglutination occurs

(Figure 8a), is highly superior to the standard

deviation of the image without agglutination (Figure

8b). Significant differences are also observed in the

other analyzed parameters, excepting as expected, in

the minimum value (0.00), corresponding to the

white tonality.

Table 1: Results of the application of the Quantify

function in Figure 8.

Figure Average

(pixels)

Standard

deviation

(pixels)

Minimum

(pixels)

Maximum

(pixels)

8a)

18.78

26.94

0.00 224.00

8b) 6.25 13.59 0.00 172.00

The standard deviation is the parameter that allows

distinguishing with superior exactness and

effectiveness, the occurrence or not of agglutination,

resultant from the quantification of the pixels

intensity deviation in the analyzed image area. It was

verified that for the image acquisition conditions

used, the agglutination occurrence is translated by a

standard deviation level superior than 20 pixels. So,

using the described algorithm, it is possible to

automatically classify the sanguineous type of a

sample of blood.

Figure 9 presents a schematic which resumes the

different image processing steps applied (Figure 7

and Figure 8) as well as their transformations, in the

initial images.

Figure 9: Application of the functions Extract Color Plane

(RGB Green) and Quantify (Figure 7 and Figure 8).

4 EXPERIMENTAL RESULTS

The methodology described in the previous section

was applied to a set of images with the same

sanguineous type, using the four different serums of

test. Figure 10 presents the acquired images and

Figure 11, the images obtained after the application

of the image processing techniques.

Figure 10: Acquired images of the samples of blood mixed

with the different serums of test.

Quantify

)

BIODEVICES 2010 - International Conference on Biomedical Electronics and Devices

Figure 11: Application of the techniques of image

processing to the images of Figure 10.

Table 2 presents the results of the statistical

parameters associated to the zones of analysis

identified in Figure 11 images.

Table 2: Results of the statistical parameters applied to

Figure 11 images.

Figure Serum Average

(pixels)

Standard

deviation

(pixels)

Maximum

(pixels)

11a)

Anti-A 16.32

26.63

200

11b) Anti-B 13.80 15.10 177

11c)

Anti-AB 10.33

21.09

176

11d) Anti-D 5.37 14.25 167

Analyzing Table 2, it is verified that agglutination

does not occur in Figures 11b) and 11d) (standard

deviation inferior the 20 pixels) but occurs in

Figures 11a) and 11c) (standard deviation superior

the 20 pixels). It is also observed that although

agglutination occurs in Figure 11c), its level is not as

significant as in Figure 11a), as a result of an inferior

value of standard deviation, 21.09 pixels < 26.63

pixels. This result from the fact that has been

administered the serum anti-AB in Figure 11c),

occurring agglutination in the presence of a lower

quantity of serum A. Thus, considering the results

obtained, it is concluded that the sanguineous type of

the analyzed blood is A (Rod, Tate and Trent, 2005),

given that the occurrence of agglutination was only

verified in the presence of the serums anti-A and

anti-AB.

5 SOFTWARE DEVELOPED

After the implementation and validation of the

previous sections described methodology, it was

developed a custom software application using

IMAQ Vision and LabVIEW

from National

Instruments (Klinger, 2003) (Relf, 2003). Figure 12

presents the interface of the developed application.

Figure 12: Interface of the developed application [a) -

initial Image, b) – final Image, c) – Selection of the region

of interest (ROI), d – Obtained results].

The interface of the developed application (Figure

12), is segmented in four different sectors: sector a),

presenting the initial acquired image, sector b),

presenting the final image obtained after the

application of image processing techniques, sector

c), for selection of the region of interest in the final

image and finally, sector d), for presenting the final

results.

6 CONCLUSIONS AND FUTURE

WORK

Considering the studies presented in this work, it can

be concluded that the applied image processing

techniques enable determining automatically, fast

and accurately, the sanguineous type of the analyzed

samples of blood. Clearly distinct zones in the pixels

intensity of the images are identified, allowing

classifying with mathematical basis quantification

the agglutination occurrence.

As a reference it was used the plate method,

adjusted conveniently to the methodology of

detection of sanguineous type using image

processing, presenting safe results in a time inferior

to 2 minutes. Thus, the use of the approach

described in this work allows eliminating the errors

committed by the technicians in the sanguineous

Quantify

c) d)

Quantify

b) a)

Quantify

AUTOMATIC DETERMINATION OF HUMAN BLOOD TYPES USING IMAGE PROCESSING TECHNIQUES

type classification. This will contribute to undertake

safe blood transfusions and to reduce the loss of

human lives.

In the near future it is intended to develop a

commercial, innovative, portable and low cost

system, likely to being used in health units. This

system will have a reduced requirement of patient

blood quantity allowing reducing the wastes of

blood in tests. Moreover, it is also desired that this

system registers, among others, the flow of

performed tests and required types of blood. These

data, associated to a health information system will

become an instrument capable of organize and

analyze the requirements that allow defining

problems in the health area. As a result, this will

stimulate the development of new solutions that

attend specifically to the necessities of the services

given to the population.

REFERENCES

Datasheet of Diamed AG,1785 Cressiers/Morat, Cressier,

2008.

Rod, R S., Tate, P., Trent, D S., 2005. Anatomia &

Fisiologia. 6ª ed. Lusociência, Loures.

Diamed (a) - Diagnostic and Medical Products [Internet

homepage]. [access in 2009 April] Available at:

www.diamed.pt/frontend/index.php?option=com_cont

ent&task=view&id=2&Itemid=3

Diamed (b) - Diagnostic and Medical Products [Internet

homepage]. [access in 2009 April] Available in:

www.diamed.ch/product_detail.aspx?id=92&navvis=

Datasheet of Diamed-ID, 2008. Card-ID. Diaclon ABO\Rh

for patients. Cressier.

Alexander, S P., 2007. An Integrated Microoptical

Microfluidic Device for Agglutination Detection and

Blood Typing [Master´s thesis]. California: University

of North of Carolina.

Muller, A., Girard, M., November 1983. Automatisation

des activités de contrôle immunohématologiques dans

les établissements de transfusion sanguine français

Revue Francaise de Transfusion et Immuno-

hématologie. Vol. 26 (5), p.517-530.

Nano2Life – Bringing Nanotechnologies to Life [Internet

homepage]. [access in 2009 April] available in:

n2lvip.tau.ac.il/attachments/023_S_ Wolff.ppt.

Kristian, J., Blouke, M., October 1982. Charged-Coupled

Devices in Astronomy. Scientific American. Vol.

247(4), p. 66-74.

IMAQ, 2004. IMAQ Vision Concepts Manual. National

Instruments, Austin.

Bisquerra, R., Martínez, F., Sarriera, J C., 2004.

Introdução à Estatística – Enfoque Informático com

Pacotes Estatístico SPSS. Artmed, São Paulo.

Klinger, T., 2003. Image Processing with LabVIEW and

IMAQ Vision. Prentice Hall, New Jersey.

Relf, Christopher G., 2003. Image Acquisition and

Processing with LabVIEW, CRC Press.

Dougherty, Geoff, 2009. Digital Image Processing for

Medical Applications, Cambridge University Press.

Bernd, Jahne, 2009. Practical Handbook on Image

Processing for Scientific and Technical Applications,

CRC Press.

Burger, Wilhelm, Burge, Mark J., 2009. Principles of

Digital Image Processing: Fundamental Techniques.

Springer.

Tinku, Acharya, Ajoy K. Ray, 2005. Image Processing:

Principles and Applications. Wiley-Interscience.

BIODEVICES 2010 - International Conference on Biomedical Electronics and Devices