How do you determine whether an individual has blood type A, B, AB or O? In Denmark, this is automatically determined in the blood banks, but in developing countries where advanced equipment for analysing a blood test is not available, the so-called EldonCards remain an important tool for physicians and health professionals.
The card, invented by the Danish physician Knud Eldon in 1954, is a piece of plastic with four circles of antibodies. If a drop of blood is transferred to each of the four circles, a grainy structure called agglutination may develop in each of the circles depending on the blood type. In case of non-agglutination, the blood in the circle remains a homogeneous red colour.
Therefore, Eldon Biologicals that sells the cards came up with the idea of developing an algorithm that can distinguish whether the blood clots, i.e. agglutinates, or remains a homogeneous red colour. In collaboration with our Visual Computing Lab, they have developed a prototype of a mobile application that is able to read the blood type based on the blood deposited on the card. The application can then send the result to a blood bank or others who have an interest in the result. It can also tell if the blood type determination is not performed correctly and therefore faulty.
Kasper Juul Hedegaard, international sales and marketing manager at Eldon Biologicals, says:
We needed a link between the analogue and the digital world. When the blood reacts with the antibodies it forms small islands that resemble a QR code, and therefore we wondered if it was possible to make an application that could scan the cards and come up with the result.
The company's customers include field hospitals and relief organisations operating primarily in third world countries. In the United States, the cards are used for blood type determination in private homes, e.g. if people want to eat according to their blood type.
Converts the texture of the blood to numbers
Senior Computer Graphics Engineer Jens Rimestad from Visual Computing Lab, who was responsible for the development, used computer techniques that use visual recognition to find the circles. Subsequently, he digitised the texture of the blood, and finally he used machine learning to determine whether or not the blood clots. Jens Rimestad says:
We have made various image operations on the cards that provide visual information about the texture of the blood. The information is combined into a descriptive code number that can be used to determine whether or not the blood in a circle is agglutinated. The algorithm is trained and validated on the basis of a lot of cards where we know what the result is.
Freely translated this means that based on training images he has defined a high-dimensional space that can be divided so that the agglutinated circles occupy part of the space and the non-agglutinated another part. This means that today we have a motor that works. It is a model that with a close to 100 percent says whether or not a result has been found.
 |  |
"The Alexandra Institute has really helped us through as this was completely new territory for us. They have managed to translate our blood cards into a result. This means that we can connect the analogue with the digital. I think the mobile application is world class, but really it is only for internal use, so we need to create a user interface that makes it easy to use," says Kasper Juul Hedegaard.
Tool against rhesus disease
Eldon Biologicals is a co-founder of and collaborates with the organisation CURhE, a group of physicians in Africa, India, the United States and Canada, who fights the so-called rhesus disease. In Denmark it is unknown, but it is a disease of neonates, which is due to the fact that a rhesus negative mother forms antibodies to the red blood cells of a rhesus positive foetus.
"It does not have great significance for the first child, the woman is pregnant with. However, if she becomes pregnant again, she may have a miscarriage, which can be dangerous in India or Africa. Subsequent pregnancies could also end with brain damage in newborns or stillbirth. In our part of the world, this is treated with an infusion of special antibodies. But this is not automatically done in e.g. Africa," says Kasper Juul Hedegaard.
Allows for collection of statistical data
CURhE is run by researchers at Stanford University, and they have particularly called for a digitisation of the cards.
"It allows for the collection of statistical data and for sending the result to the local blood banks. I can also send a text message with the blood type to the woman which she can then use if she becomes pregnant again,” he says.
It could also be an important tool in hospitals in Denmark, where health professionals must be ready to transfuse blood when patients arrive by ambulance. As it is today, they need to take a blood sample and send it to the blood bank. He says:
If you could scan the blood sample as soon as it is taken in the ambulance, then the right blood type could be ready faster, when they get to the operating room. Of course, this will require that it is approved as a medical device. But that is something we strongly consider.
The development of the mobile application and the collaboration is funded by the Videnkupon programme (now InnoBooster programme) under Innovation Fund Denmark.
I lande uden avanceret udstyr til at analysere en blodprøve udgør Eldon-kort et vigtigt værktøj for læger og sundhedsarbejdere