UCLH to develop AI systems for precision blood matching

UCLH and Cambridge University in collaboration with NHS Blood and Transplant have won government funding to apply artificial intelligence (AI) to improve blood transfusions.

The new AI systems aim to transform the quality and efficiency of blood matching, reduce complications of blood transfusions, and improve clinical care for patients. They will be a key component of a programme of work called HAEM-MATCH that aims to deliver extended blood group matching to improve outcomes for patients with sickle cell disorder and other heavily transfused patient cohorts.

The funding award is one of several announced by the government to accelerate the testing and evaluation of AI technologies which meet the aims set out in the NHS Long Term Plan.

Consultant haematologist Dr Sara Trompeter is the UCLH lead in the work. Dr Manuel Gomez from UCL will be supporting the Health Economics aspect.

Current practice is to match blood for transfusion based on the major blood groups. Blood groups are determined by antigens – molecules found on the surface of red blood cells. Being given blood with antigen mismatch between donor and recipient can cause serious complications and this is more common in people who are very dependent on blood transfusion – for example people with sickle cell disorder. These complications include formation of antibodies that can cause life-threatening reactions, and make it very difficult to transfuse patients in the future.

Ideally, researchers would like to match for the minor blood groups too, known as extended antigen matching. However, this is currently not possible as testing for the minor blood groups is very expensive and slow, and the whole process of matching is manually performed by scientists surveying some of the 2 million donations a year.  

To get round this problem, the Blood Genomics Consortium (BGC) have developed a way of testing genes for blood groups – which is a substantially cheaper, quicker and yet equally effective technique. A computer programme called bloodTyper will then convert the genetic sequences identified via the new test into blood types. This should allow researchers to determine the extended blood groups of all donors and multi-transfused patients in the future.

The AI award will enable to team to develop this computer programme, as well as automate several other processes related to blood transfusions.

The team will develop a programme called bloodMatcher to enable the best allocation of units of blood from blood stocks to patients. Currently, laboratory scientists manually pick units from blood stocks. bloodMatcher will be an AI driven process that will be able to automatically allocate the best units of blood to patients with much more extensive blood groups matching, and is based on a prototype created by collaborators in the Dutch blood service, Sanquin.

The researchers will also develop an automated system called bloodStocker which will enable the donation and stocking of blood from donors to match the specific needs of patients who require a transfusion. Currently, donation is encouraged from people with rare blood types, but these donations do not necessarily match up with the types of blood patients need. This will increase our ability to extended blood group match blood and reduce wastage also.

Dr Trompeter said: “We hope to develop a set of automated systems – building on what we have achieved so far – that will ensure donated blood is more precisely matched with patients who need it. This will reduce the complications of transfusion, improve efficiency and reduce waste.”

The AI Award is one of the programmes that make up the NHS AI Lab, led by NHSX and delivered in partnership with the Accelerated Access Collaborative (AAC) and National Institute for Health Research (NIHR).