The Oxford Martin Programme on Antimicrobial Resistance Testing is developing a new type of medical test powered by AI that is portable and accessible anywhere, from a hospital to a pharmacy to a field tent serving victims of a natural disaster. The aim is to be able to determine the cause of an infection and how best to treat it, all within 30 minutes.
On a personal level, the successful development of such a test would help people all over the world recover faster from bacterial infections.
It could also help reduce burdens on healthcare systems by ensuring care and treatment are prompt and effective, preventing or cutting the time of hospital stays, and making targeted antibiotic prescribing available in more settings like pharmacies. This would also cut countries' healthcare costs by reducing hospitalisation and hospital stay lengths, and reduce prescription costs by removing the need for ‘trial and error’ antibiotic prescribing and ending antibiotic prescriptions to people with viral infections.
On a global scale, making sure antibiotics are only prescribed when they are needed and when they will be effective could slow the process of bacteria developing the ability to defeat the drugs designed to cure them. This will preserve the effectiveness of the antibiotics we have, meaning they will work for everyone for longer, creating a healthier world and buying us time to develop new antibiotics.
A healthcare professional takes a sample from a patient with a suspected bacterial infection. This sample can be anything from blood to saliva to spinal fluid depending on the patient's symptoms.
The sample is placed into the testing device's microfluidic cassette. This device separates the bacterial cells within the sample fluid from the other cells and components. It then concentrates those bacterial cells into a small area and applies fluorescent reagents so the microscope can view and identify them.
The test automatically scans these microscopy images, identifies the type of bacteria present, and applies potential antibiotic treatments to the bacterial sample. This currently requires an internet connection, but the team hopes to develop an offline version in time.
The treated sample is analysed by AI to determine which, if any, of the potential antibiotics can successfully treat the bacterial sample or whether the bacteria is a resistant strain.
The healthcare professional performing the test receives a readout of the test results showing:
The entire testing process will be rapid, and the team hopes to get this down to within 30 minutes. The most comparable test currently available takes several hours and requires expert operators to perform the testing.
The idea for this test brings multiple challenges, including sample processing, microscopy, bacterial identification, antibiotic expertise, computer imaging, and machine learning, to name a few. The program, therefore, brings together specialists from across physics, medical sciences, and computer sciences to work on different aspects of the test. Find out more about the interdisciplinary team in this short video.
We are less than 100 years on from the discovery of penicillin. In 1928 Alexander Fleming discovered it in his lab at St Mary’s Hospital Medical School, but it wasn’t until 1940 here at Oxford that it was first developed and tested as a drug to treat bacterial infections. It was a miracle drug. Diseases like tuberculosis, cholera, and even the bubonic plague, which were once death sentences were now curable. Infected wounds were no longer a life-or-death situation, and surgery, cancer treatments, and organ transplants are all much safer as a result of antibiotics.
Yet, just as the world has been getting used to being able to treat bacterial infections and diseases, the bacteria causing them have evolved to evade our antimicrobial drugs through a process of Darwinian natural selection. Because of our over-reliance on antibiotics in medical treatment, food production, preventatives, and cures, we are eliminating the sensitive bacterial strains that are susceptible to treatment, leaving behind more resistant strains that are not treated effectively by many of our antibiotic treatments.
Using the right antibiotics for a specific infection, and using them only when we need to, is thought to be one of the best things we can do to slow the process of bacteria evolving resistance and to preserve the antibiotics we do have.
As a result, we are now on course for a future where antibiotic-resistant infections represent a major cause of infection and death. Antibiotic-resistant infections already cause 1.2 million deaths annually, and this is currently increasing. The World Health Organization has named antimicrobial resistance a Global Health Emergency and one of the biggest threats to global public health.
Using the right antibiotics for a specific infection, and using them only when we need to, is thought to be one of the best things we can do to slow the process of bacteria evolving resistance and to preserve the antibiotics we do have. That’s because the use of antibiotics when a bacterial infection isn’t present or using the wrong antibiotic for a situation can accelerate the bacteria’s process of developing resistance.
This test will:
All of this would improve global antimicrobial stewardship and help maintain our ability to tackle bacterial infections for future generations.
The test uses images taken by a microscope and an AI system to tell whether bacteria causing an infection and then exposed to different types of antibiotics can be successfully treated by those antibiotics. Susceptible bacteria will display visual changes in the cell structures, whereas resistant bacteria won't – they look the same as if they hadn’t been exposed to the antibiotic.
Humans are brilliant at spotting these sorts of changes, so the team is using people to train the AI system on what to look for in the sample images. In a citizen science project with Zooniverse.org called Infection Inspection, anyone with internet access from anywhere in the world can contribute to making this rapid antimicrobial testing a reality.
All you have to do is look at images and say whether you think the bacteria has reacted to an antibiotic challenge or not. Infection Inspection will train you on what to look for, and then you can look at 10 images or 10,000. Through this, you'll be helping to make our diagnostic test a reality and contributing to the global fight against antimicrobial resistance.
Take part in Infection InspectionThe test is already working well in lab conditions, although the research team is working on making it faster, improving the AI using Infection Inspection training, and widening the number of bacteria it can identify and antibiotics it applies.
The main challenge for the future of the project will be scale and standardisation. The Oxford Martin Programme on Antimicrobial Resistance Testing is developing the testing process, but to take it to the point of deployment and start helping patients and clinicians around the world, it needs to be manufactured at scale. That manufacture needs to be affordable to make it as widely accessible as possible, especially if, as the team hopes, the test will become a useful tool in difficult situations where bacterial infections are more likely like refugee camps, war zones, and in medical aid responses to natural disasters.
Scale manufacture also means it needs to be standardised and compatible with a wide range of cheap and accessible medical microscopes. The system needs to be easy to use and the offline version needs to be fully developed to enable its use in remote parts of the world.
All of this means the team needs a commercial partner or further collaborators to commercialise the testing system and bring it to market. With that support, they believe they could have a manufacturing standard test ready in three-to-five years.
If this is something you’d be keen to discuss you can contact Julian Laird, Oxford Martin School’s Head of Policy and Development, on +44(0)1865 287356, or at julian.laird@oxfordmartin.ox.ac.uk
If you found this page useful, sign up to our monthly digest of the latest news and events
Subscribe