- Standard tests for many common health conditions take hours or even days to return results.
- This leads to anxiety for patients and may delay effective treatment.
- In a new study, scientists have used machine learning to develop a test to detect changes in protein concentrations that indicate disease.
- By rapidly analyzing macromolecules in biofluids, such as blood, joint fluid, and saliva, the test could return a diagnosis in just 2 minutes.
- The researchers suggest it could be valuable for both diagnosing and monitoring health conditions.
Rapid diagnosis is key to the effective treatment of many health conditions. One method of diagnosis is to measure the concentration of different proteins in biofluids, such as blood,
Rheumatoid arthritis can be indicated by changes in the level of hyaluronic acid in synovial fluid. The level of
So diagnosis often relies on measuring and analyzing these proteins. However, most tests either require a large volume of fluid, are complex to carry out, or take several hours to produce results.
Now, a team from Swansea University in the United Kingdom has used artificial intelligence, in the form of machine learning, to develop a test that could diagnose conditions using only a tiny sample of biofluid. The test produces results in under 2 minutes, so it may be a rapid way to diagnose and monitor common disorders.
“We are willing to challenge the status quo in medical research to introduce new disruptive technologies that can improve diagnosis and monitoring with the ultimate aim of saving people’s lives. We believe that our platform can deliver that in the long-term, and the advent of machine learning has made our vision more possible.”
– Dr. Francesco Del Giudice, project lead and associate professor, Chemical Engineering, Faculty of Science and Engineering, Swansea University.
The study appears in Analytical Chemistry.
Biofluids, such as blood plasma, synovial fluid, and saliva, are made up of water containing macromolecules, many of which are proteins. The proteins change the properties of the fluid from Newtonian — having a constant viscosity — to non-Newtonian. When under external force, the viscosity of non-Newtonian fluids changes.
The researchers used a microfluidic rheometer to analyze tiny volumes — 100–200 microliters — of non-Newtonian fluids using temperature-dependent
Dr. Del Giudice told Medical News Today that “[t]he system was developed by combining existing pieces of commercialized technologies — e.g., microscope, pressure pumps, heating system, microfluidic device — with a machine learning algorithm trained in-house.”
In their research, the investigators used two different fluids, polyethylene oxide (PEO) and hyaluronic acid (HA). They used PEO, a standard polymer for microfluidic applications, to demonstrate the accuracy of the device. They then used the microfluidic rheometer to rapidly evaluate different concentrations of HA at various temperatures.
To mimic biofluids, the researchers used concentrations of HA that are within the normal range found in joint fluid.
The researchers identified the longest relaxation time of the fluid as a potential biomarker of several health conditions and diseases. This measure can identify the contribution of the larger macromolecules in a solution, which is important in assessing joint condition by synovial fluid.
Fibrinogen, which is an indicator of cardiovascular disease, is a very large macromolecule, so can be easily identified using this system.
Another advantage of this method is that only a very small volume of biofluid is needed for testing. In their experiments, the researchers used only 100-200 microliters — approximately 2 drops — of fluid.
In a clinical setting, this would mean that only small samples of blood, synovial fluid, or saliva would need to be taken from patients for testing.
The researchers stress that their experiments have proved only concepts and that testing with actual biofluids is essential before the method can be taken further.
“The system has not been tested in the clinical setting, and it remains a proof of concept. We currently are seeking funding to translate our proof of concept within the medical field. For studies on the joint grade, we are seeking interested partners and we could begin pre-clinical experimentation soon. For cardiovascular disease monitoring, with an estimated 3–5 years route to clinical setting and around 2 years to test in pre-clinical conditions.”
– Dr. Francesco Del Giudice
However, if the results can be replicated in real-world situations, it may provide a rapid and reliable method for diagnosing and monitoring a number of common health conditions.
Dr. Del Giudice told MNT: “We believe that our technology could become extremely important to monitor cardiovascular disease progression. Currently, mainly physiological monitoring is pursued in hospital because of the lack of specialized staff and the length of testings (12 hours or more).
“We instead believe that our platform could be essential at the point of care, near the patient bed, and it would allow a rapid analysis turnaround within 2 minutes using 2 drops of blood,” he noted.
This rapid testing method shows huge potential, but there is still work to do before it might be seen in a clinical setting.