New diagnostic test for COPD reports 91% accuracy

A British technology organisation has recently published results from a series of clinical trials, showing over 91% accuracy in diagnosing severe COPD from a single breath recording.

TidalSense, which develops AI-driven diagnostic and monitoring technologies for Chronic Obstructive Pulmonary Disease (COPD) and asthma, published results this week from studies which involved the collection of over 88,000 breath records in 295 patients who used the new N-Tidal fast response capnometer and had a variety of medical conditions which can affect breathing.

This capnometer contains a sensor measures carbon dioxide in the breath more sensitively than previously possible and transmits the ‘waveform’ to a secure cloud platform using mobile data.

TidalSense designed the device to use normal, relaxed breathing for a single minute, in a way that was comfortable for patients. This high-resolution carbon dioxide data was then used to build and validate a diagnostic algorithm using machine learning.

When the carbon dioxide data from a new patient is shared, the algorithm can tell whether that patient has COPD – an incurable medical condition that causes breathing problems because of lunch damage and affects over 1.2 million people in the UK – with a high degree of accuracy.

An updated version of the algorithm has subsequently been integrated with the device, so that it takes under five minutes from a patient breathing into the device to a diagnosis being available to a healthcare professional.

In the future, it could be used to diagnose patients with COPD during their first appointment and because it is simple to use, can be administered by any healthcare professional.

Although the disease is common, according to NHS England in the UK 1.2 million people are affected by the disease, many people do not know they have it in the UK, almost half of people affected are yet to be diagnosed.

In addition, COPD costs the NHS £1.9bn a year with the majority of the cost linked with hospitalisations in the later stages of the disease.

Chris Carlin, consultant respiratory physician at NHS Greater Glasgow and Clyde, said: “The outcomes from this study are very promising.

‘This AI-driven health technology looks to have the ability to diagnose patients with COPD far quicker and more reliably than with standard spirometry testing.

‘Reducing exacerbations and hospitalisations are patients with COPD and healthcare providers’ top priorities.

‘If N-Tidal’s diagnostic test could be introduced across the NHS, there’s a very real prospect of diagnosing patients with COPD sooner and starting optimised treatment earlier.

‘This will reduce exacerbations, prevent deterioration in health and assist with the ambition to reorientate to preventative, rather than reactive care.’

The researchers have avoided using ‘black box’ artificial intelligence, allowing them to clearly see why the algorithm is determining that some patients have COPD and others do not. The algorithm has highlighted that certain parts of the carbon dioxide waveform are more important than others in making this determination, giving the team groundbreaking insight into the link between lung disease and carbon dioxide.

This is a novel scientific discovery, only possible because of TidalSense’s first-of-its-kind combination of the sensor and artificial intelligence technology.

The team now plans to start piloting the device in clinical settings such as GP practices – bringing the benefits of the rapid, accurate AI-driven diagnostic test to NHS patients.

They also anticipate further insights into the links between the carbon dioxide waveforms produced by N-Tidal and a range of other respiratory diseases in subsequent clinical studies.

Image: Robina Weermeijer