Using AI to Detect Diseases

In 2023, Will Studholme, then 58, didn’t anticipate being diagnosed with osteoporosis when he ended up in accident and emergency at an NHS hospital in Oxford due to gastrointestinal issues.

The condition, which is closely linked to aging, makes bones weak and brittle, which raises the possibility of breakage. An abdominal CT scan was performed early in the examination of Mr. Studholme’s illness, which turned out to be a serious case of food poisoning. When artificial intelligence (AI) technology was used to analyze that scan, it revealed that Mr. Studholme’s spine had a compressed vertebra, which is a frequent early sign of osteoporosis.

After more testing, Mr. Studholme received his diagnosis and a straightforward treatment: a yearly infusion of a medication for osteoporosis, which should increase his bone density. Mr. Studholme says he feels really fortunate because he doesn’t believe this would have been discovered without artificial intelligence.

It is not unheard of for a radiologist to discover something unrelated to what they were initially looking for in a patient’s imaging, such as an undiagnosed tumor or an issue with a specific tissue or organ.

However, it is novel to use AI in the background to methodically go through scans and automatically spot early warning indicators of common, preventable chronic diseases that may be developing, regardless of why the scan was initially requested.

According to Perry Pickhardt, a professor of radiology and medical physics at the University of Wisconsin-Madison, who is responsible for inventing the algorithms, the clinical application of AI for opportunistic screening, also known as opportunistic imaging, “is just beginning.”

It is seen as opportunistic since it exploits imaging that has previously been performed for a different clinical reason, such as suspected malignancy, a chest infection, appendicitis, or abdominal pain.

Before symptoms appear, it may be able to identify diseases that were previously undetected, when they are easier to cure or can be stopped from getting worse. “We can avoid a lot of the lack of prevention that we have missed out on previously,” Professor Pickhardt states.

He says that blood tests and routine physicals frequently miss these diseases.

There is a lot of information in CT scans about body tissues and organs that we don’t really use, says Miriam Bredella, an NYU Langone radiologist who is also working on algorithms in the area. In theory, radiologists could analyze it without AI by taking measurements, but this would take a lot of time.

She adds that there are advantages to the technology in terms of reducing bias.

Because osteoporosis, for instance, is perceived as mostly afflicting elderly, slender, white women, clinicians may not always consider looking outside of that demographic. In contrast, opportunistic imaging does not make such distinction.

An excellent illustration is the case of Mr. Studholme. Given his age, gender, and lack of bone fractures, it is doubtful that he would have received a diagnosis without artificial intelligence. Apart from osteoporosis, AI is being trained to detect diabetes, fatty liver disease, heart disease, and age-related muscle loss opportunistically.

Although CT scans, such as those of the chest or abdomen, are the primary emphasis, efforts are also being made to opportunistically extract information from other imaging modalities, such as mammograms and chest x-rays.

The scientists emphasize that if the technology is to be used on a diverse variety of people, it is crucial that the training data contain scans from a large number of ethnic groups. This is because the algorithms are trained on thousands of tagged previous scans.

Additionally, there should be a degree of human scrutiny; if the AI discovers something questionable, radiologists should verify it before reporting it to doctors.

The AI system that was utilized to analyze Mr. Studholme’s scan is owned by Nanox.AI, an Israeli corporation, is one of just a few businesses developing AI for opportunistic screening; many more are more concerned with utilizing AI to help diagnose the precise conditions for which the scans are actually performed quickly and accurately.

Nanox.AI provides three opportunistic screening solutions that use routine CT scans to help detect osteoporosis, heart disease, and fatty liver disease, respectively. Before formally launching the osteoporosis-focused solution from Nanox.AI in 2020, Oxford NHS hospitals started testing it in 2018.

According to Kassim Javaid, a professor of osteoporosis and rare bone diseases at the University of Oxford who has led the algorithm’s introduction, data from Oxford hospitals indicate a six-fold increase in the number of patients diagnosed with vertebral fractures compared to the NHS average. These patients can then be evaluated for osteoporosis and begin treatment to prevent the condition.

The algorithm is currently undergoing additional testing at hospitals in Southampton, Cambridge, Cardiff, and Nottingham. “We want to build the evidence to use it across the NHS,” Prof. Javaid notes.

Sebastien Ourselin, a professor of healthcare engineering at Kings College London and the director of the AI Centre for Value Based Healthcare, believes that although the technology can help people, there are broader implications that should be taken into account.

He points out that the additional patient numbers that the technology’s utilization may produce are a significant issue that must be balanced. “This is increasing the demand on the healthcare system not reducing it,” he asserts.

First, additional confirmatory testing, which requires money, is probably required for those identified by the opportunistic screening as possibly having a disease. Additionally, there may be a lot of needless testing if the AI is imprecise or overly sensitive.

Then, for those additional individuals who do receive a diagnosis, assistance must be available. According to Prof. Javaid, the additional load is a problem that comes with the technology, but there are workarounds.

For follow-up, patients with confirmed fractures in Oxford are sent to a fracture prevention service that is primarily provided by nurses in order to avoid overburdening doctors. “The AI does force you to change your pathway,” he states.

In the long run, Prof. Javaid thinks the NHS will save money if more people with early-stage osteoporosis are recognized and given the prophylactic medication they require. According to him, fractures are among the main causes of hospitalization.

Mr. Studholme seen the devastating effects of osteoporosis firsthand when his mother broke both of her hips as a result of the condition. He claims that in the past, it was simply seen as an elderly person’s ailment for which there was nothing to be done. He claims that he feels extremely fortunate and that he can take action before his bones turn to chalk.

Source link