Using the Power of AI for Rare Diseases

An interesting article recently published in the Washington Post takes a brief look into the use of AI in the medical industry, and raises the question whether or not AI can effectively diagnose medical mysteries. Despite most of us understanding the imprecise practice of self-diagnosing medical conditions based solely on information obtained from the internet, many of us still regularly jump online to find out what’s wrong, and more often than not, one of the many reputable medical website does indeed provide the information you’re looking for.

With the rise of AI chatbots it was only a matter of time before people would attempt to utilize this next-generation technology to analyze the millions of medical literature papers with the goal of diagnosing common medical ailments, as well as rare diseases and hard to diagnosis conditions. For example, the few instances where patients reported “elevated inflammation marker in a blood test combined with pain in your left heel mean” the AI bot can find these documents and analyze them for any related information.

A frustrated mother, who had taken her son to 17 doctors for chronic pain, decided to put his medical info into ChatGPT, which to her surprised, quickly returned a suggested diagnosis of tethered cord syndrome; and of course, this diagnosis was correct. Whilst this example represents the very best results, and there’s no study to suggest how often this might occur, the process still shows great promise as it allows patients to assist in diagnosis, which can be extremely helpful when it comes to detecting rare diseases and conditions that are hard to diagnose.

Rare diseases are reported to affect an estimated 30 million people in the United States alone, and hundreds of millions of other people globally. These are all patients who could potentially benefit from AI powered software that assists in detecting these diseases. Isaac Kohane, chair of the department of biomedical informatics at Harvard Medical School, explained:

“Doctors are very good at dealing with the common things…

“But there are literally thousands of diseases that most clinicians will have never seen or even have ever heard of.”

Another benefit of this technology is that it places some power back into the hands of the patient, who may perform their own, effective searches, that will not get overlooked by medical professionals who, for whatever reason, may not take concerns seriously.

This can happen to even the most educated, according to Kate McCrann, who was a doctor in her final year at Yale when her newborn, Tess, began to show signs of developmental delays. She took Tess to visit a pediatrician, who told her it was nothing to worry about. However, years later, McCrann learned that her daughter actually had a rare disease known as Hao-Fountain syndrome, caused by a genetic mutation.

It was all thanks to a social media post in 2015, when Tess was 5, that prompted McCrann to become a researcher studying the genetic mutation. This led to the discovery of another 7 patients suffering from the same disease, and as a result, they launched a foundation that has since connected more than 200 more patients. Without assistance like this it can be almost impossible to reach the critical mass required to prompt researchers to look for treatments and start clinical trials. This leaves patient groups having to raise money for themselves, a vital part of the research process that can also be aided by such AI technology.

Patients with rare diseases often spend years of even the most experienced doctors trying, and failing, to diagnose the problem. Even diseases hiding in plain sight that becomes obvious once found can be extremely difficult to spot in the first place.

Machine learning in rare disease

Rare diseases are inherently limited in clinical cases, leading to few samples to study. It is a priority to address challenges and harness emerging solutions for using machine learning for rare diseases.
Read More

According to the National Institutes of Health Undiagnosed Diseases Network, around 11% of patients referred each year have mysterious conditions that can eventually be diagnosed by analyzing results and lab notes. Kohane, together with Matt Might, a computer scientist whose son died of a rare disease, believes that a large language modeled chatbot could help make these diagnoses much more quickly, and they are taking the first steps in an attempt to create one.

This recent development isn’t the only application of AI that’s being used to advance the medical industry. Another new tool, built using neural networks instead of large language models, has also shown promise in diagnosing and pooling similar patients.

The downside, however, is quite clear. People can become too easily reliant on these tools, and thus ignore other warning signs. There’s also a high likelihood that bad faith actors, perhaps even the AI themselves, may fabricate false information with the goal of misleading readers and encouraging certain actions. Furthermore, if/when this occurs, we’ll have no idea what information was false, or how much it might be skewed. 

Even correct information may have no clue how accurate it is, all of which could pose a major problem when it comes to diagnosing medical conditions.

These concerns raise, what many consider to be, the most important question surrounding AI in the medical industry: How to get the best out of the technology without experiencing the worst. That said, the medical industry already has bias, it’s how we manage this bias that counts, and this situation is no different.

Machine learning to solve problems

The Power of AI

AI is much more than a grammar checking software or art generator. AI and machine learning can help solve the toughest problems - rare disease cures.

One remedy being suggested is for governments to require disclosure of the data sources used to train the medical chatbots, and for any AI system using patient health records to be subject to government regulation. In order for this to work independent researchers will be required to scrutinize all medical advice fed to the AI bots. Medical boards/associations will also need to certify the credible ones.

The first wave of AI powered technologies to enter the medical industry seems to cut costs at the expense of questionable benefits to the patients. Software that could effectively detect hard to diagnose disease would be a refreshing change that would hopefully spur on more, similarly helpful technologies.

Nevertheless, one of the main concerns regarding AI in general is the risk of falling into a dystopian future in which machines and technology rule us all, and preventing such a dark outcome all rests on how we choose to utilize AI. In keeping with that, it’s important to remember that chatbot diagnoses should still be seen as second opinions.

Skip to content