Applying AI & Machine Learning to Accelerate Rare Diagnosis

Accelerating diagnosis remains the fundamental challenge in rare disease. There is no magic wand for resolving this challenge but technological advances in AI and machine learning can provide demonstrated benefits in helping to overcome it.

Those benefits reveal themselves in the highly specific medical subfield of rare disease diagnostics and treatment, which provides a model for the broader applications of AI algorithms for other treatments and conditions. In the U.S., a rare disease is defined as a condition that affects fewer than 200,000 people, and some are so rare that they affect only one person. Diagnosing and treating these very rare conditions poses unique and daunting challenges, so much so that it takes an average of 5 to 7 years to get an accurate diagnosis. Rare diseases can have widely varying symptoms, which often mimic other, more common conditions. This, coupled with a lack of awareness/education and inefficient pathways, causes delays in diagnostic decisions and blocks access to appropriate supportive care and treatments.

The algorithms of today’s advanced AI applications can address the challenges of this unique subfamily of health conditions. Rare disease research represents the biggest application of AI predictive technologies in healthcare because this is an area in which the ability to accurately identify a small group of potential patients from a worldwide population becomes crucial for timely diagnosis and treatment.

Operating on massive sets of health data collected from all available sources, including electronic records, manual case notes, and an array of real-world data from a variety of secondary sources, AI-powered algorithms can search these datasets for information on symptoms, past treatments, and a variety of other data points in the history of any patient’s interaction with the healthcare system. The results of this kind of operation can be shared with practitioners on the “front lines” of patient care to provide more complete lists of symptoms, past treatment options, and outcomes.

This kind of data mining and analysis can also be used to conduct research and run clinical trials of potential treatments. AI-driven data searches can reveal past treatment efforts, point the way toward new ones, and identify patients who could participate in trials. For very rare diseases with a small patient base, AI-powered algorithms can search records from around the globe to find potential patients and collect information about them. The predictive power of well-designed AI models can be invaluable for reducing the time to diagnosis and correct treatment for the world’s rarest diseases. That power also extends to the management and treatment of many more common diseases and conditions such as diabetes and cancer. In that way, AI technologies can play a key role in a new era of precision, patient-centered medicine.

It is clear that we should be excited about the potential of AI and machine learning in providing solutions that address barriers in rare disease diagnosis. We must continue to pressure test these applications and ensure that data-driven solutions are part of a suite of activities that support patients, caregivers, and physicians along the diagnostic pathway.

This article was first published in November 2021 as part of the FierceBiotech whitepaper titled ‘How AI is Changing the Health Communications Landscape’.

Access the full whitepaper ‘How AI is Changing the Health Communications Landscape’

If you would like to learn more about the potential of data in diagnostics, watch our webinar on ‘Generating data led diagnosis accelerators in rare disease’.