Scroll Top

How Machine Learning is Changing Rare Disease Diagnosis

The TV medical drama House, which aired from 2004-2012, portrayed a group of doctors, led by the irascible Dr. House, solving medical mysteries1. Patients with obscure or nonspecific symptoms would undergo a diagnostic odyssey over the course of the episode, often resulting in the diagnosis of a rare illness or unusual presentation of a more common disease. In reality, the diagnostic odyssey of a patient with some autoimmune, genetic, or metabolic diseases can take months or years, not days or weeks. The situation is complicated by the relatively few numbers of patients any individual doctor will see over their career, so doctors generally look for the common disorders first, rather than a rare explanation for the patient’s symptoms. Even when a diagnosis can be made, there may be no currently available treatment and the patient could be one of only a handful of patients worldwide with the disorder. For patients (and their families) and providers alike, this situation is frustrating, lengthy, and expensive. An early diagnosis can sometimes mean the difference between a treatable illness and lifelong complications or even death.

In MoneyBall Medicine: Thriving in the New Data-Driven Market, I describe how research and diagnosis are being transformed by new ways of collecting and analyzing data and technologies, including genomic and other advanced molecular diagnostics, machine learning/artificial intelligence, and pattern recognition/image analysis. There are numerous examples of these techniques being used for oncology or in other specialties: radiologists and pathologists are utilizing image analysis software to help them identify areas of concern for cancer or kidney disease2 more rapidly than a human can, often with accuracy that meets or exceeds doctors3. Diabetic retinopathy can be diagnosed by an artificial intelligence-based system without the need for a doctor to analyze the retinal scan4. But facial recognition technology is tackling rare disease diagnosis with surprising results.

In 2012, the Israeli software company Face.com sold its facial recognition platform to Facebook5. The company’s founders and attorney Dekel Gelbman had no healthcare experience but wanted to make a difference using technology. After talking to a variety of medical specialists, they learned that facial dysmorphologies sometimes offered clues to help geneticists diagnose young children. With a clear need to improve and accelerate rare disease diagnosis and their technical prowess, the trio launched FDNA in 2011 and their Face2Gene platform. Similar to the way that your photo application can recognize the same person in multiple pictures, even if the person’s hair color might change over time, Face2Gene uses artificial intelligence to find patterns and similarities in the facial structure of patients and uses that data along with existing clinical information to help doctors reach a diagnosis.

As artificial intelligence-based technology like Face2Gene is trained on more patients, the answers it can give to doctors improves. Instead of the long diagnostic odyssey that patients and their families currently face, this technology can point doctors in the direction of specific disorders or even particular genes that could be responsible for the patient’s symptoms, potentially reducing the number of diagnostic tests needed to confirm the diagnosis. Join me for Episode 4 of the MoneyBall Medicine Podcast where I talk to Dekel Gelbman, CEO of FDNA where we discuss how their Face2Gene platform is transforming rare disease diagnosis and the difficulties the group faced entering a clinical field with a technical background. We’ll touch on the unique challenges of hiring talent in this competitive industry and conveying the value proposition of technology to clinicians and payers.

  1. https://www.imdb.com/title/tt0412142/
  2. Siwicki, B. 2018.Mass General, Brigham and Women’s to apply deep learning to medical records and images. Healthcare IT News. https://www.healthcareitnews.com/news/mass-general-brigham-and-womens-apply-deep-learning-medical-records-and-images
  3. Schatz, R. D. 2018.Mount Sinai partners with AI startup to detect and manage kidney disease. Modern Healthcare. http://www.modernhealthcare.com/article/20180604/NEWS/180609972
  4. U.S. Food and Drug Administration.FDA permits marketing of artificial intelligence-based device to detect certain diabetes-related eye problems. 2018-04-11. https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm604357.htm
  5. https://m.jpost.com/Jpost-Tech/From-Facebook-facial-recognition-to-genetic-disease-identification-569555