AI-Driven Platforms for Improving Diagnostic Accuracy in Rare Diseases: Utilizing Machine Learning to Identify and Diagnose Underrecognized Medical Conditions

Abstract

Rare diseases, also known as orphan diseases, are conditions that affect a small percentage of the population. Despite their low prevalence, collectively they can have a huge impact on society: one in 17 individuals is estimated to live with a rare diagnosis, which adds up to a burden of rare diseases. Additionally, the rare disease population typically includes families and children who are faced with complex unmanaged diseases that can have considerable personal and financial effects, and who are in urgent need of effective treatments and care. However, to get treatment for these complex and life-altering conditions, an enormous amount of time and money is spent when simply trying to figure out the correct diagnosis. Rare diseases display a vast array of manifestations, with some degree of overlap with common diseases. On top of this, they are hard to identify, given that healthcare providers do not have patient cohorts large enough to learn from their peers or from electronic health records. Thus, rare diseases are very difficult to diagnose and can have significant diagnostic delays, which can even lead to misdiagnoses. Clinicians, however, are aware that there exists a large number of underrecognized medical conditions, and that many efforts are underway to improve rare disease diagnoses both at a clinical level and at a large scale. Artificial intelligence offers immense potential to assist clinicians in meeting this unmet need. This paper reviews and discusses the applications of AI to rare disease diagnostics.