As a child, my father had a deadly, unidentified illness. My mother drove him furiously from one hospital to the next as the paralysis ascended throughout his body, yet doctor after doctor offered only blank stares.
Finally, one physician asked the crucial question: “Have you checked for Guillain-Barré?” As my mother frantically researched the condition, the symptoms aligned perfectly.
My father was lucky to finally get an answer. Even a few days’ wait could have made the outcome tragic.
My dad is one of 30 million in the US suffering from a rare disease. Impacting only a small percentage of people, these diseases are inadequately studied and difficult to diagnose, with patients averaging 5.6 years before proper diagnosis.
AI can change all of that.
Currently, rare disease identification requires incredible time investment and effort. With over 7,000 rare diseases worldwide, though, the solution cannot be grounded exclusively in training each physician to single-handedly diagnose thousands of diseases they might never even encounter.
Technology can expedite these life-saving diagnoses. Genome testing advancements accelerate the identification of rare genetic diseases, connecting over 25% of undiagnosed patients to answers. Meanwhile, cutting-edge algorithms are analyzing disease symptoms. PhenIX, for example, determines macromolecular structure based on algorithms like heavy-atom search and maximum-likelihood molecular replacement, providing new insight into our biological systems and potential therapeutics. Additional methods are needed to assist the remaining population, but AI is a promising tool for families in the same position mine was in.
Building off success from other fields, the medical community must prioritize the creation of a specialized AI database that catalogs rare diseases and their characteristics, including symptoms, treatment methods, and vulnerable populations. AI has not developed to the point where it can accurately diagnose and treat thousands of diseases, but this database could serve as a resource for physicians to consult when facing an unfamiliar group of symptoms.
Over time, AI can be trained to identify clusters of symptoms prevalent in certain rare diseases, then present a list of possible diseases for the physician to inspect. This effectively reduces the number of diseases physicians need to investigate from thousands to a dozen, massively shortening physicians’ time commitment and patients’ agonizing waits, while increasing the probability of a correct diagnosis.
In order for this AI dataset to work effectively, physicians around the world must collaborate, inputting and corroborating disease data for uncommon diseases they encounter.
AI gets a bad rap for many good reasons: replacing jobs, using environmental resources, introducing hallucinations — those “alternative facts” that plague us — into places as varied as public discourse, court records, and even the news.
But its potential for people struggling to identify illnesses outweighs all of that. At least that’s what my family’s ordeal with a rare disease taught me.
Rather than a 5-year wait, AI has the potential to make the diagnostic process more streamlined and effective, offering patients the diagnoses that may save or prolong their lives. For my family, one doctor’s peripheral knowledge of a rare disease saved my dad’s life. But patients should not have to move from doctor to doctor in pursuit of someone with the insight they need. An AI tool could empower individual doctors to make connections far beyond their own experiential knowledge, tapping into a shared knowledge base of other physicians across the world.
We have a duty to alleviate human suffering. Let’s utilize every resource at our disposal in pursuit of that aim.