10-year-old son's height 122cm, doctor says 'It's okay'... AI found clues to rare disease

Artificial intelligence is emerging as a powerful ally in the swift diagnosis of rare diseases, a process that has historically been lengthy and arduous, often spanning over five years. Medical professionals are increasingly leveraging AI tools, including facial analysis applications, general-purpose AI chatbots, and FDA-cleared clinical AI, to identify potential rare or difficult-to-diagnose conditions more rapidly.

These AI tools primarily function by suggesting a range of possible diseases for medical experts to consider, thereby narrowing down the diagnostic possibilities. The Wall Street Journal highlighted the case of Oliver, a child whose delayed development and consistently low height went undiagnosed for years despite medical reassurance. His mother, Rachel Hinken, used a facial analysis app called Face2Gene, which identified a type of TRPS (a rare genetic disorder) as a likely candidate. Subsequent genetic testing and specialist consultations confirmed the diagnosis for both Oliver and his mother.

Experts acknowledge the transformative potential of AI in this field. Dr. Sarah Dykman, a neurologist specializing in nervous system disorders, expressed astonishment at AI's ability to uncover clues for rare diseases. She noted that patients with conditions like POTS (Postural Orthostatic Tachycardia Syndrome), a disorder of the autonomic nervous system, are now seeking diagnosis within months of symptom onset, thanks to AI chatbots flagging potential issues. This contrasts sharply with previous cases where patients suffered for decades without a diagnosis.

While AI demonstrates significant promise, particularly in pattern recognition by connecting physical characteristics in medical images with vast databases of disease information, human oversight remains crucial. Dr. Xiao P. Feng, a clinical geneticist, cautioned that AI can still make errors and is best suited for suggesting disease candidates, explaining complex medical terms, and organizing research. He emphasized that human intervention is indispensable for synthesizing information and making definitive medical judgments. Furthermore, the accuracy of AI is contingent on the quality and quantity of training data, which is often scarce for rare diseases, necessitating ongoing efforts to digitize and expand these datasets.