Cracking the code: a head-to-head comparison of expert clinicians and artificial intelligence in diagnosing rare diseases

Abstract

BACKGROUND: Patients with rare diseases often face prolonged diagnostic journeys due to the low prevalence and diverse clinical presentations of these conditions. In Germany, specialized centers for rare diseases, established at university hospitals, offer targeted diagnostic and therapeutic care to reduce diagnostic delays. Tools like "Isabel Healthcare" can support clinicians by streamlining the differential diagnosis process and aiding in the accurate identification of rare conditions. RESULTS: The study included 100 patients with a mean age of 44 years. "Isabel Healthcare DDx companion" and the interdisciplinary case conferences generated a total of 727 diagnosis suggestions. Among the top ten diagnoses suggested by "Isabel Healthcare DDx companion", 28% matched at least one diagnosis identified during the interdisciplinary case conferences. The diagnoses suggested as "more likely" by "Isabel Healthcare DDx companion" showed a higher correlation with the differential diagnoses and procedures identified during the interdisciplinary case conferences, suggesting a potential alignment in clinical decision-making processes. CONCLUSION: This study has demonstrated the potential of the differential diagnostic tool "Isabel Healthcare DDx companion" to assist in patient diagnosis. However, discrepancies between the tool's findings and expert decisions suggest that, although it can support clinicians in decision-making, its independent effectiveness may be limited by accurately filtering and interpreting the essential medical history required for a precise diagnosis.

Ähnliche Arbeiten