Evaluating AI Diagnostic Accuracy for Early-Stage Guillain-Barré Syndrome
Kavya Mohanasundaram
09/09/2026
Guillain-Barré syndrome (GBS) is a rare neurological disorder in which the immune system attacks the peripheral nervous system, which can cause paralysis and sometimes respiratory failure. Clinicians miss it over 51% of the time, and 16% of patients get the wrong diagnosis, leading to a delay in treatment up to 63 days. AI symptom checkers generate ranked diagnoses from symptom inputs, but research shows they only correctly identify rare diseases 24.2% of the time. No study has tested AI tools on GBS detection against mimic conditions compared to human benchmarks. This study fills that gap.
This study tested how accurately AI tools detect early-stage GBS using only symptom information, compared to published clinician misdiagnosis rates.
60 vignettes were built from published case reports, 30 confirmed GBS and 30 non-GBS mimic cases, describing only early symptoms before any testing. All 60 were run through three tools: a custom ML model, WebMD Symptom Checker, and Isabel DDx.
WebMD got zero out of 30 correct, a 0.0% sensitivity. Isabel DDx got 2 out of 30, a 6.7% sensitivity. The custom ML model got 17 out of 30, a 56.7% sensitivity, not statistically different from the human benchmark of 49%. However, custom ML incorrectly flagged GBS in 63.3% of non-GBS cases. All three tools scored AUC values at or below 0.5, meaning none performed better than random chance.
Current AI symptom checkers cannot reliably detect GBS. The custom ML model shows promise but needs major specificity improvements before clinical use