What question did this study set out to answer?

This research aims to evaluate the diagnostic performance of AI models in recognizing paroxysmal nocturnal hemoglobinuria (PNH) using clinical vignettes.

December 8, 2025BloodOpen Access

AI-assisted diagnosis of paroxysmal nocturnal hemoglobinuria

View Full Paper

Ask AI

Bookmark

This research aims to evaluate the diagnostic performance of AI models in recognizing paroxysmal nocturnal hemoglobinuria (PNH) using clinical vignettes.
Systematic review of case reports documenting PNH diagnosis confirmed by flow cytometry or Ham testing.
Creation of standardized clinical vignettes from 68 patients with documented diagnostic delays.
Independent evaluation of diagnostic accuracy by GPT-4.5, Claude, and Gemini using complete and modified vignettes.
Statistical analyses performed included Mann-Whitney U tests and Bonferroni adjustments for multiple comparisons.
GPT-4.5 achieved 89.7% accuracy in identifying PNH among top-five differential diagnoses from complete vignettes.
GPT-4.5 ranked PNH as the top diagnosis in 70.6% of cases, outperforming Claude and Gemini.
AI accuracy was stable when single diagnostic features were removed, but dropped significantly when dual hemolytic indicators were absent.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

AI models improve diagnostic accuracy for paroxysmal nocturnal hemoglobinuria in clinical vignettes, highlighting key diagnostic indicators.

Cite This Study

Bowen et al. (2025) studied this question.

synapsesocial.com/papers/69362f574fa91c937236dacd https://doi.org/10.1182/blood-2025-4670

View Full Paper

Ask AI

Bookmark

Implication

AI models improve diagnostic accuracy for paroxysmal nocturnal hemoglobinuria in clinical vignettes, highlighting key diagnostic indicators.

AI-assisted diagnosis of paroxysmal nocturnal hemoglobinuria

Key Points

Discussion

Implication

Cite This Study

AI-assisted diagnosis of paroxysmal nocturnal hemoglobinuria

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider