Systematic Review: Fairness Reporting Lacking in Studies of Commercial Stroke CT AI Tools

Only one of thirteen studies on regulator-cleared AI for acute stroke CT reported subgroup performance, showing lower PPV for female vs male (67.3% vs 72.7%, p=0.02) and Black vs non-Black patients (66.7% vs 73.2%, p=0.005).

• Thirteen studies from nine vendors and four CT tasks were included; only one reported demographic subgroup performance, and one reported performance by site, scanner, or protocol.
• No included study described the demographic composition of the AI tool's training data; four studies acknowledged potential bias or data gaps without quantifying them.
• The review does not conclude the evaluated systems are inequitable—only that missing subgroup reporting limits fairness assessment.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.