AI generates synthetic CT angiography from noncontrast head CT for vascular screening

An AI model generated synthetic CT angiography from noncontrast head CT, achieving 92.7% accuracy for detecting intracranial vascular disease. On a 110-patient test set, sensitivity was 86.0% and specificity 98.3%. Radiologists rated the synthetic image quality 4.45 out of 5.

• Developed and validated across five Chinese hospitals using 3,709 patients with head NCCT and paired CTA.
• Differentiation accuracy between intracranial aneurysm and intracranial atherosclerotic stenosis within diseased arteries was 90.7%.
• Retrospective design; external validation performed but no prospective or real-world triage study yet conducted.

Automated summary

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