Perfusion-derived cerebral CT angiography with artificial intelligence iterative reconstruction as a replacement for...

To investigate the feasibility of replacing routine cerebral CT angiography (CTA) with CT perfusion (CTP)-derived cerebral CTA by using artificial intelligence iterative reconstruction (AIIR) for acute ischemic stroke (AIS) patients.

Patients with AIS undergoing cerebral CTP and CTA were prospectively collected. CTA images derived from CTP at arterial phase (100kVp/150mAs) were reconstructed with hybrid iterative reconstruction (HIR, Group A1) and AIIR (Group A2), whereas routine CTA images (100kVp/ref. 200mAs) were obtained with HIR (Group B). Two radiologists independently assessed these three image sets for intracranial artery occlusion, with digital subtraction angiography (DSA) serving as the reference standard. Image noise suppression, vascular edge sharpness, small vessel visibility and overall diagnostic acceptability were graded using five-point Likert scales. Noise and contrast-to-noise ratio (CNR) of internal carotid artery, middle cerebral artery and basilar artery were also measured.

Of the 62 enrolled patients (33-93 y), DSA identified 63 arterial occlusions in 54 patients (87.1%). For occlusion detection, Groups A2 and B provided comparable diagnostic performance (Sensitivity: 93.7% vs. 93.7%, p = 1.000; Specificity: 97.5% vs. 98.0%, p = 0.383; Accuracy: 97.3% vs. 97.7%, p = 0.383). Both Groups A2 and B outperformed Group A1 on all diagnostic metrics (all p < 0.001). For qualitative analysis, Group A2 received significantly higher scores than both Groups A1 and B (all p < 0.001), particularly for small vessel visibility (5 [5,5] vs. 3 [3,3] vs. 4 [4,5]). Additionally, Group A2 exhibited the lowest image noise and highest CNR across all arterial measurements.

Perfusion-derived cerebral CTA with AIIR achieves comparable diagnostic image quality to that of routine CTA for AIS patients.