Copper filtration reduces radiation dose while maintaining image quality in cerebral flat-detector CT.

BACKGROUND

Flat-detector computed tomography (FDCT) is increasingly used for peri-interventional cerebral imaging but is associated with a relatively high radiation exposure. Copper (Cu) filtration may reduce radiation dose. However, its impact on cerebral image quality and intracranial hemorrhage detection remains unclear.

METHODS

In this retrospective single-center study, 31 patients undergoing neurointerventional procedures with intraindividual FDCT acquisitions with and without Cu filtration were analyzed. Quantitative image quality was assessed using contrast-to-noise ratio (CNR). Qualitative image analysis and intracranial hemorrhage detection were independently evaluated by two readers blinded to Cu filtration status using five-point scales.

RESULTS

Cu filtration resulted in a significant radiation dose reduction of 25.9% for both entrance skin dose (145.19±13.18 mGy vs 195.89±18.05 mGy) and dose-area product (41.52±3.77 Gy·cm² vs 56.01±5.16 Gy·cm²), respectively (P<0.001). No differences in CNR were observed for unfiltered vs Cu-filtered FDCT (basal ganglia: 4.73±2.04 vs 4.37±1.99, P=0.419). Qualitative image ratings were similar between techniques (supratentorial cortex: 2.27±0.66 vs 2.08±0.75, P=0.089), with very good inter-reader agreement (κ=0.86; 95% CI: 0.80 to 0.91). All intracranial hemorrhages were correctly identified by both techniques. Correct exclusion of intracranial hemorrhage was 15/16 with Cu filtration and 14/16 without Cu filtration, without statistically significant difference. Differences were limited to hemorrhage mimics (n=2) and minor variations in diagnostic confidence without affecting binary classification.

CONCLUSION

Cu filtration in cerebral FDCT enables substantial radiation dose reduction while preserving image quality and intracranial hemorrhage detection, supporting its clinical implementation as a practical dose optimization strategy for peri-interventional imaging.