Sustainable four-dimensional cardiac CT: Minimizing data storage via reduced matrix size with optimized...

The increasing volume of four-dimensional cardiac CT data complicates transcatheter aortic valve repair (TAVR) planning, particularly regarding data storage. This study evaluates a modified 256 × 256 reconstruction method for TAVR-CT that reduces storage requirements while maintaining aortic valve measurement accuracy.

A retrospective analysis was performed on 75 TAVR-CT scans obtained using the dual-source energy-integrating detector CT. The modified 256 × 256 reconstruction used a sharper Bv44 kernel (iterative construction strength 4) and narrowed the field of view to 80 %. We assessed virtual basal ring (BR) areas using reference 512 × 512, standard 256 × 256, and modified 256 × 256 images. Five technologists-three from the CT team and two from the interventional radiology (IR) team-manually evaluated the BR areas. The measurements were averaged within each team. The intraclass correlation coefficient (ICC), bias, and consistency in device selection were assessed.

Among 75 patients (48 women, median age 85), the mean BR area was 416.2 ± 77.9 mm². The mean bias between the modified 256 × 256 and reference 512 × 512 images was -2.5 mm² (95 % CI: -5.2 to 0.3), significantly smaller (P = 0.04) than the bias between standard 256 × 256 and reference (4.1 mm², 95 % CI: 0.7 to 7.5). BR measurements demonstrated excellent reproducibility (ICC = 0.996-0.997), with limits of agreement (LOA) for the modified 256 × 256 ranging from -13.0 to 12.2 mm² for the IR team and -14.5 to 11.6 mm² for the CT team, comparable to the inter-team LOA of -18.1 to 13.1 mm². Valve selection consistency was 95 % (Cohen's Kappa = 0.92).

The modified 256 × 256 reconstruction method effectively preserved BR measurements and valve selection reliability.

The proposed 256 × 256 method could cut data storage while ensuring accurate valve selection.