Singular value-guided similarity filter improves detection of vessels in low-dose dynamic CT angiography: application to DIEP flap studies

: Dynamic CT angiography provides much more information compared to static 3D CT angiography, in several clinical applications. The prolonged exposure, on the other hand, results in higher radiation dose for to the patient. Care should be taken when designing noise reduction algorithms so that temporal profiles and spatial resolution is not significantly compromised. In particular, in deep inferior epigastric perforators (DIEP) flap studies, the temporal enhancement of small abdominal subcutaneous and intramuscular vessels needs to be evaluated from the surgeons to plan breast reconstruction. We propose a new approach to exploit the data redundancy in the temporal direction and successfully reduce noise via singular value decomposition. The similarity of voxels temporal profiles is then measured on the noise-reduced dataset and a spatial smoothing is performed with an additional edge-preserving weight coming from the noise-reduced tMIP. We named our filter the singular value decomposition guided similarity (SVGS) filter, and compared it to the TIPS (time-intensity profile similarity) filtered data and to the highly constrained filtered backprojection (HYPR) processed data. This institutional review board-approved study included 17 DIEP flap clinical cases, that were retrospectively processed with HYPR, TIPS and SVGS and then compared both qualitatively and quantitatively. Contrast-to-noise ratio (CNR) was used for quantitative evaluation, while a blinded subjective evaluation was performed by an experienced radiologist. We reported significantly improved results with the SVGS filter when compared to the other approaches, both in terms of CNR, as well as with the subjective clinical evaluation. The results over 17 clinical cases suggest that a potential dose reduction of a factor of 9 could be achieved, while preserving the same image quality. In conclusion, we developed an efficient algorithm for dynamic CT angiographic data processing, which successfully enables a reduction of spatial and temporal noise, while preserving and enhancing the visualization of small vessels. We demonstrated the clinical usefulness of the SVGS for the DIEP flap application. The proposed algorithm can be employed to significantly improve image quality and diagnostic acceptability at same dose level or alternatively to reduce radiation dose, to which extent needs to be evaluated in future studies.