Mathematical analysis of DICOM CT datasets: can endograft sizing be automated for complex anatomy?

Objectives: To validate the use of a novel mathematical algorithm applied to digital imaging and communication in medicine (DICOM) computed tomography (CT) data to automate the generation of complex endovascular graft planning.

Methods: An algorithm was developed enabling the creation of patient-specific mathematical model based upon DICOM CT data to allow for detailed efficient geometric analysis with repeatable results. This algorithm was applied to high resolution DICOM CT datasets of 15 patients, selected at random from 350 patients with aneurysms involving the visceral arteries. The longitudinal and rotational relationships of the visceral vessels were determined by the algorithm. For comparison purposes, the same measurements were acquired manually using centerline of flow software by a blinded investigator. The distance between the renal arteries, and location of the renal origins calculated with each method were then compared.

Results: Automated results were readily created for all 15 randomly selected patients. The measured versus calculated mean inter-renal artery distances were exceptionally close, differing by a mean of only 1.3 mm with a maximal range of 3.0 mm. The rotational position of the renal origins differed by only 10.5 degrees of arc (21 clock-face minutes) on average and by 32.5 degrees in the worst case.

Conclusions: The generation of an automated mathematical model to represent complex aortic geometry is feasible and reproducible in the context of high-resolution CT data. This process has been validated in 15 patients, where results corresponded with manual measurements that were used to successfully implant customized devices that accommodate the imaged vessels. Additional attributes include the expression of the 3D aorta in a compact form (on the order of kilobytes) for purposes of data storage, transfer, and other manipulations.

Clinical Relevance: The construct of mathematical representation of patient specific anatomy from CT data is both feasible and applicable, allowing for automated endograft device design even in the setting of markedly tortuous anatomy where the repair must incorporate major aortic branches. The process has been validated in 15 patients, where results corresponded with manual measurements that were used to successfully implant customized devices that accommodate the imaged vessels. The application of this technology must now be studied in a prospective manner and incorporated into a system users may readily apply to patients undergoing evaluation. Additional attributes of a mathematical representation of the arterial tree include the 3D expression from a very compact set of parameters (on the order of kilobytes) for purposes of data storage, transfer, and other manipulations.