Validation study of an automated algorithm to calculate the aortic angulations based on the analysis of tortuosity.

GAUDRIC, JULIEN; MARIA TERESA POLITI; FERNÁNDEZ, JUAN M.

Lille

Congreso; Annals of Vascular Surgery 60 (2019): 5. French Society for Vascular and Endovascular Surgery; 2019

French Society for Vascular and Endovascular Surgery

Objectives: The aortic angulations are associated withthe risk of endoleak after EVAR. All the current methodsof measurement of the angulations are based on a twodimensionalvisual analysis. However, aortic reconstructionsfrom CT acquisitions allow three-dimensional analysisof data. The aim of this study was to show that it ispossible to automatically identify the significant points ofcurvature of the central line (CL) using an algorithm.Materials and methods: An algorithm was developed toanalyze point by point the radius of curvature of the aorticCL. The CL was generated semi automatically from fivenormal thoracic aortic CT-scans and the CT-scans offour thoracic aortic aneurysms. The three-dimensional coordinatesof each point of the CL were extracted and thevariations of the radius of curvature were calculated. Thepoints presenting the smallest radius of curvature werethus identified. These points were then linked together bylines whose angles were calculated. The localization ofthe points of inflection of the CL was also determined twicevisually by two independent observers and the angulationsof the aorta were measured as it is in clinical practice. Thecoordinates of the identified points were compared betweenthe observers and with the calculation algorithm.Results: For each CT-scan five to seven points wereidentified. The method of Bland and Altman was used forinter- and intra-observer comparison of the measurementsand to compare the results of the visual and mathematicalmeasurement techniques. In all the cases the highlighteddifferences were all within the limits of agreement (1.96SD). The average angles measured by the observers orcalculated by the algorithm were comparable for thenormal aortas (127.5 vs. 131.8 (p¼0.53), respectively)and for the aneurysmal aortas (121.2 vs. 124.3(p¼0.43), respectively). The simplification of the centralline in straight lines was realistic since the average distancebetween the line joining together each point andthe CL was 0.073 mm for the normal aortas and 0.085mm for the aneurysmal aortas when determining mathematicallythe points of inflection.Conclusion: The mathematical localization of the pointsof inflection of the CL is efficient when compared to visualmeasurement. The coordinates of the points of inflection ofthe CL can be obtained from the radius of curvature of theCL.