Eva Maria Gassner MD, Presenter: Nothing to Disclose

Razvan Ionasec PhD, Abstract Co-Author: Employee, Siemens AG

Bogdan Georgescu PhD, Abstract Co-Author: Employee, Siemens AG

Sebastian Vogt, Abstract Co-Author: Employee, Siemens AG

U. Joseph Schoepf MD, Abstract Co-Author: Speakers Bureau, Bracco Group Speakers Bureau, General Electric Company Speakers Bureau, Bayer AG Speakers Bureau, TeraRecon, Inc Medical Advisory Board, Bracco Group Medical Advisory Board, General Electric Company Medical Advisory Board, Bayer AG Research grant, Bayer AG Research grant, Bracco Group Research grant, General Electric Company Research grant, Bayer AG Research grant, Siemens AG

Dorin Comaniciu, Abstract Co-Author: Employee, Siemens AG

Philip Costello MD, Abstract Co-Author: Research grant, Bracco Group Research grant, Siemens AG Research grant, General Electric Company

et al, Abstract Co-Author: Nothing to Disclose

The usefulness of cardiac CT to assess the aortic valve opening area (AVA) has been exhaustively documented. However, manual valve planimetry is cumbersome and time-consuming. We evaluated the accuracy and time-effectiveness of automated model-based AVA computation compared to manual planimetry.  

Retrospectively ECG-gated cardiac CT data of 32 patients scanned with dual-source CT (n=21) or 64-slice CT (n=11) were included. Data were reconstructed at 10% increments across the cardiac cycle with 1.5mm section thickness and 1mm increment. Two independent observers performed manual planimetric measurements by tracing the maximal systolic orifice on double oblique short axis multiplanar reconstructions. The same data were then analyzed using an automatic 4D aortic valve model (Siemens Corporate Research). The model contains the surface of the aortic root, valve leaflets, the commissures, hinges and coronary ostia, and allows for morphologic and dynamic quantification of the aortic valve. The model is fitted to CT data based on discriminative learning methods and incremental search. The leaflets’ geometries during maximal opening define the course of the free margins. The encompassed AVA can be computed as a surface integral. Data were analyzed using linear regression and Bland Altman plots. Interobserver and intermethod variances were calculated. Analysis times for both methods were recorded.  

Mean AVA by CT planimetry was 3.62±1.21cm². Mean AVA derived from the model was 3.74±1.34cm². Excellent correlation was found between planimetric and automated quantification (r=0.963, p<0.0001). Bland Altman plots revealed a systematic bias of 0.12±0.38cm². Intermethod variance did not differ significantly from interobserver variance (0.28 vs 0.25cm², p>0.05), placing 82% of model measurements between user measurements. Mean analysis time was significantly (p<0.05) reduced for model-based measurements (mean 125sec), compared with manual planimetry (mean 230sec).

Our 4D aortic valve model allows automated, patient specific morphologic and dynamic quantification of AVA. Measurement results are within the inter-observer variance of manual planimetry.    

Quantification of AVA derived from an aortic valve model enables fast, accurate assessment in excellent agreement with manual planimetry and has the potential to improve cardiac imaging workflow.

Gassner, E, Ionasec, R, Georgescu, B, Vogt, S, Schoepf, U, Comaniciu, D, Costello, P, et al, , Performance of a Dynamic Aortic Valve Model for Quantification of the Opening Area at Cardiac MDCT: Comparison to Manual Planimetry.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6011906.html