Differentiation between thrombus formation and degeneration after biological heart valve replacement based on computed tomography angiography

Jonas Hein (Bad Krozingen)1, R. Schmitt (Bad Krozingen)1, J. Brado (Bad Krozingen)1, M. Apweiler (Bad Krozingen)1, M. Hein (Bad Krozingen)1, D. Westermann (Bad Krozingen)1, P. Ruile (Bad Krozingen)1, P. Breitbart (Bad Krozingen)1

1Universitäts-Herzzentrum Freiburg / Bad Krozingen Klinik für Kardiologie und Angiologie Bad Krozingen, Deutschland


The differentiation between thrombus and pannus (=degeneration) on valvular leaflets in patients with structural valve prosthesis dysfunction is essential for treatment decisions. The aim of this retrospective monocentric study is to determine a cut-off value of Hounsfield units (HU) that can be used in case of visual deposits for differentiation between pannus and thrombus after biological heart valve replacement.



All patients who received a biological heart valve prosthesis with subsequent computed tomography angiography (CTA) at any time after implantation between 03.2015 and 09.2023 were included in this study. A total of 404 CTAs from 344 patients were analyzed.

The thickest visual deposit was used to determine the HUs for each leaflet. The maximum number of leaflets per prosthetic valve equals three. The diagnosis of valve thrombosis was assigned in two scenarios: 1) if the deposits were decreasing after oral anticoagulation, or 2) if a visual deposit appeared within three months after valve replacement. The diagnosis of pannus was also assigned in two scenarios: 1) if visual deposits appeared under appropriate OAC treatment, or 2) if the visual deposits were stable or progressed after the initiation of OAC treatment. 

Differences between groups were assessed by Mann-Whitney tests. Receiver operating characteristic curves (ROC) analyses were done to assess the area under the ROC curves (AUROC) for HUs to discriminate between pannus and thrombus. Best performing cut-off values were then derived from the ROC curves and used to determine discriminatory accuracy.



Valve thrombosis was assigned in 168 prosthetic leaflets, assessed in 128 CTAs (15,7 ± 27,4 days after implantation) from 95 patients (63,2% women, mean age 82.4 ± 5.8 years). Pannus was assigned in 221 prosthetic leaflets which stem from 109 CTAs (6,1 ± 3,4 years after implantation) from 79 patients (50,6 % women, mean age 72.4 ± 8.4). 78 patients were excluded due to poor image quality or unmeasurable valve deposits. Visual deposits in 107 CTA from 92 patients could not be assigned to any group.

The Hounsfield Units of the pannus group were significantly higher compared to the thrombosis group (Median 151.0, IQR 42 HU vs. 86.5, IQR 23 HU), U (N1=168, N2=221) = 389; z = -15.40; p < 0.001. HUs had an AUROC of 0.96 (95 % CI: 0.94-0.98; p<0.001) to discriminate between valvular thrombosis and pannus. The best performing cut-off for the diagnosis of pannus was ≥ 127,5 HU. Using this cut-off led to a sensitivity of 81,9 % and a specificity of 95,2 % for the diagnosis of pannus in our cohort.



In this systematic retrospective analysis of patients after bioprosthetic valve replacement the HUs of visual valvular deposits in CTAs had considerable accuracy in discriminating between valvular thrombus and pannus. A cut-off values of ≥ 127.5 HU indicate a pannus, while a lower value is associated with thrombus formation. After external validation, these data may be used to define treatment paths for patients with structural valve prosthesis dysfunction and visual leaflet deposits.

Diese Seite teilen