Investigative radiology, Volume 32, Issue 9, 1 1 1997, Pages 540-549 Quantitative assessment of the presence of a single leg separation in Björk-Shiley convexoconcave prosthetic heart valves. Vrooman HA, Maliepaard C, van der Linden LP, Jessurun ER, Ludwig JW, Plokker HW, Schalij MJ, Weeda HW, Laufer JL, Huysmans HA, Reiber JH
Rationale and objectives
The authors developed an analytic software package for the objective and reproducible assessment of a single leg separation (SLS) in the outlet strut of Björk-Shiley convexoconcave (BSCC) prosthetic heart valves.
Methods
The radiographic cinefilm recordings of 18 phantom valves (12 intact and 6 SLS) and of 43 patient valves were acquired. After digitization of regions of interest in a cineframe, several processing steps were carried out to obtain a one-dimensional corrected and averaged density profile along the central axis of each strut leg. To characterize the degree of possible separation, two quantitative measures were introduced: the normalized pit depth (NPD) and the depth-sigma ratio (DSR). The group of 43 patient studies was divided into a learning set (25 patients) and a test set (18 patients).
Results
All phantom valves with an SLS were detected (sensitivity, 100%) at a specificity of 100%. The threshold values for the NPD and the DSR to decide whether a fracture was present or not were 3.6 and 2.5, respectively. On the basis of the visual interpretations of the 25 patient studies (learning set) by an expert panel, it was concluded that none of the patients had an SLS. To achieve a 100% specificity by quantitative analysis, the threshold values for the NPD and the DSR were set at 5.8 and 2.5, respectively, for the patient data. Based on these threshold values, the analysis of patient data from the test set resulted in one false-negative detection and three false-positive detections.
Conclusions
An analytic software package for the detection of an SLS was developed. Phantom data showed excellent sensitivity (100%) and specificity (100%). Further research and software development is needed to increase the sensitivity and specificity for patient data.
