Abstract
Poster session I
2.14: Volumetric quantitation of left ventricular perfusion and function from myocardial perfusion SPECT: Validation of a new algorithm

https://doi.org/10.1016/j.nuclcard.2006.06.016Get rights and content

Section snippets

Background

We have validated a new quantitation algorithm for left ventricular surface detection (LVSD). The algorithm segments the LV from myocardial perfusion SPECT short axis images and measures its volume. For ECG-gated images, the LV volume is recorded per each gated interval and the ejection fraction is calculated. The new segmentation aims at robust success rates in studies with hot extra-cardiac activity where discerning the LV can be challenging.

Methods

The algorithm detects the preliminary LV center coordinates by combining two methods: 1) The image is thresholded, clusterized and. the selected cluster centroid is noted. 2) The SA slices are summed and a high-order polynomial is fitted to the line profiles of the summed image. The center is calculated by comparing the polynomial extramum points to the result of method (1). The image is sampled spherically around this center. The radial profiles are fitted to Gaussians and their respective

Results

Comparison of EF and EDV measurements by LVSD showed close correlation to the corresponding QGS and ECToolbox measurements. LVSD EF correlation curve to QGS was y=0.9774x+0.024 (r2=0.9148). In comparison, ECToolbox EF correlation curve to QGS was y=0.9022x+0.0491 (r2=0.8949).

Conclusions

The algorithm was able to report reliably EF and EDV according to the comparison with leading applications. The novel centroid detection, sampling and outliers’ removal techniques could potentially improve LV detectability and robustness in face of hot extra-cardiac activity.

References (0)

Cited by (0)

View full text