Volumetric analysis of cerebral hypoperfusion on SPECT: validation and reliability

A method of volumetric analysis of hypoperfusion on 99mTc-labeled hexamethylpropylene amine oxime SPECT has been developed. This analysis integrates both the size and severity of perfusion reduction, yielding an equivalent volume of cortical tissue having zero blood flow, or the effective hypoperfusion volume. This study aimed to validate the methodology in vitro using the Hoffman brain phantom and two different camera systems, to examine the relationship between spatial resolution and accuracy of volume measurement and to assess the interobserver variability in SPECT studies of stroke patients.

Simulated cortical lesions of three different sizes were sequentially incorporated into the Hoffman brain phantom and imaged using both single- and triple-head camera systems. For each system and for each lesion size, successive acquisitions were performed using three progressively larger radii of camera rotation. The hypoperfusion volume for each study was measured three times by a blinded observer, and the percentage difference from the true lesion volume was then calculated. SPECT studies of 32 stroke patients were independently analyzed by two blinded observers.

At the smallest radii of rotation, mean (s.d.) percentage difference between observed and true volumes was 0.90% (4.80%) for the triple-head and 4.50% (11.58%) for the single-head system. The degree of overestimation was similar for both systems. Percentage overestimation was strongly associated with radius of rotation (B = 0.71 +/- 0.32, p = 0.04 for the triple-head system; B = 1.26 +/- 0.55, p = 0.03 for the single-head system) but not with lesion size. The mean difference between hypoperfusion volumes of stroke patients obtained by two observers was 0.045 (3.240) cm3, which was not significant.

This study has shown that this technique of volumetric analysis of regional hypoperfusion on SPECT is both highly accurate and reproducible between two different camera systems. The degree of overestimation chiefly relates to diminishing spatial resolution. Despite some element of subjectivity, the interobserver variability is negligible.