|
|
|
Development and clinical validation of normal Tc-99m
sestamibi database: Comparison of 4D-MSPECT to CEqualTM
Objective: The purpose of this study was to
develop and validate a normal database for quantitative perfusion
imaging with Tc-99m Sestamibi using the normal database generator of
4D-MSPECT and to compare it to a previously validated program, CEqualTM.
Methods: Gender specific normal databases (180°
and 360° reconstructions of same acquisitions) were created from 30
low-likelihood normal males and 30 females using the normal database
generator of 4D-MSPECT. Following automatic rendering of the
endocardial and epicardial LV surfaces, myocardial activity is sampled
using a cylindrical spherical coordinate system where each polar map
ring is equidistantly sampled along the mid-myocardial surface from
base to apex. Using ROC analysis regional thresholds in SD below the
normal mean were optimized. Clinical validation was performed using
data from 317 patients, low likelihood normal (=5% pre-test likelihood)
(n=114), angiographically normal (n=36), angiographic coronary stenoses
(=50% stenoses) (n=167). A rest Tl-201/stress Tc-99m Sestamibi imaging
protocol was performed. Patients with prior MI, LBBB and CABG were
excluded. Their were no exclusions based on body habitus or bra size.
The mean age of the normals was 49±16 and the mean age of the patients
was 57±13 years. All studies were analyzed using 4D-MSPECT (180° and
360°) and CEqual (180°).
Results: Normalcy rates were significantly
higher for 4D-MSPECT, 58% for 180° data, 76° for 360° data as compared
to CEqual 34% (p<0.001). The low CEqual normalcy value can be
attributed to breast attenuation that is not accurately represented in
the CEqual female database. Sensitivities (81%, 83%, 77%),
specificities (42%, 42%, 42%) and accuracies (74%, 75%, 71%) for the
identification of patients with CHD were comparable for the three
methods respectively. Sensitivities, specificities and accuracies for
individual vessel stenoses were comparable for 180° and 360° data
analyzed with 4D-MSPECT in comparison to CEqual data. The ROC curves
generated by varying the defect threshold in 4D-MSPECT are presented
for the detection coronary stenoses in each of the vascular territories
and for the detection of disease by patient. Since CEqual does not
allow the defect threshold to be changed, a single
sensitivity/specificity point is plotted on each of the 4D-MSPECT ROC
curves for comparison.
Conclusions: Databases created and optimized
with the normal database generator of 4D-MSPECT provide increased
normalcy rates and comparable sensitivities, specificities and
accuracies compared to CEqual for a non-selective clinical patient
population.
|
