Correct spatial normalization of myocardial perfusion SPECT improves detection of multivessel coronary artery disease

This study sought to improve the detection of multivessel coronary artery disease (CAD) with the use of rest and exercise single photon emission computed tomography (SPECT) perfusion scintigraphy by developing a processing scheme, which provides proper regional normalization of the images for interpretation. When SPECT perfusion images are interpreted, one area of myocardium serves as "normal." We hypothesized that if this "normal" region changes location from rest to stress, the stress images must be adjusted for proper interpretation. By taking into account the level of tracer activity in this "normal" area on the resting images, we could more accurately identify patients with multivessel CAD. Methods and results Dual-isotope rest (thallium 201) and exercise dobutamine or adenosine stress (technetium 99m sestamibi) perfusion SPECT studies were examined in 258 patients with 2- or 3-vessel CAD on coronary arteriography performed within 6 months of each other (mean interval, 19 days). If a shift in regional location of the "normal" segment from rest to stress was present, the images were (1) interpreted in the usual fashion for the number of vessels with ischemia (PRE-NORM) and (2) reinterpreted after quantitative normalization (ie, adjusting the display window until the intensity of the "normal" segment was matched at rest and stress [POST-NORM]). Interpretation was performed with blinding to arteriographic results. An angiographic stenosis was defined as luminal diameter stenosis greater than 50%. Three control groups comprising (1) single-vessel CAD (n = 119), (2) no significant angiographic CAD (n = 118), and (3) a normalcy group of low pre- and post-test probability of CAD (n = 44) were also studied to determine the incidence of false-positive results induced by the renormalization technique. A shift in the "normal" segment occurred in 81 studies of 258 patients (31%), 80 of which were read as abnormal PRE-NORM (sensitivity, 99%); however, for their 216 stenosed vessels, only 143 were detected PRE-NORM (vessel sensitivity, 66%; accuracy, 65%). The mean POST-NORM change in the display was 11%. POST-NORM, all 82 patients' studies were interpreted as abnormal (sensitivity, 100%), and 196 of 216 vascular territories were abnormal (vessel sensitivity, 91%; accuracy, 83%; both P <.0001 vs PRE-NORM). In the single-vessel disease, no significant disease, and normalcy groups, 19 of 119, 15 of 118, and 11 of 44 patients, respectively, demonstrated a shift in the peak pixel location. However, there were no significant changes in single-vessel sensitivity, angiographic specificity, or normalcy in these patients.

With multivessel CAD, the "normal" region on SPECT often changes in location from rest to stress, potentially masking the extent and severity of multivessel ischemia. Renormalization of the images to match their resting level before image interpretation allows diagnosis of contralateral ischemia and strikingly improves the detection of multivessel CAD, without a substantive loss in specificity.