Cancellation of unnecessary rest imaging (CURI) following separate- day stress Tc-99m mibi tomography: multi-reader evaluation of stress-only scans, and quantitative programs

Aim: In myocardial scintigraphy, performance of a high-dose stress as the first study (HDSF), with follow-up rest scanning has several advantages. Canceling unnecessary rest imaging (CURI) following normal stress studies may improve targeting of resources; CURI could be aided by quantitative software with adequate specificity to distinguish the most normal scans. At one hospital site, CURI has been performed 1000 times in a four-year interval. Methods: We reviewed the frequency and appropriateness of CURI with attention to the use of PERFIT, software for fully-automated fitting to gender-matched rest and stress templates. We compared 2 forms of the quantitative output; (i) the original software (P3D) using 3D region-growing for demarcation of hypoperfusion, and (ii) a polar-plot paradigm (PPP). A consecutive series of 1200 sestamibi tomographic studies dated 3/96 to 5/97 constituted an intense-study set; CURI had been performed in only 95 of 594 HDSF cases (16%) in this early series, but later its use increased to 40%. To characterize the stress/rest study pairs in a random subset of 95 cases, blinded ratings of aligned slices were performed without quantitation by 3 experienced observers. The stress component of these scans and a cohort of 30 corresponding CURI cases then underwent review; the task (identifying blinded stress scans), for CURI, was performed 3 times by each of 6 readers in random order - (i) without quantitation, (ii) with P3D, and (iii) with PPP. Each case was assigned a CURI-score for each technique based on the 6 ratings, and an overall score based on 18 readings. Results: The extent of abnormal myocardium by P3D differed among 3 groups of scans, averaging 0.7% in the rest-cancelled cases, 2.5% in the 78 normal stress/rest cases (3/3 agreement), and 11% in the 17 borderline stress/rest cases (discrepancy among the original 3 readers). Corresponding values for PPP were 3%, 6% and 20% respectively. Overall CURI-scores for these groups differed (p<0.01 by analysis of variance) at 90%, 72% and 16% respectively. Reading with P3D showed the highest CURI-scores and inter-observer agreement of the 3 interpretive protocols; sensitivity for mild scan abnormalities seemed similar with P3D and PPP. In a follow-up series, P3D confirmed the normal status of the rest-cancelled cases, and identified still further cases where CURI could have been employed. Conclusion: CURI can be applied in a substantial proportion of HDSF cases. P3D, due to its high specificity, can assist to confirm the most normal cases and thereby to redirect resources