Intragastric meal distribution during routine gastric emptying scintigraphy: Validation of visual qualitative and semi-automated quantitative analysis

Objectives: Abnormal fundal accommodation (FA) may potentially explain patient dyspeptic symptoms. Currently, there are no well-established criteria for diagnosing abnormal FA during routine solid-meal gastric emptying scintigraphy (GES). Goals of this study included: 1) Assessing the consistency among Nuclear Medicine/Radiology Physicians (NMRMDs) of visually assessing FA from images obtained during routine 4-hour GES of a Tc-99m labeled solid meal; 2) Developing computer software for quantitative analysis of FA during GES.

Methods: To assess whether FA can be consistently assessed visually during GES an instruction guide was first developed and sent to four NMRMDs of the NIH Gastroparesis Clinical Research Consortium. Readers first completed an initial training set of studies to learn consensus interpretive criteria (24 patients: 16 normal; 8 abnormal FA). Each NMRMD then interpreted 118 studies: 19 normal volunteers and 99 patients (18 impaired FA, 81 normal). The studies were analyzed using a five-level score: 1=definitely abnormal accommodation; 2=probably abnormal accommodation; 3=possibly abnormal accommodation; 4=probably normal accommodation; and 5=definitely normal accommodation. Mixture distribution analysis (MDA) (Kundel, Acad Radiol 1997(4):1-7) and Kappa agreement analysis were employed. Semi-automated computer software using MatLab was developed to quantitate intragastric meal distribution(IMD) as the ratio of the gastric counts in the proximal half of the stomach after dividing the stomach into two equal halves by bisecting the longitudinal axis of the entire stomach. The NMRMDs assessments of FA were compared to the computer-derived IMD via ROC analysis.

Results: MDA of the 99 patients’ data showed that 11% of cases were easy positives for impaired FA, 62% were easy negatives with remaining 27% a difficult call. Pairwise weighted Kappas among the 4 different NMRMDs to visually assess FA using the 5 levels of normal/impaired accommodation based on the 99 patients’ data averaged 0.38 (fair agreement; 95% CI: 0.25-0.51). Kappa agreement measures among the 4 different NMRMDs for normal vs. impaired accommodation classifications (i.e., 4-5 vs. 1-3) averaged 0.43 (moderate agreement; 95% CI: 0.25-0.62). Overall, NMRMDs achieved a consensus about normal and abnormal FA in 84% of patients studied. Using the semi-automated MatLab software, patients with consensus normal accommodation averaged 0.809±0.083 (SD) proximal to total gastric count ratio compared to 0.447±0.132 (p<0.01) for those with impaired accommodation, defined as 蠅3 of the 4 readers reading the image as impaired. Based on ROC analysis, the AUC was 0.93 and the optimal cutoff was 0.57 for the ratio of proximal to total gastric counts to best discriminate normal versus impaired FA. This yielded a sensitivity of 87% and specificity of 92%.

Conclusion: Gastric FA can be assessed visually during routine GES with good agreement(Kappa) and high consistency(MDA) among trained NMRMDs. Using the time 0 minute image and a division of the stomach into proximal and distal halves, the relative amount in the proximal half (IMD) can be calculated with >0.57 defining normal IMD and FA whereas <0.57 defining abnormal IMD and impaired FA. Based on these results we conclude that normal and abnormal IMD and fundal accommodation can be assessed as a part of a routine solid-meal gastric emptying study. Research Support: NIH Consortium