OBJECTIVE Bowel uptake is a frequent artefact in myocardial perfusion scans (MPS) that can impede the assessment of the inferior wall, necessitating repeat acquisitions. This study is a retrospective analysis of MPS reconstructed with different algorithms to see whether corrections for scatter or depth-dependent loss of resolution may help differentiate apparently joined uptake in the inferior wall of the left ventricle from that in the bowel. METHODS A total of 79 sequential stress MPS acquired 1 h after injection of 600 MBq Tc-tetrofosmin on a Siemens Symbia T were reconstructed using filtered back projection (FBP), iteratively with ordered subset expectation maximization (OSEM) and iteratively with three-dimensional collimator beam modelling (Siemens Flash3D), Flash3D with scatter correction (SC) and/or attenuation correction (AC). Single-photon emission computed tomography studies were classified by a nonblinded observer as follows: I, no artefact; II, abdominal activity just touching the inferior wall (not interfering with interpretation); III, abdominal activity covering some myocardium (interpretation still meaningful); and IV, nondiagnostic scan for large parts of the myocardium. RESULTS The following numbers of scans were placed in categories I/II/III/IV for each algorithm: FBP 58/15/2/4; OSEM 54/16/5/4; Flash3D 54/20/5/0; Flash3D+SC 61/17/1/0; Flash3D+AC 21/36/17/5; and Flash3D+SC+AC 29/39/9/2. These differences were significant (Friedman test, P<0.0001). CONCLUSION We conclude that the use of Flash3D+SC produces the least number of nondiagnostic or difficult-to-interpret scans.