SPM comparison of brain SPECT from two groups of normal controls acquired on the same 3-head gammacamera, but 3 years apart, showed significant unexpected differences. Previously we simulated these differences by imposing in one group a 10% decrease in sensitivity in one camera head. Here we describe detection and correction of actual inter-head sensitivity differences in individual scans. Sensitivity differences were detected in the curve of decay-corrected total counts versus acquisition angle. This curve should vary smoothly and be closely approximated by the sum of its first few cosine basis functions. We applied a downhill simplex iterative adjustment to the sensitivity of each head to fit the adjusted curve and the sum of its first 4 cosine basis functions. After multiplicative correction of the SPECT projections in the two groups of normal controls, their reconstruction was repeated. This method reliably detects inter-head sensitivity differences of 1%. Sensitivity differences of up to 25% were detected in the subject scans. SPM analysis of the sensitivity-corrected normal controls from the two different periods showed the differences between them had been eliminated. Inter-head sensitivity differences in multi-head gammacameras, when each head scans a different arc of the acquisition circle and when the sensitivity differences drift, can have serious consequences in quantitative SPECT. We have developed a method that retrospectively detects small changes in inter-head sensitivity in individual SPECT scans on multi-head gammacameras, and that permits correction of the acquired scans. This method promises to be a useful QA tool for quantitative multi-head SPECT.