A method for calibrating a nuclear medicine tomography detector module using principal component analysis is based on the idea that calibration beam data lies on a one-dimensional path within the higher dimensional dataspace of output data. The module includes a weighted multiplexing circuit that generates a small number of multiplexed signals for each photon event. Calibration data for the module is generated and analyzed using several iterations of principal component analyses, to filter scattering events, noise, and other spurious signals. The direction of depth-of-interaction information has been found in the high-dimensional dataspace to be indicated by the primary principal component of the calibration data. The primary principal components, principal components from filtered datasets, intermediate thresholds, and DOI or inner product values are recorded for calibrating the module.
