A method and apparatus are provided for positron emission imaging to calibrate energy measurements of a pixilated gamma-ray detector using energy calibration based on a calibration with a distribution energy signature (i.e., having more spectral features than just a single full-energy peak). The energy calibration can be performed using a deep learning (DL) network or a physics-based model. Using the DL network, a calibration spectrum is applied to either generate the measured-signal values of known energy values (e.g., spectral peaks for spectra of various radioactive isotopes) or the parameters of an energy-calibration function/model.