Dear Editor,rnSince calcium imaging can simultaneously record the cal-cium signals of several hundred neurons,it has been a powerful tool in measuring the functional activity of neural populations.However,calcium imaging is an indirect mea-surement of neuronal firing activity,and this technique re-quires reconstructing the AP trains from the corresponding calcium fluorescence signals of the recorded neurons.During reconstruction,a threshold operation is necessary to trans-form the reconstructed signals into AP trains,and threshold selection largely determines the reconstruction accuracy;however,the optimal threshold is hard to find.Here,we consider the all-or-nothing nature of an AP train and in-corporate it into an optimization problem for AP train re-construction.After the above modification,the reconstructed signals are 0-1 sequences,which can be directly regarded as AP trains.Moreover,we model the baseline of the calcium signal and regard it as a linear combination of several low-frequency Fourier series.This eliminates the impact of the baseline fluctuations during reconstruction.Based on the above two improvements,we obtain a new AP reconstruc-tion model,the signal-shape composition(SSC)model.The SSC model is briefly described as follows.