Reconstruction of burst activity from calcium imaging of neuronal population via Lq minimization and interval screening

Calcium imaging is becoming an increasingly popular technology to indirectly measure activity patterns in local neuronal networks. Based on the dependence of calcium fluorescence on neuronal spiking, two-photon calcium imaging affords single-cell resolution of neuronal population activity. However, it is still difficult to reconstruct neuronal activity from complex calcium fluorescence traces, particularly for traces contaminated by noise. Here, we describe a robust and efficient neuronal-activity reconstruction method that utilizes Lq minimization and interval screening (IS), which we refer to as LqIS. The simulation results show that LqIS performs satisfactorily in terms of both accuracy and speed of reconstruction. Reconstruction of simulation and experimental data also shows that LqIS has advantages in terms of the recall rate, precision rate, and timing error. Finally, LqIS is demonstrated to effectively reconstruct neuronal burst activity from calcium fluorescence traces recorded from large-size neuronal population.