Increasing measurement information can effectively reduce the ill-posedness of the fluorescence molecular tomography(FMT) reconstruction. With the increase of data, the time of FMT reconstruction will increase significantly. In order to reduce the ill-posedness of FMT reconstruction and enhance reconstruction efficiency under big data sets, a reconstruction method of improved stochastic variant of alternating direction method of multipliers (ADMM) is proposed by combining dual coordinate ascent (DCA) method and ADMM method. The proposed method offers a stochastic update rule base on the original ADMM method where each iteration requires only one or few sample observations, thus gives speed up of convergence, so that the objective function can get the optimal solution rapidly and achieve the effect of rapid reconstruction. Simulation experiments of digital mouse and real mouse experiment show that the proposed method can guarantee FWT reconstruction images′s accuracy and improve reconstruction efficiency.