In optical projection tomography (OPT), for longitudinal living model studies, multiple measurements of the same sample are required at different time points. It is important to decrease both the total acquisition time and the light dose to the sample. We improved the ordered subsets expectation maximization reconstruction algorithm for OPT, which reduces the acquisition time and number of projections greatly compared with filtered back projection (FBP), and obtained satisfactory reconstructed images. Using zebrafish, in transmission and fluorescence mode, we demonstrate the capability of the method to reconstruct image from downsampled projection subsets. The result shows that the reconstruction image quality of the proposed method using 30 projections is comparable to that of FBP using 720 projections. The total acquisition procedure can be finished in a few seconds. The method also provides OPT with the remarkable capability to resist noises and artifacts. Projection image and fused image of zebrafish.