In this paper, we propose two ordered-subset simultaneous algebraic reconstruction techniques (OS-SART). First, we describe the heuristics in support of the two OS-SART formulas. Then, we study the convergence in the framework of our recent work on the OS version of the Landweber scheme. It is shown that our first OS-SART is a special case of the OS version of the Landweber scheme. Hence, it converges in the weighted least square sense even in the case of inconsistent data. Both the OS-SART formulas are tested for reconstruction of CT images from practical data.