Conventional (SRS) and fractionated (FSRS) stereotactic radiosurgery necessarily require stringent overall target point accuracy and precision. We determine three-dimensional intracranial target point deviations (TPDs) in a whole treatment procedure using magnetic resonance image (MRI)-based polymer-gel dosimetry, and suggest a technique for overall system tests. TPDs were measured using a custom-made head phantom and gel dosimetry. We calculated TPDs using a treatment planning system. Then, we compared TPDs using mid bi-plane and three-dimensional volume methods with spherical and elliptical targets to determine their inherent analysis errors; finally, we analyzed regional TPDs using the latter method. Average and maximum additive errors for ellipses were 0.62 and 0.69 mm, respectively. Total displacements were 0.92 ${pm}$ 0.25 and 0.77 ${pm}$ 0.15 mm for virtual SRS and FSRS, respectively. Average TPDtotal at peripheral regions was greater than that at central regions for both. Overall system accuracy was similar to that reported previously. Our technique could be used as an overall system accuracy test that considers the real radiation field shape.