In radionuclide treatment, tumor cells are primarily destroyed by charged particles emitted by the compound while associated higher energy photons are used to image the tumor in order to determine radiation dose and monitor shrinkage. However, the higher energy photons are difficult to image with conventional collimated Anger cameras, since a tradeoff exists between resolution and sensitivity, and the collimator septal penetration and scattering is increased due to the high energy photons. This research compares imaging performance of the conventional Anger camera to a Compton imaging system that can have improved spatial resolution and sensitivity for high energy photons because this tradeoff is decoupled, and the effect of Doppler broadening at higher gamma energies is decreased. System performance is analyzed by the modified uniform Cramer-Rao bound (M-UCRB) algorithms based on the developed system modeling. The bound shows that the effect of Doppler broadening is the limiting factor for Compton camera performance for imaging 364.4 keV photons emitted from 131I. According to the bound, the Compton camera outperforms the collimated system for an equal number of detected events when the desired spatial resolution for a 26 cm diameter uniform disk object is better than 12 mm FWHM. For a 3D cylindrical phantom, the lower bound on variance for the collimated camera is greater than for the Compton imaginer over the resolution range from 0.5 to 2 cm FWHM. Furthermore, the detection sensitivity of the proposed Compton imaging system is about 15-20 times higher than that of the collimated Anger camera.