Two numerical models for predicting the temperature elevations resulting from focused ultrasound heating of muscle tissue were tested against experimental data. Both models use the Rayleigh-Sommerfeld integral to calculate the pressure field from a source distribution. The first method assumes a source distribution derived from a uniformly radiating transducer whereas the second uses a source distribution obtained by numerically projecting pressure field measurements from an area near the focus backward toward the transducer surface. Both of these calculated ultrasound fields were used as heat sources in the bioheat equation to calculate the temperature elevation in vivo. Experimental results were obtained from in vivo rabbit experiments using eight-element sector-vortex transducers at 1.61 and 1.7 MHz and noninvasive temperature mapping with MRI. Results showed that the uniformly radiating transducer model over-predicted the peak temperature by a factor ranging from 1.4 to 2.8, depending on the operating mode. Simulations run using the back-projected sources were much closer to experimental values, ranging from 1.0 to 1.7 times the experimental results, again varying with mode. Thus, a significant improvement in the treatment planning can be obtained by using actual measured ultrasound field distributions in combination with backward projection.