This paper presents the development of filtering methods for positron emission tomography (PET) signals contaminated by radio frequency (RF) pulses for combined PET and clinical 3-T magnetic resonance imaging (MRI). The filtering methods include software, hardware, and hybrid correction methods. In the software correction method, PET signals are assessed, and valid signals are identified based on the characteristics of a typical PET signal using Field-Programmable Gate Array (FPGA)-based programming. The hardware correction method makes use of differential-to-single-ended and low-pass filter circuits for PET analog signals. The hybrid correction method involves the sequential application of both the hardware and software methods. Both valid and contaminated PET signals are measured with an oscilloscope. An evaluation is then made of the performance (energy resolution, photopeak channel, total counts, and coincidence timing resolution) of the PET detector modules with and without various MR sequences (gradient echo, spin echo T1 sequence). For all correction methods, the energy resolution, photopeak position, and coincidence timing resolution with MR sequences are similar (<; 3%) to those without MR sequences. However, the total count of each module depends greatly on the method applied. The hybrid correction method displays an ability to preserve (<; 1%) the total counts of the modules during various MR sequences. The results show that this filtering method, which can reject noise signals and reduce count loss while preserving the valid analog signals of MR sequences, is reliable and useful for the development of simultaneous PET-MRI.