The accurate determination of mean ionization potential (I m) has the potential to reduce range uncertainty based margins and therefore allow for more focal treatments in proton radiotherapy. Many methods have been proposed to reduce uncertainty in I m and stopping power ratios (SPR), each with varying degrees of accuracy and issues. In this work, we present a simple parameterized model to determine I m in human biological tissue, allowing for the computation of patient-specific I m at the voxel level using magnetic resonance imaging (MRI). The model requires the measurement of three parameters by MRI, with only two parameters, mass percent water content and mass percent hydrogen content in organic molecules, required for the special case of soft tissues. The accuracy of this I m determination method was evaluated in available 'standard' (ICRU Report #44, (ICRU 1989 Tissue Substitutes in Radiation Dosimetry and Measurement (Bethesda, MD: International Commission on Radiation Units and Measurements))) human tissues. The sensitivity of this I m determination method to in vivo perturbations was also tested by calculating the effect of 10% variations in the experimentally measurable parameters on I m and SPR. For the human tissues modeled in this work, a high level of accuracy with low susceptibility to perturbations in measurement error was achieved in the prediction of I m. Root-mean-square errors in I m were within 0.77% and 1.8% for both soft and bony tissues, and were 0.09% and 0.2% for the SPR of soft and bony tissues, respectively, assuming knowledge of electron density. Proof of principle MR measurements and model-based computations of I m and SPR were taken in phantom for a series of hydrogenous solutions and compared against expected I m and SPR calculations from known elemental composition. MR determined I m and SPR values in a known composition solution were determined to within 5% and 0.52%, respectively. We present a novel model to accurately calculate mean ionization potential from measurements acquirable by MRI and show its feasibility in a phantom.