Data harmonization is an important issue in computerized analysis of medical images, and especially with magnetic resonance imaging (MRI). MRI consists of a complex system of radiofrequency pulses at various times that allow for incredible control over the imaging contrast. However, with that comes a lack of uniformity in imaging datasets. While radiologists do not necessarily need a uniform imaging dataset, this presents a problem when trying to train machine learning algorithms to perform tasks using these heterogeneous datasets. Without any uniformity, certain imaging features may be skewed or not present, even if the training dataset consists of the same MR weightings. Relaxometry or quantitative MRI is a potential solution to this issue. The contrast seen in T1-weighted and T2-weighted MRIs between tissues is primarily due to this difference in T1 and/or T2 relaxation between tissues of different types. The other parameters used to generate the MR image (such as the repetition time, echo time, flip angle, etc) are used to highlight certain aspects of the image or affect the image quality or duration. Therefore, if it were possible to remove the user-defined scan parameters and generate an image of only T1 or T2, the issue of a heterogeneous dataset could be solved. Current relaxometry techniques are slow, computationally expensive, and not often performed. Most clinicians find no additional value in performing these scans, so very few datasets include relaxometry images. Because of this, there are few MR imaging datasets that can be used for machine learning applications. It would be beneficial, then, to be able to retrospectively remove the effect of scan parameters on previously-acquired MR images and artificially produce quantitative T1 and T2 maps for big data studies. By doing this, it would not disrupt clinical workflow and instead be a step in a machine learning pipeline. The research presented in this dissertation presents the following results. First, an algorithm to retrospectively quantitate T1 from T1-weighted MR images was developed using a combination of sequence-specific MR signal equations and literature values for healthy tissue as references. This algorithm, referred to as T1-REtrospective Quantification Using Internal REferences (T1-REQUIRE), was tested for two specific MR sequences: T1-weighted spin-echo and T1-weighted magnetization prepared gradient echo (MPRAGE). The results demonstrate that T1-REQUIRE is comparable to the reference standard to a clinically-relevant degree and within a range of T1 values that are inclusive of most brain tissue and neuropathology. In addition, it was shown that T1-REQUIRE also effectively harmonizes data across both sequences and across scanners. This is shown by comparing the results from T1-REQUIRE for both the spin-echo and MPRAGE images, before completing a study consisting of two subjects having repeated scans across multiple scanner models and manufacturers. Finally, T1-REQUIRE was applied for a pilot study to determine if it was effective at data harmonization and if it provided any additional predictive power.