90 patients diagnosed with primary osteoarthritis were randomly assigned to undergo total knee arthroplasty using either conventional instrumentation, MRI-based patient-specific instrumentation, or CT-based patient-specific instrumentation. Patients were assessed after surgery for coronal alignment (mechanical axis, femoral component, and tibial component), sagittal alignment (posterior slope), and axial alignment (rotation of the femoral component|). Clinical scores were obtained at 3 months postoperatively. Surgical duration was also documented and compared. Postoperatively, significantly fewer cases in which TKA was completed with MRI-based PSI demonstrated outliers (>3deg deviation from planned alignment) for mechanical axis, coronal alignment of the femoral component and axial alignment of the femoral component when compared to conventional instrument. CT-based PSI differed significantly compared to standard instruments only in the number of outliers in axial alignment of the femoral component. Comparisons of coronal alignment of the tibial component and posterior tibial slope did not differ significantly between any two groups.