Minimizing MRI Geometric Distortions for Improved Stereotactic Surgical Planning Accuracy: a Theoretical and Experimental Analysis

Accurate localization of internal structures is essential for successful stereotactic surgical planning. Magnetic resonance imaging (MRI) is an attractive modality for stereotactic imaging because it is highly sensitive to soft-tissue differences. Unfortunately, it has been shown to be susceptible to geometric distortions. These distortions contribute to object shifting and also to object -shape deformations. Research analyzing these distortions has found them to be complex and attributable to a variety of different sources. Furthermore, a wide range of stereotactic errors has been reported in numerous clinical studies, intimating that these errors are site specific. Because of the complexity of these distortions and the uncertainty of their effects from one imaging site to the next, most clinical sites choose either to ignore the likelihood that distortions exist, or image concurrently with MRI (for tissue specificity) and x-ray computed tomography (for geometric accuracy). Both of these strategies are unsatisfactory, however, as they either compromise patient care or induce unnecessary cost and inconvenience. This uncertainty in the accuracy of MRI was the impetus that prompted a comprehensive study of geometric distortion contributors and their subsequent effects on both imaged objects and stereotactic accuracy. This dissertation is a report of that study. As far as is known, it is the first comprehensive analysis (theoretical and experimental) of all of the individual distortion contributors that affect MRI geometric accuracy. Additionally, it is the first work that individually analyzes distortion effects on the stereotactic referencing system as well as on imaged objects. The dissertation begins with a brief history of the role of medical imaging in stereotactic surgical planning. Individual contributors to MRI geometric distortions are then analyzed theoretically and experimentally. How they affect both object distortions and stereotactic accuracy is addressed. From this analysis it is shown that significant geometric distortions often exist in MR images. It is pointed out, however, that the net geometric distortion in a stereotactic image can be minimized to an acceptable level (less than 1 mm error) with careful selection of MR sequences and parameters. This improved accuracy is accompanied by a reduction in the signal-to-noise ratio. This can be overcome by signal averaging, but this is clinically unattractive because it requires extra time. A novel method that yields a high degree of accuracy without sacrificing signal-to-noise is then introduced. Finally, the concepts developed throughout this dissertation are validated using stereotactic images of a cadaver head. It is concluded that by implementing the steps outlined in this dissertation, any MR site can achieve a high degree of image stereotactic accuracy without the cost, inconvenience, or patient discomfort associated with concurrent x-ray computed tomography (CT) imaging.