Evolution of methodology for obtaining hydrogen magnetic resonance spectra of thyroid nodules in vivo
Thyroid nodules are common (up to 10% of the population), but only a small fraction (<5%) are found to be malignant. Cytology of fine needle aspiration biopsies (FNAB), the standard initial diagnostic modality for thyroid cancer, is unable to discriminate benign from malignant follicular thyroid nodules, which are recognized only by pathologic evidence of capsular or vascular invasion. Therefore, many thyroid surgeries are performed simply to exclude a diagnosis of malignancy. In the benign case, the thyroid gland is removed solely for diagnostic purposes. 'Ex vivo' proton MRS on resected tissue as well as on FNAB has been reported to accurately discriminate malignant thyroid nodules from normal tissue. The ability to localize and differentiate normal or benign tissue from their malignant counterpart with MRS ' in vivo' would provide a non-invasive and non-subjective diagnosis, reduce surgery for nonmalignant glands, and aid in the clinical management of thyroid nodules. This thesis describes methods for obtaining 1H MR spectra ' in vivo' from normal thyroid and thyroid nodules. Many technical barriers, such as low signal-to-noise, poor resolution, poor localization, and magnetic susceptibility and motion-related artifacts, had to be overcome. The research began by demonstrating that the poor SNR, resolution and/or water suppression of 1H MR spectra of the thyroid at magnetic fields of 1.5T would not allow for classification of thyroid spectra ' in vivo'. Considerable improvement in SNR and resolution was apparent at a magnetic field of 3T compared to results at 1.5T. Optimizing the slice-selecti n order, as well as automatic localized shimming and frequency/phase correction, were necessary for achieving optimal spectral resolution. Using large spoiler gradients, with a large difference between TE and TM gradient areas, minimized the amount of contamination present in the spectra from unwanted coherences. The combined use of STEAM localization and a multi-ring surface coil offered the SNR advantages of a surface coil with homogeneous volume localization, comparable to that of volume coils. With an appropriate multi-ring surface-coil design and optimal placement with respect to the thyroid or tumor, sensitivity to lipid contamination from subcutaneous neck fat can be reduced. Together, the multi-ring surface coil characteristics allowed high-quality spectra to be obtained. 'In vivo' 1H MR spectra from normal thyroid at 3T had spectral characteristics similar to those of 'ex vivo' biopsy spectra at 8.5T. When the same criteria used for classification of the 'ex vivo' spectra were applied to the 'in vivo' spectra obtained from normal volunteers, all were classified as "normal". Similarly for the benign cases, the 1.7ppm/0.9ppm and the 2.0ppm/0.9ppm metabolite peak height ratios were greater that 1.1 and 1.12, respectively, which would also classify these spectra as "normal". Therefore, the 1H MR spectra obtained from normal thyroid as well as thyroid nodules show that there is potential for classification of thyroid tumors ' in vivo' using the previously described 3T methods.