In this project, we propose a novel T1 and T2 quantification method that generates quantitative T1 or T2 maps from weighted MR images. Magnetic resonance imaging (MRI) is commonly used as a tool to diagnose Multiple Sclerosis (MS) and track lesional changes over time. Because MRI has various contrasts that display different information about the underlying tissue microstructure and physiology, it can potentially be used as a tool to predict MS disease progression and even disability. However, there is no known measure derived from MR images of MS that correlates well with clinical disability as described by the Expanded Disability Status Score (EDSS). Previous efforts to correlate MRI features and EDSS have included calculating total lesion load on T1- and T2-weighted images, measuring the variations in the magnetic transfer ration of normal-appearing brain tissues, and calculating cerebral atrophy, each with a varying level of success. Yet, there has been little study of the evolution of relaxation times of the lesions over time and how it relates to disability. Because changes in the T1 (spin-lattice) and T2 (spin-spin) relaxation times of a tissue can reflect pathological changes in that tissue over time, quantitative T1 and T2 maps derived from MR images may be more indicative of microscopic changes that manifest as disability in MS patients.
The specific aims of this proposal are: (1) develop and validate novel T1 and T2 quantification method on spin-echo MR images, (2) extend the novel quantification method to common MS imaging sequences, and (3) apply the novel quantification method to MS datasets to predict EDSS using machine learning.
Aim 1 will involve the validation of the quantification pipeline on both T1- and T2-weighted spin-echo MR images in vivo, resulting in a range of acceptable parameters for the novel quantification method.
Aim 2 will extend the quantification pipeline to include commonly used and more complicated MS imaging sequences, again resulting in a range of acceptable parameters for the quantification method.
Aim 3 will use the quantification pipeline to compare machine learning algorithms with and without quantification to determine the added value of quantification in the imaging of MS. Additionally, Aim 3 will result in a predictive machine learning model utilizing multiple imaging contrasts for the prediction of disability in MS. These results will provide a more thorough understanding of the role of MR quantification in the evaluation of neurological diseases, such as MS, and will offer a scientific foundation to extend the use of MR quantification as a potential imaging biomarker for other diseases and pathologies.
The proposed research aims to develop and validate a T1 and T2 quantification method using internal reference values derived from T1- or T2-weighted MR images. This method will be applied to weighted images of patients who have been diagnosed Multiple Sclerosis and input into various classification algorithms to determine which method is most predictive of worsening clinical disability as described by the Expanded Disability Status Score. By doing this, we will determine the impact of this novel quantification method as a tool for both the analysis of patients with Multiple Sclerosis as well as a tool for the normalization of big MR datasets before being input into machine learning algorithms.
