Integrated Magnetic Resonance Imaging (iMRI): Integrating Multi-sequence and Multi-visit Information to Accelerate MRI Exams and Detect Temporal Image Changes

点击翻译按钮获取中文摘要

Magnetic resonance imaging (MRI) is a cornerstone non-invasive imaging technology that allows diagnosing several conditions and answering relevant research questions. Nearly two million MRI exams are done yearly in Canada at an average cost of $700/exam, making it a billion-dollar industry. During MRI acquisition, the scanner collects raw data, known as k-space, which is a representation of the image in the Fourier-domain. Reconstruction consists of transforming the k-space into interpretable images in the spatial domain. Traditionally, k-space is sampled following the Nyquist theorem. In this case, a simple inverse Fourier Transform operation is sufficient to reconstruct the spatial domain images. However, these fully sampled acquisitions lead to long MRI exam times (> 45 minutes/exam), which creates long wait times (> 9 weeks) compared to other imaging modalities like computed tomography. The MRI scanner collects multiple sequences during a scan session, resulting in images of the same location that have complementary information. Subjects are also often scanned numerous times (i.e., multi-visit) to monitor disease progression. Compressed Sensing (CS) and Parallel Imaging (PI) reconstruction methods reconstruct MRI from sub-Nyquist acquisitions, resulting in faster MRI exams. Conventional PI and CS methods disregard multi-sequence and multi-visit information. My research program focuses on developing methods to efficiently integrate information of different types and from different sources to advance knowledge and better support imaging applications' decision-making. In this Discovery Grant, I will investigate the integration of multi-sequence and multi-visit information to improve MRI reconstruction. The multi-visit information will also be used to detect temporal changes across images, which will expedite the analysis of MRI by the radiologists (i.e., professionals that interpret MRI). The specific objectives of this Discovery Grant proposal are: 1) To develop multi-sequence MRI reconstruction models from undersampled k-spaces; 2) To develop models that incorporate multi-visit MRI information to further improve the MRI reconstruction from objective one and detect temporal image changes; 3) To embed the methods developed into an MRI scanner and deploy them in real-time. My research program will train a large and diverse group of highly qualified personnel in Canada's strategic sectors. Multi-sequence MRI reconstruction will enable 10-fold faster scanning of new subjects and my pilot results indicate up to 20-fold faster scanning of multi-visit subjects (i.e., subjects with a previous scan). Faster MRI exams will decrease the price and reduce wait times of MRI exams. The detection of image changes across time will facilitate image interpretation. The outcomes of my research program have great potential for commercialization. The cost of embedding my models into an MRI scanner will be less than 2% of the scanner installation cost.