Compressed sensing is an advanced acceleration technique that under-samples k-space data by exploiting the intrinsic domain sparsity of imaged structures, and we are utilizing this technique to reduce the scan time for time-of-flight MRA while preserving diagnostic quality. This also eliminates the need for contrast administration.
Min Lang, MD, MSc, was awarded the 2022 Strategic Radiology Arl Van Moore Jr. MD Research Resident Grant by the RSNA Research & Education Foundation (R&E) at the end of 2022. The grant provided Dr. Lang with $30,000 for one year to devote 50% of his time to a research project under the guidance of a scientific advisor. Dr. Lang, a radiology resident at Massachusetts General Hospital, will work to develop an ultrafast 3-minute MRI/MRA protocol for acute ischemic stroke evaluation using new rapid MR imaging techniques; a novel AI-assisted reconstruction technique that enhances image quality will occur in tandem. His goal is to improve timely diagnosis and treatment of acute ischemic stroke, and perhaps be implemented for other neuroimaging indications.
Dr. Lang shared a bit about the project and his path in radiology in a Q & A with SR, which fully funded an $800,000 RSNA R&E named grant in 2020 that will fund a total of 20 annual research grants.
SR: What are the shortcomings of CT in stroke imaging and what made you believe that there was a way to more widely deploy MRI in this setting?
Dr. Lang: Ischemic stroke is one of the leading causes of morbidity and mortality worldwide. Neuroimaging plays a critical role in the initial evaluation of acute ischemic stroke. CT is often the first imaging modality of choice as it is widely accessible, fast acquisition of images, and does not require screening questionnaire for non-contrast enhanced exams. However, CT imaging has a low sensitivity for detecting acute ischemic stroke with an overall sensitivity of 57-71% in the first 24 hours and only 12% in the first 3 hours.
MRI offers a much greater sensitivity in detecting acute ischemic stroke than CT, with reported sensitivity of 73-92% in the first 3 hours and 95-100% in the first 6 hours. Furthermore, identification of diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) mismatch on MRI is suggestive of ischemic stroke at less than 4.5 hours, which is a crucial patient selection criterion for IV-tPA and is extremely helpful when symptom onset is unknown.
MRI utilization is increasing in the emergency setting. From 1999 to 2008, there has been a 38% increase of MRI utilization for ischemic stroke evaluation. Also, an increasing number of hospitals are moving toward 24/7 on-site MRI availability to meet requirements of The Joint Commission for Comprehensive Stroke Centerrequirements. Interest is growing in the adoption fast brain MRI techniques to meet the demand for 24/7 MR availability in acute clinical settings to quickly and accurately evaluate acute ischemic strokes.
SR: The first objective of your research is to develop an ultrafast 3-minute MRI/MRA protocol for acute ischemic evaluation using rapid MR imaging techniques developed at MGH. What is the basis of those techniques and how they have been used to date?
Dr. Lang: MRI is a powerful tool for ischemic stroke evaluation, but its utility has been limited by accessibility, long scan time, and sensitivity to motion. Conventional methods for accelerating MR sequences include 2D turbo-spin-echo sequences, single-shot echoplanar imaging, and parallel imaging. These techniques have been utilized in previous fast brain MRI protocols for evaluation of patients presenting with acute neurological symptoms and pediatric patients. But at higher acceleration factors (R>2), these techniques can suffer from increased noise, geometric distortion, and residual aliasing artifacts. Another limitation of single-shot echoplanar FLAIR imaging is the poor tissue contrast between white and grey matter.
Multi-shot echoplanar imaging (ms-EPI) is a highly efficiency interleaved EPI imaging technique that utilizes multiple excitations rather than a single excitation to acquire portions of the k-space simultaneously. This results in significantly reduced geometric distortion and higher SNR than single shot-EPI. Our institution implemented the ms-EPI technique on key brain MR sequences, including T2/T2*-weighted imaging, T1-weighted imaging, FLAIR imaging, and diffusion weighted imaging.
Compressed sensing (CS) is an advanced acceleration technique that under-samples k-space data by exploiting the intrinsic domain sparsity of imaged structures, and we are utilizing this technique to reduce the scan time for time-of-flight (TOF) MRA while preserving diagnostic quality. This also eliminates the need for contrast administration. This technique has been applied in cardiac MRI, dynamic contrast enhanced MRI, and spectroscopy. It also has been used in head MRA to evaluate vessel stenosis, small vessel pathology, vascular malformations, and post-bypass changes. Little has been reported on the use of CS TOF MRA for vessel occlusion evaluation in the clinical context of acute ischemic stroke.
So far, the ms-EPI and CS acceleration techniques have been successfully implemented in our ultrafast brain MR/MRA protocol.
SR: What are the protocol elements (including AI) that you intend to explore as you develop the 3-minute protocol for stroke imaging and will your team include researchers from other disciplines and sciences?
Dr. Lang: Our research group has developed an AI-assisted reconstruction framework that was trained on volunteer subjects scanned on a 3T MRI system. The AI-assisted reconstruction algorithm was applied to the ms-EPI imaging sequences to limit g-factor noise amplification and improve SNR of the highly accelerated images. The machine-learning based reconstruction incorporated a tunable parameter for controlling the level of denoising and has been previously validated across various acceleration factors, contrasts, and SNR conditions.
We developed this reconstruction framework through a collaborative research effort between radiologists from Massachusetts General Hospital and scientists from the Athinoula A. Martinos Center (Massachusetts Institute of Technology) and Siemens Healthineers.
You also intend to compare the quality of the 3-minute protocol to the standard MRA protocol in 65 patients who will receive neuroimaging using both protocols. Do you have a plan for selecting patients?
We intend to perform both the ultrafast brain MR/MRA protocol and the reference standard brain MR/MRA protocol in patients with clinical suspicion for acute ischemic stroke or transient ischemic attack (TIA). These patients must be in the emergency department or inpatient setting and have no contraindications to MR imaging. Based on our power analysis, we calculated that approximately 65 patients will be needed to complete the study.
SR: If you are successful, which patients will benefit most from your 3-minute stroke protocol?
Dr. Lang: We specifically chose patients from the emergency department and inpatient setting for this study because time to diagnosis and management is crucial in these patients. Currently, the guideline states that IV thrombolytics can be administered in patients up to 4.5 hours after symptom onset and mechanical thrombectomy can be considered in patients up to 24 hours after symptom onset. Of course, there are more nuances to patient selection for both IV thrombolytics and mechanical thrombectomy, which I won’t go into here.
Furthermore, patients with symptoms of acute stroke are often mentally altered and cannot lay still for a lengthy brain MR/MRA exam. With only 3 minutes of scan time, the ultrafast brain MRI will be able to complete the exam quickly, limit motion artifact, and provide key information to the stroke team. It is these patients that we believe will benefit the most from this ultrafast brain MR/MRA protocol.
SR: What led you into research and do you intend to continue this pursuit after training?
Dr. Lang: After completing my undergraduate training at the University of Toronto in Physiology and Human Biology, I was debating whether to pursue a career in research or medicine. I eventually enrolled in a research-focused Masters program at the University of Toronto, where I investigated mechanisms of neural circuit dysfunction in autism spectrum disorders and mechanisms and treatment strategies of epileptic seizures in rodent models. While I enjoyed basic science research, it lacked the immediate clinical impact that I sought.
The skills and passion for research carried over to medical school and residency training and led me to conduct translational and clinical research projects, including development of novel techniques to enhance MR image quality, COVID-19 associated imaging findings, and operations in health care. I plan on continuing these research pursuits after residency and fellowship training, and I also plan on venturing into other research areas, such as motion correction in MR imaging and AI-power synthetic data generation.
Hub is the monthly newsletter published for the membership of Strategic Radiology practices. It includes coalition and practice news as well as news and commentary of interest to radiology professionals.
If you want to know more about Strategic Radiology, you are invited to subscribe to our monthly newsletter. Your email will not be shared.