Autonomous Low-field Magnetic Resonance Imaging - Volume II

This Research Topic is the second volume of the Autonomous Low-field Magnetic Resonance Imaging series, the first volume can be found here.

Magnetic Resonance Imaging (MRI) has played a critical role in medicine and science. However, at present, two-thirds of the world does not have access to MRI. This inaccessibility is exacerbated in developing countries and low-resource settings. The unavailability of MRI needs immediate attention, given that approximately eighty percent of the world’s population and ninety-seven percent of the population growth will be in developing countries. This low accessibility results from the cost, siting, and electrical power requirements of high-field MRI systems. Further, the lack of human expertise in these underserved settings negatively impacts accessible neuroimaging, especially given the need for diverse and skilled human resources to install, operate, use, and maintain MRI systems. In contrast, portable low-field imaging provides a point-of-care device that significantly improves accessibility and affordability with improved safety. These advantages have led to a resurgence in low-field MRI (< 1T) development.

However, this resurgence is constrained by a lack of standardization in the associated hardware and software that impacts consistent scanner operation and image quality. Therefore, there is a need to validate protocols and ensure comparable image quality. Identifying and mitigating artifacts in MR images requires expertise and experience. The challenges to ensure scanner consistency and image quality are exacerbated at very low field due to the following four factors: (i) significantly longer acquisition times – reduced SNR and spatial resolution; (ii) inconsistency in scanner operation due to effects of thermal drift, electromagnetic interference (EMI), etc ; (iii) lack of MR expertise in remote and unconventional locations; (iv) lack of repeatability and reproducibility of MRI data acquired at very low-field. Therefore, there is a critical need to develop automated methods that can intervene more frequently to update acquisition parameters to control variability in scanner operation. These developments are necessary and timely as groups across the globe invest time and resources to develop low-field MRI systems and methods.

In this second volume of the Research Topic on Autonomous MRI (AMRI), we explore AMRI methods and applications at low-field that use artificial intelligence (AI) approaches to augment human MR expertise required for accessible neuroimaging investigations. This research overview will comprise research reports, mini-reviews, and data reports related to using AI techniques to automate operation and enable interpretation of neuropathologies such as Hydrocephalus, stroke, brain tumors, and dementia. In particular, the Research Topic will focus on low-field image quality and variability assessment, artifact identification and mitigation, sensors and robotics for characterizing low-field scanners, image reconstruction, and associated analyses. These technical reports will be complemented by articles reporting on the usage of low-field MRI for in vivo human brain investigations in disease and health, neuroethics of portable MRI, and good ML practices at low-field.

In summary, the Research Topic will strive to provide a state-of-the-art collection of scientific works focusing on the methods and in vivo uses of AI in low-field MRI of the human brain.

The subtopics will be categorized based on the challenges that must be overcome to make low-field MRI more consistent for use and image quality. The list of potential subtopics includes:

1. Low-field Autonomous Brain MRI: Why, what, and how?

2. Low-field Autonomous MRI: Opportunities and challenges

3. Automating low-field image acquisition and quality assessment

4. Robotics and sensors for low-field MRI

5. Low-field MRI simulations and applications

7. Super-resolution reconstruction for low-field MRI

8. Neuroethics of portable and autonomous MRI

9. Good machine learning practices for low-field MRI

10. Automating Hydrocephalus detection

11. Automating Dementia screening

12. Automating Stroke detection

13. Automating Brain tumor detection and follow-up

14. Semi-automated monitoring of mental health and substance use disorders

15. Semi-automated monitoring of infectious diseases

Dr. Geethanath is the co-founder of MR Access, the inventor of multiple US patent applications, and the recipient of funding from GE Healthcare, USA, and GE Healthcare, India. The other Topic Editors declare no competing interests with the topic.