Healthcare providers face increasing demand on services, and radiology departments are no exception. There remains a continuous demand for medical imaging, and healthcare providers are turning their attention to technology that seek to lessen this burden. X-rays and CT scans are relatively short (typically, a matter of minutes), with the read times generally proportional (some use-cases for CT scans may have longer read times). For MR imaging, the scan times and read times are much longer (typically, tens of minutes).

Modality vendors have looked to overcome this challenge and reduce the scan time for MR imaging, using computer vision and classical machine learning methods. Several manufacturers, including GE Healthcare (HyperSense), Philips (Compressed SENSE), and Siemens Healthineers (Compressed Sensing), have adopted compressed sensing technologies, which aim to reduce scan times. These have led to incremental gains to reduce scan times, whilst ensuring image resolution is maintained (which is of paramount importance), especially as MR imaging remains the gold-standard for many imaging exams. However, compressed sensing has reached maturity due to computational complexities, and any further gains in reducing scan times are expected to be minimal.

More recently, deep learning-based image reconstruction technologies have emerged. For example, in June 2019, Canon Medical Systems received US-FDA 510(K) clearance for Advanced intelligent Clear-IQ Engine (AiCE), a deep learning image reconstruction technology. This technology attempts to bring about a significant step-change to reduce scan times, whilst maintaining the scan resolution and a high signal-to-noise ratio (SNR). In another development, Facebook AI Research (FAIR) are working with New York University Langone Health on a project known as fastMRI. This is a collaborative approach aiming to speed up the process in MR imaging. In December 2019, fastMRI released a fully anonymised dataset enabling broader research in this area.

These deep learning technologies will no doubt be sought after by healthcare providers, especially if scan times are significantly reduced with no visibly detectable differences resolution or image quality. Patients, in turn, will benefit the most; it will mean less time spent in the scanner, minimising the discomfort experienced. It will also mean fewer rescans, as artefact created, typically due to patients moving in the scanner, will be minimised.

Less time in the scanner will be especially beneficial for paediatric, claustrophobic or seriously ill patients, all of whom may have difficultly remaining still for long periods. It also reduces the need to sedate patients (to keep them still), reducing their risk exposure. Reducing MR scan times will also lead to improved efficiencies for clinicians, whilst healthcare providers can alleviate the pressures in the system through more effective use of the scanner. This translates to a monetary return on investment for such technologies for the healthcare provider.

However, reducing the scan time for MR imaging is not without its potential drawbacks. A shorter scan sequence will result in an increased number of scans delivered by the hospital. Consider the following hypothetical scenario: if the scan time for MR imaging is significantly reduced, it may allow a hospital to deliver an additional MR scan per day. If the hospital has three MR scanners this is approximately 1,000 additional scans per year. Assuming it takes a radiologist an average of 20–30 minutes to read a scan, this translates to between 330 to 500 extra hours of radiologist time required to read the additional scans. Of course, healthcare providers may not choose to increase the volume of MR scans delivered, but the scenario starts to highlight how a potential bottleneck in the imaging pathway may be created.

Inevitably, such bottlenecks become more pronounced in countries with a shortage of radiologists, where they are already faced with the problem of a backlog of unread scans.

A more prudent approach from healthcare providers would be to consider the entire imaging pathway and implement strategies to deal with potential bottlenecks caused by reducing MR scan times and the associated increase in scan volumes. Three suggested approaches, downstream of image acquisition, include: (1) investing in caseload management; (2) deploying automated solutions for image registration and/or image analysis (detection, segmentation, and quantification tools); and (3) outsourcing the increased workload to teleradiology services. Ideally, healthcare providers should consider implementing automated solutions across the imaging pathway, from image acquisition to image analysis.

Having said that, the market for automated image analysis software solutions, particularly the new breed of deep learning-based solutions, remains nascent and only a limited number have received regulatory approval and are commercially available.

In the long term, reducing MR scan times will have a significant impact on radiology, but healthcare providers should consider solutions addressing the whole imaging pathway to avoid potential bottlenecks downstream of image acquisition.

Sanjay Parekh is Senior Analyst at Signify Research, a UK-based independent supplier of market intelligence and consultancy to the global healthcare technology industry.