Economical and Diagnostic Quality Aspects Relevant for a Radiology Department Considering Implementation of Deep Learning – Augmented MRI

Motivation: Patient throughput, while maintaining image quality, remains a major concern of radiology departments, especially with ever-increasing referral numbers/waitlists. Goal(s): To comparatively assess the diagnostic quality of neuroimaging studies with/without commercial deep learning MRI and its economic impact. Approach: 20 most-frequent MRI neuroimaging-protocols underwent comparative assessments of patient scan quality and speed with/without 2D/3D Deep Resolve and assessed its economic impact due to freed-up time. Results: Average scan time was reduced by almost 34%, enabling us to shorten scan slots from 30 to 20min while improving quality. At current software costs and reimbursement levels, initial expenses can be earned back within <5 months. Impact: We identified the most time-consuming scan protocols by weighting the frequency of utilization by protocol duration. Leveraging 2D+3D DL reconstruction, these protocols were dramatically shortened, freeing up approximately 34% of MRI time, allowing increased throughput while improving diagnostic quality.