Can algorithms increase cost-effectiveness in radiology? : how to resolve magnetic resonance imaging bottlenecks with computing

Abstract

Artificial intelligence and deep learning solutions are increasingly utilized in healthcare and radiology. However, the literature addressing their productivity enhancement and financial impact remains limited.

Motivated by a need to enhance magnetic resonance imaging (MRI) scanner throughput in our hospital, this thesis investigates the diagnostic adoption of a new commercial deep learning reconstruction (DLR) algorithm. In this endeavor, the impact of the widespread deployment of this algorithm across our magnetic resonance imaging scanner fleet is evaluated.

The focus of this analysis is on the implications of the DLR algorithm for patient throughput and investment costs, while comparing it to alternative strategies for capacity expansion, such as the acquisition of an additional MRI scanner and the weekend utilization of the existing scanner fleet. A method to quantify and estimate technology’s monetary impact prior to the investment decision is provided for the reader, facilitating informed decision-making in healthcare investments.

Cost-reductions are demonstrated when compared to other methods of enhancing capacity. Procuring the DLR tecnology for each of our scanners represents only 11% of the cost associated with acquiring an additional scanner and 20% of the costs utilizing our existing devices during the weekend while producing the same scanner capacity.

The adoption of DLR for five MRI scanners was shown to yield MRI capacity similar to that of six non-DLR-enabled scanners. Consequently, procuring the technology would realize substantial cost savings. These reductions underscore the efficiency and economic feasibility of employing DLR to optimize MRI service delivery.