1663
clinical study

Evaluation of a Large-Scale Multi-Year AI Rollout in an Academic Radiology Department.

Materials & Methods

Key analytics, trends, and milestones that are critical for a successful roll out of AI at scale in a large academic radiologist department were extracted from our department. The study provides a set of guidelines applicable to help other institutions for metrics and best practice for monitoring of AI. 

Results

During this time period, the use of AI grew from 2 pathologies/modules to a total of 9 pathologies/modules, deployed throughout the hospital system. Monthly analyzed cases grew from 1,687 cases/month to 18,307 cases/month over a period of 30 months. The total number of active users, defined as a user who received a prioritization alert, grew from 50+ users to 150+ from 01/2021 to 06/2021. 

Over 18% of monthly studies were analyzed by two or more AI modules during advanced deployment. The trend showed the added value of scaling up AI on multiple pathologies of common anatomical regions.  A key shift occurred when the LVO module was activated around 01/2021. An example of a multi-module use case is a head CT protocol containing both CTA and non-contrast CT series analyzed for LVO, BA and ICH,

Conclusions

Our case study may provide insights and value to other radiology practices and hospital systems as a basis for defining best practice guidelines.