Abstract
Background
Using artificial intelligence (AI) to augment knowledge is key to establishing precision education in modern radiology training. Our department has developed a novel AI-derived knowledge recommender, “The system” (RH), that augments the training of radiology residents and fellows, leading to the first precision education program in radiology.
Purpose
To assess the impact on trainees of an AI-based knowledge recommender compared to traditional knowledge sourcing for radiology reporting through reporting time, quality, cognitive load, and learning experiences.
Materials and Methods
A mixed methods prospective study allocated trainees to intervention and control groups, working with and without access to RH, respectively. Validated questionnaires and observed and graded simulated PACS-based reporting at the start and end of a month’s rotation assessed technology acceptance, image interpretation quality, turnaround time, cognitive load, and attitudes toward modified learning strategies.
Results
The RH group showed a statistically significant reduction in mean case reading time by 161.5 seconds for every case (~ 2 ½ minutes; p = 0.022) and mean case-sourcing time by 113 seconds for difficult cases (~ 2 minutes; p = 0.026). The intervention group showed a 14% increase in image interpretation accuracy (p < 0.001) as well as reduced overall workload (p = 0.029), mental demand (p = 0.030), and effort (p = 0.039). Additionally, 45–65% positively rated productivity and effectiveness with over 80% finding it flexible and easy to use, with strong optimism towards technology (3.9–4.3/5) and moderate to high intrinsic motivation (5.1-6.0/7). Eighty-four percent of participants requested access to RH for their next rotation.
Conclusion
This study supports the growing philosophy that AI will boost rather than replace human intelligence in medical training with enhanced quality and productivity. Our knowledge recommender can effectively augment the knowledge and performance of radiology trainees, and it is highly likely that the learner will use RH to promote self-directed learning. Further testing of a larger external cohort will support more widespread implementation of RH for precision education.