Integrating AI into Radiology workflow: Levels of research, production,\n and feedback maturity

Abstract

This report represents a roadmap for integrating Artificial Intelligence\n(AI)-based image analysis algorithms into existing Radiology workflows such\nthat: (1) radiologists can significantly benefit from enhanced automation in\nvarious imaging tasks due to AI; and (2) radiologists' feedback is utilized to\nfurther improve the AI application. This is achieved by establishing three\nmaturity levels where: (1) research enables the visualization of AI-based\nresults/annotations by radiologists without generating new patient records; (2)\nproduction allows the AI-based system to generate results stored in an\ninstitution's Picture Archiving and Communication System; and (3) feedback\nequips radiologists with tools for editing the AI inference results for\nperiodic retraining of the deployed AI systems, thereby allowing the continuous\norganic improvement of AI-based radiology-workflow solutions. A case study\n(i.e., detection of brain metastases with T1-weighted contrast-enhanced 3D MRI)\nillustrates the deployment details of a particular AI-based application\naccording to the aforementioned maturity levels. It is shown that the given AI\napplication significantly improves with the feedback coming from radiologists;\nthe number of incorrectly detected brain metastases (false positives) reduces\nfrom 14.2 to 9.12 per patient with the number of subsequently annotated\ndatasets increasing from 93 to 217 as a result of radiologist adjudication.\n

Ähnliche Arbeiten