The excitement surrounding clinical AI has spread rapidly. Many skeptics have now become believers, and conversations are focusing on establishing best practices for AI governance structures and determining which stakeholders should be involved in creating a comprehensive AI strategy.
Radiologists are well-versed in engaging in deep discussions about AI. They have long recognized the potential for AI to significantly impact workflows, such as reducing reading time and reprioritizing worklists. The reading room has served as the starting point for integrating new healthcare technologies, with AI being no exception.
Given the radiology department’s importance as a key stakeholder in developing a sustainable AI strategy, here are three essential considerations regardless of where you are in your AI journey:
Championing the Enterprise-Wide AI Potential
AI algorithms have the power to transform specific use cases, and facilities have successfully implemented individual solutions to address real clinical challenges. As AI proves its value, the focus shifts to ensuring that these solutions are stable and scalable.
When scaling AI beyond a single use case, radiology leaders must address key questions such as potential complications from adding algorithms from different vendors, ensuring compatibility between algorithms, and avoiding conflicts with existing protocols.
For instance, consider an ED patient following a car accident receiving a chest and abdomen contrast-enhanced CT scan. How would your AI system handle orchestrating algorithms for each exam?
Which algorithms will be orchestrated to run on each exam?
How does your AI system decide which algorithms to run on each exam? The challenge lies in configuring and maintaining the orchestration of exams, ensuring optimal performance as changes occur at your site.
What is the radiologist’s experience?
How does the system provide information on the status of each algorithm running on an exam? How does it prioritize urgent findings and handle automatic insertion of AI results into reports? A unified interface should present AI status and results to radiologists in a seamless manner.
For the ED patient scenario, an enterprise-wide platform that orchestrates different algorithms based on scan type and anatomy can enhance patient outcomes by detecting unexpected pathologies alongside expected ones.
Looking ahead, as industry experts predict consolidation among AI vendors, health systems must carefully evaluate potential partners. Data normalization, single interfaces, and workflow integrations are crucial for ensuring a sustainable return on AI investment, both clinically and financially.
