Integrating Al Algorithms into the Clinical Workflow

Area of Science:

• Medical Informatics
• Radiology
• Artificial Intelligence in Medicine

Background:

• Integrating artificial intelligence (AI) algorithms into clinical workflows is crucial for practical application.
• A pilot study was conducted to deploy AI systems in clinical practice using lymphoscintigraphy examinations.

Purpose of the Study:

• To describe generalizable components for deploying AI systems into clinical practice.
• To evaluate the effectiveness and user satisfaction of an AI integration framework in a clinical setting.

Main Methods:

• Developed a seven-component AI deployment system: image delivery, quality control, results database, processing, presentation, error correction, and performance monitoring dashboard.
• Implemented the system in a clinical pilot study involving 14 users (radiologists and trainees) over 16 months.
• Analyzed the number of examinations processed, error rates, and correction times.

Main Results:

• The AI system processed 1748 lymphoscintigraphy examinations.
• Radiologists corrected 146 AI results, enabling real-time updates to radiology reports.
• All 14 users reported satisfaction with the AI system's integration into the clinical workflow.

Conclusions:

• A framework for integrating AI algorithms into clinical workflows was successfully developed and implemented.
• The described AI integration facilitates the assessment and real-time monitoring of AI system performance in clinical practice.
• This approach supports the seamless incorporation of AI tools into diagnostic radiology.