Nuclear Medicine and Artificial Intelligence: Best Practices for Evaluation (the RELAINCE Guidelines)

Area of Science:

• Nuclear Medicine
• Artificial Intelligence
• Medical Imaging

Background:

• A critical need exists for standardized, objective methods to evaluate artificial intelligence (AI) algorithms in clinical settings.
• Current evaluation strategies for AI in nuclear medicine lack rigor and task-based clinical relevance.

Purpose of the Study:

• To propose a comprehensive 4-class framework for evaluating AI algorithms in nuclear medicine.
• To establish best practices for assessing AI algorithm promise, technical efficacy, clinical decision-making impact, and post-deployment performance.

Main Methods:

• Development of a 4-class evaluation framework: promise, technical task-specific efficacy, clinical decision making, and post-deployment efficacy.
• Tabulation of best practices into RELAINCE (Recommendations for EvaLuation of AI for NuClear medicinE) guidelines.
• Multidisciplinary collaboration involving nuclear medicine physicians, physicists, computational imaging scientists, and regulatory experts.

Main Results:

• The proposed framework provides a structured approach to AI algorithm evaluation in nuclear medicine.
• Each evaluation class generates a specific claim detailing the AI algorithm's performance characteristics.
• The RELAINCE guidelines offer actionable best practices for rigorous AI assessment.

Conclusions:

• The 4-class framework and RELAINCE guidelines address the need for objective, clinical-task-based evaluation of AI in nuclear medicine.
• This systematic approach will enhance the reliable integration of AI into clinical practice.
• Standardized evaluation promotes trust and efficacy in AI-driven nuclear medicine.