Artificial intelligence in medicine: a clinician-oriented guide and operational framework to evaluate new AI devices

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support

Background:

• Artificial Intelligence (AI) is increasingly used in medicine, but understanding is often limited to interactive 'front-end' tools.
• The prevalence of 'back-end' AI, embedded invisibly in systems, is often overlooked by clinicians.
• A clear distinction and evaluation framework for front-end and back-end AI are needed for clinical practice.

Purpose of the Study:

• To present a clinician-oriented framework distinguishing between front-end and back-end AI.
• To propose a clinically guided approach for evaluating these AI systems.
• To enhance clinicians' understanding and safe application of AI in medicine.

Main Methods:

• Conducted a narrative review of AI applications in minimally invasive therapy and medical imaging.
• Categorized AI systems into 'front-end' (interactive) and 'back-end' (embedded) modalities.
• Synthesized relevant evaluation metrics from computer science and engineering for clinical safety.

Main Results:

• Front-end and back-end AI systems require distinct validation strategies for safety.
• Front-end evaluation focuses on decision quality and human-computer interaction; back-end evaluation requires technical benchmarking of signal fidelity and latency.
• A structured inquiry framework was developed to audit AI systems for data provenance, transparency, and failure modes.

Conclusions:

• Clinical safety in the AI era necessitates 'algorithmic literacy' among medical professionals.
• Applying the front-end/back-end framework aids in identifying AI failure modes and ensuring data integrity.
• Clinicians must transition from passive consumers to active evaluators of medical AI technology for accountability.