The Gemini Eye in Microsurgery: Video-Based Capillary Refill and Chromatic Assessment for Free Flap Monitoring

Area of Science:

• Medical Technology
• Artificial Intelligence in Healthcare
• Plastic Surgery

Background:

• Postoperative free flap monitoring traditionally relies on subjective clinical assessments (color, turgor, CRT).
• Current methods are labor-intensive, subjective, and dependent on observer experience, leading to potential inconsistencies.
• Developing an objective, AI-driven secondary safety net is crucial for augmenting surgeon assessment.

Purpose of the Study:

• To evaluate the feasibility of a large multimodal AI model for postoperative free flap assessment without task-specific training.
• To determine if meaningful diagnostic agreement can be achieved between AI and expert surgeons.
• To explore AI's potential as an adjunct to clinical assessment in free flap surveillance.

Main Methods:

• Analysis of 143 video recordings of fasciocutaneous free flaps in patients with extraoral defects.
• AI model (Gemini 3 Flash) assessed flap color, turgor, CRT interpretation, and clinical diagnosis via intrapatient comparison.
• AI outputs were compared against consensus assessments from three senior plastic surgeons using Cohen's kappa and ROC analysis.

Main Results:

• Statistically significant agreement (p < 0.001) was found between AI and surgeons across all four parameters.
• Cohen's kappa values ranged from 0.487 (CRT interpretation) to 0.676 (turgor).
• AI correctly identified normal flaps in 81.0% of cases; ROC analysis showed an AUC of 0.667 for discriminating compromised flaps.

Conclusions:

• A multimodal AI model demonstrated significant agreement with expert surgical judgment for free flap assessment without prior training.
• This study serves as a proof-of-concept for AI-based video analysis in augmenting postoperative surveillance.
• Further prospective validation with clinical outcomes is required to establish clinical utility.