Related Experiment Video
Updated: May 8, 2025
A Pre-clinical Rat Model for the Study of Ischemia-reperfusion Injury in Reconstructive Microsurgery
Published on: November 8, 2019
An Artificial Intelligence Model for Vascularity Monitoring of Postoperative Flap
Ming Lu1,2, Aijie Zhang3, Yao Wang4
1School of Biomedical Engineering, Tsinghua University.
Background:
Postoperative flap monitoring is a labor-intensive task that relies heavily on the subjective judgment of skin color change by both medical and nursing staff. Human errors, such as misinterpretation of visual cues, may result in flap failure due to missed or delayed detection of vascular compromise.
Objective:
To simulate clinical situations of vascular compromise and investigate the feasibility of using an artificial intelligence (AI) model with skin color monitoring algorithms (SCMAs) and pulse wave monitoring algorithms (PWMAs) in identifying changes to site vascularity.
Materials And Methods:
A tourniquet machine was used on human participants to simulate clinical conditions of venous and arterial obstruction by applying 6 different pressure conditions (0, 40, 70, 100, 160, and 200 mm Hg). The tourniquet was placed 10 cm above each participant's elbow crease, and each pressure condition was maintained for 3 minutes. The color changes of the participant's forearm and arm were assessed by 5 clinical specialists of level 2 according to the Tang and Giddins classification, and the SCMAs and PWMAs were recorded as well with their respective algorithms. The results from the SCMAs were then compared with the manual assessments, while the results from the PWMAs were compared with the pulse oximeter readings.
Results:
The 100 mm Hg pressure condition simulated a clinical venous blockage, while the 200 mm Hg pressure condition simulated an arterial blockage. The results from the SCMAs and PWMAs correlated with manual assessments of skin color changes and pulse oximetry readings, respectively. The simulation of a venous compromise showed a gradual deepening of skin color, whereas an arterial obstruction demonstrated decreased peripheral oxygen saturation instead of significant changes to skin color.
Conclusion:
These AI algorithms are comparable with manual assessment and contact-based devices in identifying vascular compromise, which portends feasibility in future research and development of such technology.
