Classification of Craniomaxillofacial Free Flap: Mechanism and Model.

Timely detection and effective management of postoperative flap crises are critical for improving flap salvage rates. Flap crisis often stems from impaired blood circulation, leading to changes in the flap's color, texture, and temperature. Therefore, we analyzed flap crises using anatomical and colorimetric parameters and designed pixel curve features using a biologically derived foundation model. To mitigate the challenges posed by the complex craniomaxillofacial environment, we developed a dual-segmentation preprocessing approach combined with image morphology operations. During classification, a clustering-based constrained line extraction method was introduced to accurately identify effective feature regions. A voting-based decision mechanism was further employed to maximize the reliability of feature curve extraction and analysis. The experimental results demonstrate that the proposed classification model based on extracted pixel curve features, effectively distinguishes flap status and reduces the incidence of missed true-positive crisis cases. Continuous monitoring tests further validated the model's clinical utility. The code and dataset used in this study are publicly available at https://github.com/zhenhun1/freeflapclassfication.