Investigation of in vivo silk scaffold degradation by decoupling tissue ingrowth using a Gaussian process regression-driven digital twin framework.

Pelvic organ prolapse (POP) reconstruction is increasingly performed by utilizing knitted silk meshes (KSM), yet tracking in vivo degradation kinetics remains challenging due to complex host tissue integration. This study developed an AI-driven semi-empirical framework utilizing Gaussian Process Regression (GPR) to bridge the kinetic mismatch between in vitro and in vivo environments. KSM scaffolds underwent 32 weeks of accelerated in vitro enzymatic degradation, with morphology (SEM), molecular conformation (FTIR), and mass loss being coupled with mechanical decay to train the GPR model. In vitro results revealed a multi-stage physical disintegration via a topochemical erosion pathway that preserved crystalline β-sheet structures despite macro-scale mass and mechanical loss. When validated in a rat abdominal wall defect model, traditional tracking metrics encountered severe bottlenecks. Heterogeneous dye labeling caused premature fluorescence quenching after 16 weeks, while extensive tissue ingrowth masked gravimetric and SEM signatures. Intriguingly, a bi-phasic in vivo mechanical trajectory was identified, where initial degradation-led failure was followed by a secondary mechanical recovery driven by biomechanical synergy with neo-muscular tissue. Importantly, despite premature quenching, this work presents the first optical imaging approach to visually mapping the complete chronological breakdown of the scaffold's peripheral boundary layer in vivo, proving that outer functionalized layers eroded prior to internal silk cores. Furthermore, our GPR framework offers a promising approach to address the perennial technical barrier of tissue-mesh overlapping. By mathematically decoupling intrinsic polymer degradation from confounding tissue ingrowth, this model successfully achieved a prediction of the bare scaffold's long-term structural fate in a non-adhered state, providing a preliminary digital twin methodology for lifetime predictions of degradable biomaterials.