Machine learning-assisted design of a novel crocin-loaded poloxamer 407/hyaluronic acid in situ forming hydrogel for endometriosis treatment: in vitro characterization and in vivo efficacy evaluation

Our study focused on fabricating and characterizing a crocin in situ forming hydrogel for endometriosis treatment. An artificial neural network (ANN)-based machine learning (ML) approach was employed to determine the effects of poloxamer 407 (P407) and hyaluronic acid (HA) concentrations on the gelation behavior. The optimum formulation was prepared using 17.9% w/v P407 and 0.5% w/v HA, with gelation temperature and time of 34 °C and 61 s, respectively. The spreadability values of the formulations were approximately 2.1–2.6 cm. Crocin was released in a controlled manner from the in situ gel, reaching ∼100% after 72 h (versus 1 h for the crocin solution), which fitted the Hixson-Crowell model. Furthermore, FTIR, XRD, TGA, and DTG showed characteristic peaks of optimized formulations and their components. The pore diameters of the crocin-loaded and bare gels were approximately 0.8–2.5 and 1.0–2.9 μm, respectively. A dose-dependent scavenging activity (88.2% for 200 µg/mL) was also observed for crocin-loaded in situ gels in the antioxidant assay. The formulations were cytocompatible up to concentrations of 100 µg/mL. Compared to the crocin solution, pathological lesions were reduced in the crocin gel group by prolonging crocin retention at the application site. H&E results demonstrated that the crocin-loaded in situ gel effectively reduced endometriosis scores from ∼3 to ∼0–1 compared with the control. The expression of TNF-α and IL-6 was also reduced in vivo , which was attributed to the anti-inflammatory properties of crocin in situ gel. The optimized system can be considered an effective endometriosis treatment.