The study developed an ensemble learning model combining conventional radiomic features (CR) and machine learning (ML) features to assess periorbital fat status across the age spectrum. They analyzed preoperative MRI of the skull and face in 237 patients with meningioma, divided into a training set (165 patients) and a test set (72 patients). Patients were categorized into youth (age 28.5 ± 5.0 in training, 28.6 ± 5.6 in test), middle-aged (42.9 ± 4.7; 43.9 ± 4.1) and seniors (60.0 ± 6.5; 58.8 ± 6.7) groups. Features from three periorbital regions were extracted from CR and ML. The ensemble model outperformed the separate CR, ML, and fusion CR-ML models, with a macro AUC of 0.833 (95% CI: 0.737–0.902), F1 score of 0.614, precision of 0.597, and positive predictive value of 0.6 on the test set. The model demonstrated optimal capabilities in multi-classification tasks, improved generalization and robustness. It has achieved non-invasive and reliable evaluation for rejuvenation surgery.