Predictability Of The Vault After Implantable Collamer Lens Implantation, Using Anterior Segment Oct And Artificial Intelligence

To compare the predicted vault using machine learning with the achieved vault using the online manufacturer’s nomogram in patients undergoing posterior chamber implantation with an implantable collamer lens (ICL; STAAR Surgical).

This multicenter study was conducted at the Centro Oculistico Bresciano (Brescia, Italy) and at the I.R.C.C.S. – G.B. Bietti Foundation (Rome, Italy), where we performed a retrospective analysis of patients who had undergone ICL implantation between 2018 and 2021.

This retrospective study included 449 eyes from 238 consecutive patients who underwent ICL placement surgery. All preoperative and postoperative measurements were obtained by anterior segment optical coherence tomography (AS-OCT; MS-39 Costruzioni Strumenti Oftalmici C.S.O. SRL, Italy). The actual vault was quantitatively measured and compared with the predicted vault using machine leaning of AS-OCT metrics.

A strong correlation between the model predictions and achieved vaulting was detected by random forest regression (RF; R² = 0.42), extra tree regression (ET; R² = 0.50), and extreme gradient boosting regression (R² = 0.41). Conversely, a high residual difference was observed between the achieved vaulting values and those predicted by the multilinear regression (R² = 0.32) and ridge regression (R² = 0.32). The ET and RF regression showed significantly lower mean absolute errors and higher percentages of eyes within ±250 µm of the intended ICL vault than the conventional nomogram (P < .001). The ET regression provided the lowest mean absolute error and the best vault prediction.

Machine learning of the preoperative AS-OCT metrics, especially the ET regression, provided significantly higher predictability of the ICL vault than the online manufacturer’s nomogram, providing the surgeon with a valuable aid for predicting the ICL vault in clinical practice.