Validation of an Artificial Intelligence-Based Anterior Chamber Depth Estimation Using a Smartphone-Compatible Slit Lamp Device.

PURPOSE

Accurate evaluation of anterior chamber depth (ACD), a major risk factor for angle closure, is clinically important. Although standard techniques provide reliable measurements, they are often labor-intensive, technically demanding, and time-consuming. To address this, we previously developed an artificial intelligence (AI) algorithm capable of estimating ACD from slit lamp photographs. This study sought to assess the performance of this AI model when applied via the Smart Eye Camera (SEC), a smartphone-compatible slit lamp imaging system, by comparing its estimates to those obtained with anterior segment OCT (AS-OCT) at a separate institution.

DESIGN

An evaluation of diagnostic test.

PARTICIPANTS

Five hundred fifty-six phakic eyes (268 nondilated and 288 dilated eyes) from 329 Asian patients.

METHODS

A retrospective analysis was performed on images captured using both the SEC and AS-OCT. Anterior chamber depth values generated by the AI model embedded in the SEC were compared with corresponding measurements obtained using AS-OCT.

MAIN OUTCOME MEASURES

Metrics, including mean absolute error (MAE), mean squared error (MSE), Pearson correlation coefficient, and the intraclass correlation coefficient (ICC), were calculated to evaluate model performance.

RESULTS

The AI algorithm integrated into the SEC demonstrated the capability to estimate the ACD with an MAE of 0.119 ± 0.0949 mm and an MSE of 0.0233 ± 0.0557 mm. A strong correlation was observed between ACD measurements obtained using AS-OCT and those estimated by AI ( = 0.922; 95% confidence interval, 0.908-0.933). The ICC for agreement between AI-estimated and AS-OCT-measured ACD was 0.903, indicating excellent reliability.

CONCLUSIONS

Our AI model, embedded within the SEC platform, demonstrated high accuracy in estimating ACD when benchmarked against AS-OCT. Given its portability and user-friendliness, the SEC presents a promising option for accessible ACD screening across diverse clinical environments.

FINANCIAL DISCLOSURES

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.