Enhanced Tomographic Assessment to Detect Corneal Ectasia Based on Artificial Intelligence.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Corneal Imaging

Background:

• Corneal ectasia susceptibility detection is crucial for preventing post-surgical complications.
• Current diagnostic methods may not fully capture the risk of developing ectasia.
• Tomographic data offers a rich source for developing advanced diagnostic tools.

Purpose of the Study:

• To enhance the detection of corneal ectasia susceptibility using tomographic data.
• To develop and validate an artificial intelligence (AI) model for improved ectasia risk assessment.
• To compare the performance of the AI model against existing diagnostic indices.

Main Methods:

• A multicenter case-control study involving 5 clinics across South America, the US, and Europe.
• AI models were generated using Pentacam HR parameters to analyze preoperative data from stable LASIK, ectasia susceptibility, and keratoconus groups.
• Model accuracy was validated on independent datasets of stable LASIK and unoperated patients with very asymmetric ectasia.

Main Results:

• The random forest (RF) model, named Pentacam Random Forest Index (PRFI), achieved 100% sensitivity for clinical ectasia.
• PRFI demonstrated a superior area under the curve (AUC) of 0.992 compared to Belin/Ambrósio deviation (BAD-D; AUC=0.960).
• An optimized PRFI cutoff showed high sensitivity for detecting susceptibility to post-LASIK ectasia and very asymmetric ectasia.

Conclusions:

• The PRFI significantly enhances the diagnosis of corneal ectasia.
• This AI-driven index offers a more accurate method for identifying ectasia risk.
• Future research should integrate PRFI with biomechanical parameters and laser vision correction impact for comprehensive ectasia risk assessment.