Ai-Driven Predictions Vs. Clinical Reality In Multifocal Corneal Profiles For Myopic And Hyperopic Presbyopia

To evaluate the predictive accuracy of AI-driven topography modeling (Schwind Foresight) in achieving the negative spherical aberration (SA) targets required for effective depth-of-focus. In PresbyMAX treatments(femto-lasik), near-vision performance is the primary driver of patient satisfaction. This study compares the AI-simulated postoperative wavefront with actual clinical outcomes in both myopic and hyperopic cohorts to determine if a systematic "SA Deficit" exists, thereby providing a data-driven framework for optimizing future multifocal laser vision correction nomograms.

EyeLaser Clinic Vienna, SwissLaser Clinic Warsaw & Krakow

Nineteen eyes of six nineteen patients (10 hyperopes, mean SE +1.87D; 9 myopes, mean SE -3.78D) underwent monocular PresbyMAX (Schwind Amaris) in the non-dominant eye with a addition betwee 1.6-1.9D. Preoperative corneal topography (MS39) was processed through the Foresight AI-software to generate a predicted postoperative wavefront profile (6.0mm pupil). One month postoperatively, actual corneal wavefront measurements (MS39) were compared to these AI estimates. Key metrics included the induction of negative SA (Z4,0), Coma (Z3,-1}, (Z3,1), and total higher-order aberration (HOA) RMS.

While the AI-model targeted a mean negative SA shift of -0.55 µm, the actual achieved postoperative SA was significantly lower at -0.34 µ (p < 0.05), a 38% under-induction. This "SA Deficit" was most prominent in hyperopic eyes (42%) compared to myopic eyes. Conversely, whereas the AI predicted stable or reduced asymmetric aberrations, actual postoperative Coma tripled from 0.21 µm to 0.43 µm. Total HOA RMS was 18% higher than the AI-predicted value (Actual: 1.05 µm vs. AI: 0.89 µm), suggesting a more irregular multifocal surface than modeled.

There is a systematic discrepancy between AI-simulated corneal profiles and actual clinical outcomes in PresbyMAX treatments. The 38% under-induction of negative spherical aberration suggests that the intended depth-of-focus is frequently "undershot," potentially limiting near-vision range. This gap likely reflects the influence of corneal biomechanics and epithelial remodeling—factors not yet fully integrated into current AI algorithms. To optimize patient satisfaction, surgeons should consider "over-planning" the target negative SA in the Foresight software. These findings serve as a vital benchmark for refining AI-driven presbyopic planning.