Using a multilayer perceptron in intraocular lens power calculation.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Cataract surgery requires precise intraocular lens (IOL) power calculation for optimal visual outcomes.
• Current methods for IOL power prediction can lead to refractive surprises.
• Advancements in artificial intelligence offer potential for enhanced predictive accuracy.

Purpose of the Study:

• To evaluate the capability of a multilayer perceptron (MLP) in calculating intraocular lens (IOL) power.
• To assess the MLP's effectiveness in achieving target postoperative refraction after cataract surgery.
• To determine if MLP can improve upon existing IOL power prediction methods.

Main Methods:

• A retrospective review of 15,728 cataract surgeries using a specific foldable IOL was conducted.
• Data including axial length, keratometry, and refractive outcomes were used to train an MLP.
• The MLP model was designed to predict the required IOL power based on preoperative and postoperative data.

Main Results:

• The trained MLP accurately predicted IOL power with an error of less than 0.5 diopters in over 95% of cases.
• The model demonstrated high predictive accuracy even for surgeons not included in the training dataset.
• The MLP's performance suggests a significant improvement over conventional IOL power calculation techniques.

Conclusions:

• The multilayer perceptron (MLP) demonstrates high accuracy in predicting intraocular lens (IOL) power.
• This AI-driven approach shows feasibility for a new method in refractive surgery prediction.
• The MLP method has the potential to achieve lower prediction errors than current techniques.