Surgical planning of horizontal strabismus using multiple output regression tree

Area of Science:

• Ophthalmology
• Medical Technology
• Artificial Intelligence

Background:

• Strabismus affects a significant population percentage, potentially causing vision and psychological issues.
• Surgical intervention is crucial for severe strabismus, aiming to restore binocular vision.
• Human error in pre-operative measurements is a leading cause of surgical failure.

Purpose of the Study:

• To develop an AI-driven method for precise surgical planning in horizontal strabismus.
• To predict accurate recoil and resection measurements for lateral and medial rectus muscles.
• To minimize human error in strabismus surgery planning.

Main Methods:

• Utilized Decision Tree Regressor algorithms for predictive modeling.
• Implemented two approaches: multiple single-target models and one multi-target model.
• Trained and validated the models using medical data for strabismus surgical planning.

Main Results:

• Achieved an average error of 0.66 mm in predicting surgical measurements across all procedures.
• Demonstrated high accuracy in predicting both resection and recoil values for horizontal strabismus.
• The AI method proved effective in predicting correction values based on data analysis.

Conclusions:

• Decision Tree Regressor algorithms are feasible for planning strabismus surgeries.
• The proposed method enhances surgical planning accuracy and reduces potential errors.
• This AI approach shows potential to surpass current state-of-the-art methods in strabismus surgery planning.