What amount of data is required to develop artificial intelligence that can accurately predict soft tissue changes...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Orthodontics

Background:

• Orthognathic surgery involves complex soft tissue changes.
• Accurate prediction of these changes is crucial for treatment planning.
• Current AI models may require substantial data for reliable predictions.

Purpose of the Study:

• To determine the sample size needed for developing accurate AI models for predicting soft tissue changes after orthognathic surgery.
• To establish a sample size calculation method for AI in this field.
• To achieve clinically acceptable prediction accuracy (≤1.5 mm error).

Main Methods:

• Utilized data from 705 patients undergoing combined surgical-orthodontic treatment.
• Generated 40 AI models using a deep-learning algorithm with varying sample sizes (75-700 patients).
• Evaluated prediction error on the lower lip landmark and used linear regression to correlate sample size with error.

Main Results:

• Prediction error decreased as sample size increased.
• A sample size exceeding 1700 datasets was estimated to be necessary for AI models with <1.5 mm prediction error.
• The lower lip was used as the primary outcome variable for evaluation.

Conclusions:

• A substantial amount of orthognathic surgery data is essential for developing AI software.
• This AI software aims to visualize surgical treatment objectives with high accuracy.
• The findings provide a basis for sample size estimation in similar AI-driven medical applications.