Keras Slim ResNet-Based Prediction of Furcation Management Recommendations Among Dentists

Area of Science:

• Dentistry
• Machine Learning
• Periodontology

Background:

• Furcation involvement complicates diagnosis and treatment of multirooted teeth.
• Management strategies have shifted, sometimes towards unnecessary extraction or prosthetic replacement.
• Accurate prediction models are needed to guide treatment decisions.

Purpose of the Study:

• To develop a machine learning model for predicting dentists' treatment recommendations for furcation involvement.
• To utilize Keras ResNet for forecasting management strategies based on questionnaire data.

Main Methods:

• The study included 437 South Indian dentists with at least 5 years of experience.
• Data were split into 80% training and 20% testing sets.
• Keras ResNet and Light Gradient Boosting models were compared using a data robot tool.

Main Results:

• The Keras ResNet and Light Gradient Boosting models achieved 84% accuracy.
• The models successfully predicted treatment suggestions for furcation-involved teeth.

Conclusions:

• A Keras Slim ResNet-based model accurately predicts dentists' management recommendations for furcation-involved teeth.
• The model can help predict referral patterns, encouraging general dentists to refer complex cases to periodontists.