Haematologic Data Improves Long-Term Prediction Accuracy of Artificial Intelligence Models for Temporomandibular...

Area of Science:

• Biomedical informatics
• Artificial intelligence in healthcare
• Oral and maxillofacial surgery

Background:

• Temporomandibular disorder (TMD) presents complex challenges in predicting long-term treatment outcomes.
• Current predictive models often lack comprehensive data integration.
• The role of systemic inflammation in TMD prognosis requires further elucidation.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) model for predicting long-term treatment outcomes in TMD patients.
• To assess the impact of incorporating hematologic data on the predictive accuracy of AI models for TMD.
• To identify key clinical and hematologic features crucial for accurate TMD prognosis.

Main Methods:

• Utilized medical records of 132 TMD patients treated between 2013 and 2019.
• Employed a decision tree algorithm for feature selection, followed by a deep neural network (DNN) for model development.
• Evaluated model performance using accuracy and F1-score metrics.

Main Results:

• The decision tree model achieved 90.6% accuracy and an F1-score of 0.800.
• Subjective pain features and hematologic markers of systemic inflammation were identified as significant predictors.
• The DNN model's predictive performance improved with the addition of hematologic features, reaching 90.6% accuracy and an F1-score of 0.769.

Conclusions:

• Machine learning models demonstrate significant potential for predicting long-term TMD prognosis.
• Integrating subjective pain assessments and systemic hematologic markers enhances diagnostic systems for TMD.
• These findings support the development of etiology-based diagnostic systems to improve clinical decision-making and prognosis prediction in TMD.