Objective: Accurate assessment of the repair outcomes of maxillofacial trauma is crucial for surgical success. However, existing assessment methods heavily rely on the subjective experience of doctors. Although automatic assessment methods based on deep learning have high precision, they lack interpretability and are difficult to gain clinical trust. Methods: This paper proposes a hybrid intelligent assessment system integrating symbolic machine learning and case-based reasoning (CBR). The system adopts a dual-driven architecture: the symbolic machine learning module learns interpretable assessment rules from expert-annotated CT images and clinical guidelines, forming a transparent decision-making knowledge base; the CBR module constructs a historical case base and provides analogical references for complex cases through similarity retrieval. The system ultimately integrates rule-based reasoning and case-based analogy results based on a confidence-weighted strategy, generating assessment reports that combine quantitative scores with logical basis. Results: In the validation experiment with 512 clinical data cases, the assessment accuracy of this system reached 94.7%, significantly higher than traditional deep-learning models (ResNet-3D: 89.2%) and single-reasoning methods (symbolic-only: 91.0%, CBR-only: 89.6%). Ablation experiments further demonstrated that the integration of the two modules has a synergistic enhancement effect (P < 0.05). Conclusion: This study first realized the effective integration of symbolic reasoning and case-based analogy in maxillofacial trauma assessment, providing a new idea for solving the “black-box” problem in medical artificial intelligence (AI) and having strong clinical practicability and promotion value.
Li et al. (Mon,) studied this question.