Explainable Diagnosis Prediction through Neuro-Symbolic Integration

2024-10-01

arXiv.org

AI Summary

The “black-box” nature of medical AI models undermines clinical trust and hinders real-world deployment. Method: This paper proposes an interpretable neuro-symbolic model that integrates medical prior knowledge with data-driven learning, introducing Logic Neural Networks (LNNs) to diabetes prediction for the first time. The model embeds structured clinical rules via a learnable threshold mechanism and employs a multi-path, full-feature fusion architecture to jointly optimize accuracy and interpretability. Contribution/Results: Evaluated on a real-world diabetes prediction task, the model achieves 80.52% accuracy and an AUROC of 0.8457—significantly outperforming logistic regression, SVM, and random forests. Crucially, its learned weights and thresholds admit direct mapping to established clinical decision criteria, enabling symbol-level interpretability without sacrificing predictive performance. This work establishes a novel paradigm for co-optimizing accuracy and clinical interpretability in medical AI.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Diagnosis prediction is a critical task in healthcare, where timely and accurate identification of medical conditions can significantly impact patient outcomes. Traditional machine learning and deep learning models have achieved notable success in this domain but often lack interpretability which is a crucial requirement in clinical settings. In this study, we explore the use of neuro-symbolic methods, specifically Logical Neural Networks (LNNs), to develop explainable models for diagnosis prediction. Essentially, we design and implement LNN-based models that integrate domain-specific knowledge through logical rules with learnable thresholds. Our models, particularly $M_{ ext{multi-pathway}}$ and $M_{ ext{comprehensive}}$, demonstrate superior performance over traditional models such as Logistic Regression, SVM, and Random Forest, achieving higher accuracy (up to 80.52%) and AUROC scores (up to 0.8457) in the case study of diabetes prediction. The learned weights and thresholds within the LNN models provide direct insights into feature contributions, enhancing interpretability without compromising predictive power. These findings highlight the potential of neuro-symbolic approaches in bridging the gap between accuracy and explainability in healthcare AI applications. By offering transparent and adaptable diagnostic models, our work contributes to the advancement of precision medicine and supports the development of equitable healthcare solutions. Future research will focus on extending these methods to larger and more diverse datasets to further validate their applicability across different medical conditions and populations.

Problem

• Predictive Model
• Interpretable AI
• Healthcare Application

Innovation

• Logical Neural Networks
• Interpretable Machine Learning
• Disease Prediction