Scholarlink

<p>This research presents a fully integrated neuro-symbolic system for clinical decision support, focusing on successfully diagnosing comorbid neurological presentations. The framework ingests complex medical information such as EEG, fMRI, structural MRI, and sentiment analysis, while functioning within structured medical ontologies such as SNOMED CT or ICD-11. Through supporting joint perception and reasoning, the neuro-symbolic framework addresses some of the key issues in clinical AI: explainability, generalization, and semantic coherence. The model connects functional deep neural perception modules to an associative symbolic logic engine, which acts on dynamically evolving knowledge graphs to provide an intelligent assess-and-adapt network. The architecture supports bi-directional information flow, in a way that perception can inform reasoning and vice-versa. In contrast to previous AI architectures that may be modular, even if integrated or can only be modal logic, the proposed framework supports real-time symbolic logic induction. It dynamically revises its logic based on new patterns observed in the various source data and context, while the neural embeddings receive symbolic feedback. The main contributions of this study include: (i) the creation of a bidirectional neuro-symbolic framework that enables symbolic reasoning and neural perception to cooperate in real-time; (ii) a testbed application using real-world neurologic patient data (EEG, MRI); (iii) a dynamic knowledge graph enabling in-coming data representation and use of logical reasoning for interpretable and trusted decision making; and (iv) symbolic integration with medical ontologies to validate clinical validity. The framework is validated through mathematical modeling and empirical testing, demonstrating high diagnostic accuracy, logical consistency between symbolic and neural reasoning, and symbolic fidelity. This solution also offers motivation and direction for demonstrating legitimate, explainable AI in neurology and other high-risk healthcare contexts by combining subsymbolic learning with medical reasoning.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A bidirectional neuro-symbolic framework for clinical decision support via dynamic integration of deep learning and symbolic reasoning

  • 期刊论文

  • 发表日期：

    2026年3月7日

查看全文
  • Rinku Chavda,
  • K. Suresh,
  • Sushil Kumar,
  • K. L. Raghavender Reddy,
  • B. Jayaprakash,
  • Prabhat Kumar Sahu,
  • B. Bharathi,
  • Devendra Singh,
  • Saurabh Namdev

摘要

This research presents a fully integrated neuro-symbolic system for clinical decision support, focusing on successfully diagnosing comorbid neurological presentations. The framework ingests complex medical information such as EEG, fMRI, structural MRI, and sentiment analysis, while functioning within structured medical ontologies such as SNOMED CT or ICD-11. Through supporting joint perception and reasoning, the neuro-symbolic framework addresses some of the key issues in clinical AI: explainability, generalization, and semantic coherence. The model connects functional deep neural perception modules to an associative symbolic logic engine, which acts on dynamically evolving knowledge graphs to provide an intelligent assess-and-adapt network. The architecture supports bi-directional information flow, in a way that perception can inform reasoning and vice-versa. In contrast to previous AI architectures that may be modular, even if integrated or can only be modal logic, the proposed framework supports real-time symbolic logic induction. It dynamically revises its logic based on new patterns observed in the various source data and context, while the neural embeddings receive symbolic feedback. The main contributions of this study include: (i) the creation of a bidirectional neuro-symbolic framework that enables symbolic reasoning and neural perception to cooperate in real-time; (ii) a testbed application using real-world neurologic patient data (EEG, MRI); (iii) a dynamic knowledge graph enabling in-coming data representation and use of logical reasoning for interpretable and trusted decision making; and (iv) symbolic integration with medical ontologies to validate clinical validity. The framework is validated through mathematical modeling and empirical testing, demonstrating high diagnostic accuracy, logical consistency between symbolic and neural reasoning, and symbolic fidelity. This solution also offers motivation and direction for demonstrating legitimate, explainable AI in neurology and other high-risk healthcare contexts by combining subsymbolic learning with medical reasoning.