Scholarlink

<p>Accurate and timely patient diagnosis in healthcare is both critical and challenging due to the vast complexity of medical data. This paper introduces a deep learning approach to patient interpretation and diagnosis, utilizing improved Graph Neural Network (GNN) methods within a collaborative recommendation framework. Traditional diagnostic systems often rely on rule-based or statistical methods, which may struggle to encompass the dynamic interplay of diverse patient features, medical history, and evolving medical knowledge. This paper presents three novel methods for patient interpretation and diagnosis by harnessing the power of GNNs within a collaborative recommendation framework. We developed the Graph-based Hybrid Recommendation System (GHRS), Graph Neural Network-based Collaborative Filtering - No Attention (GCFNA), and Graph Neural Network-based Collaborative Filtering - Yes Attention (GCFYA). A key novelty of our approach is applying GNNs to capture intricate relationships within patient data through adaptive, data-driven methods. This empowers a more accurate, interpretable, and collaborative approach to diagnosis, moving beyond traditional methodologies. Our system employs GNN-based Collaborative Filtering (CF) to model complex patient-patient and patient-symptom relationships within a comprehensive graph structure. This representation not only accommodates various types of medical data but also allows for the integration of external medical knowledge, effectively capturing nuanced disease manifestations and symptom patterns. Another significant innovation is the system’s emphasis on providing interpretable recommendations. It elucidates the rationale behind diagnostic suggestions, addressing a common challenge in healthcare where the reasoning behind recommendations is often obscured. Results on one real hospital dataset and two public datasets were evaluated and compared with other recommendation methods; our methods show better results in all evaluation factors. This interpretability is crucial for gaining the trust of medical professionals and ensuring the acceptance and adherence to recommended treatments. Furthermore, our focus on elderly patients, a demographic with unique healthcare challenges, and the inclusion of diverse factors such as family history, lifestyle, and past medical records, distinguishes our work. The complexity and variety of diseases in this context, as well as the presence of incomplete or missing patient information, make our approach particularly relevant.</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

Enhancing chronic disease management: hybrid graph networks and explainable AI for intelligent diagnosis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月4日

查看全文
  • Muhammad Aamir,
  • Yang Ke Yu,
  • Nomica Choudhry,
  • Uzair Aslam Bhatti

摘要

Accurate and timely patient diagnosis in healthcare is both critical and challenging due to the vast complexity of medical data. This paper introduces a deep learning approach to patient interpretation and diagnosis, utilizing improved Graph Neural Network (GNN) methods within a collaborative recommendation framework. Traditional diagnostic systems often rely on rule-based or statistical methods, which may struggle to encompass the dynamic interplay of diverse patient features, medical history, and evolving medical knowledge. This paper presents three novel methods for patient interpretation and diagnosis by harnessing the power of GNNs within a collaborative recommendation framework. We developed the Graph-based Hybrid Recommendation System (GHRS), Graph Neural Network-based Collaborative Filtering - No Attention (GCFNA), and Graph Neural Network-based Collaborative Filtering - Yes Attention (GCFYA). A key novelty of our approach is applying GNNs to capture intricate relationships within patient data through adaptive, data-driven methods. This empowers a more accurate, interpretable, and collaborative approach to diagnosis, moving beyond traditional methodologies. Our system employs GNN-based Collaborative Filtering (CF) to model complex patient-patient and patient-symptom relationships within a comprehensive graph structure. This representation not only accommodates various types of medical data but also allows for the integration of external medical knowledge, effectively capturing nuanced disease manifestations and symptom patterns. Another significant innovation is the system’s emphasis on providing interpretable recommendations. It elucidates the rationale behind diagnostic suggestions, addressing a common challenge in healthcare where the reasoning behind recommendations is often obscured. Results on one real hospital dataset and two public datasets were evaluated and compared with other recommendation methods; our methods show better results in all evaluation factors. This interpretability is crucial for gaining the trust of medical professionals and ensuring the acceptance and adherence to recommended treatments. Furthermore, our focus on elderly patients, a demographic with unique healthcare challenges, and the inclusion of diverse factors such as family history, lifestyle, and past medical records, distinguishes our work. The complexity and variety of diseases in this context, as well as the presence of incomplete or missing patient information, make our approach particularly relevant.