What question did this study set out to answer?

This research aims to improve cross-modal patient data retrieval by incorporating graph-enhanced hashing networks.

May 20, 2026Journal of Computer-Aided Design & Computer Graphics0 citationsOpen Access

Graph Enhanced Hashing Networks for Cross-Modal Patient Retrieval

YGYifan Gu XYXuebing Yang CZChengzhang Zhu

Key Points

This research aims to improve cross-modal patient data retrieval by incorporating graph-enhanced hashing networks.
Developed a label encoder using co-occurrence relationships to construct a label graph.
Introduced a cross-modal attention fusion mechanism to enhance embedding representations.
Designed a similarity loss function to optimize both intra- and inter-modal relationships.
In the chest X-ray and diagnostic report dataset MIMIC-CXR, the proposed model outperformed three classical and four advanced hashing methods.
Achieved average precision improvements of 2.98%, 1.21%, 0.63%, and 0.53% at hash lengths of 16, 32, 64, and 128 bits respectively.
In the inverse task of retrieving X-ray images from reports, average precision improvements of 0.91%, 0.75%, 1.03%, and 0.57% were observed.

Abstract

通过对图像、文本等多种模态的患者数据进行检索，医生能够更全面、深入地分析患者的病程状态，做出更精准的临床诊断和治疗决策。随着医疗数据规模的迅猛增长、低存储需求与高效率检索的挑战，哈希方法逐渐在患者数据跨模态检索领域成为流行范式。针对现有跨模态哈希方法在利用患者诊断类别的多标签信息以及实现模态间协同建模方面存在的不足，提出引入跨模态注意力的图增强哈希网络模型。首先设计图信息增强的标签编码器，利用诊断类别标签的共现关系构建标签图，并通过图卷积网络有效地提取患者的多标签嵌入表征；为了缓解多标签矩阵的稀疏性问题并丰富融合表征的语义信息，引入图像、文本与多标签嵌入表征之间的跨模态注意力融合机制；最后将诊断类别的多标签共现关系与语义信息有效地融入患者图像、文本数据的哈希编码中，进一步提升哈希编码的判别能力，并设计了模态内与模态间关联互补的相似性损失函数。在胸部x光成像与诊断报告数据集mimic-Cxr上进行实验的结果表明，与3种经典跨模态哈希方法和4种先进的深度跨模态哈希方法相比，所提模型在多个编码位数下均展现出了性能优势。在x光成像检索诊断报告任务上，所提模型在哈希编码长度为16、32、64和128位时，平均精度均值相比次优方法分别提高了2.98%、1.21%、0.63%和0.53%；在诊断报告检索x光成像任务上，所提模型在哈希编码为上述4个长度时，平均精度均值相比次优方法分别提高了0.91%、0.75%、1.03%和0.57%。

Ask AI

Helpful

Bookmark

View Full Paper

Cite This Study

Gu et al. (2024) studied this question.

synapsesocial.com/papers/6a0d4f62f03e14405aa9ab89 https://doi.org/10.3724/sp.j.1089.2024-00351

Ask AI

Helpful

Bookmark

View Full Paper

May 20, 2026Journal of Computer-Aided Design & Computer Graphics0 citationsOpen Access

Graph Enhanced Hashing Networks for Cross-Modal Patient Retrieval

YGYifan Gu XYXuebing Yang CZChengzhang Zhu

Key Points

This research aims to improve cross-modal patient data retrieval by incorporating graph-enhanced hashing networks.
Developed a label encoder using co-occurrence relationships to construct a label graph.
Introduced a cross-modal attention fusion mechanism to enhance embedding representations.
Designed a similarity loss function to optimize both intra- and inter-modal relationships.
In the chest X-ray and diagnostic report dataset MIMIC-CXR, the proposed model outperformed three classical and four advanced hashing methods.
Achieved average precision improvements of 2.98%, 1.21%, 0.63%, and 0.53% at hash lengths of 16, 32, 64, and 128 bits respectively.
In the inverse task of retrieving X-ray images from reports, average precision improvements of 0.91%, 0.75%, 1.03%, and 0.57% were observed.

Abstract

Ask AI

Helpful

Bookmark

View Full Paper

Cite This Study

Gu et al. (2024) studied this question.

synapsesocial.com/papers/6a0d4f62f03e14405aa9ab89 https://doi.org/10.3724/sp.j.1089.2024-00351

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

Ask AI

Helpful

Bookmark

View Full Paper