Deep learning-based precision nursing training effectiveness evaluation method

Abstract

The invention pertains to the field of artificial intelligence technology and provides a deep learning-based precision nursing training effect evaluation method and system. The method synchronously collects four-dimensional data—operational behavior, physiological state, instrument usage, and environmental context—through a multimodal sensor array, and generates a structured representation of operational behavior, a fused representation of physiological state, a quantified representation of instrument operation compliance, and contextual constraint conditions, respectively. These four types of representations are input into a multi-branch heterogeneous graph neural network to construct a nursing operational skill knowledge graph, achieving deep fusion of multi-source heterogeneous data through node embedding and cross-modal edge relationship modeling. A dynamic gated attention mechanism is employed to adaptively assign weights to each evaluation dimension, generating a comprehensive evaluation score vector. An interpretable decoder outputs a structured feedback report that includes localization of skill deficiencies, traceability of operational deviations, and improvement suggestions. The system comprises corresponding functional modules to realize holographic perception, dynamic evaluation, and precise feedback, addressing issues in existing technologies such as single evaluation dimensions, shallow multi-source data fusion, insufficient model generalization capability, and delayed feedback. This significantly enhances the objectivity, granularity, and clinical guidance value of nursing training evaluation.