Cross-Domain Feature Expansion for Tabular Medical Data via Knowledge Graphs Injection
06:00 · July 1, 2026 · arXiv cs.AI RSS
Acquiring comprehensive cross-domain biomedical profiles is often costly and time-consuming, resulting in severe data scarcity in medical research. To address this challenge, we propose MedKGTab, a knowledge-injected framework specifically engineered for cross-domain feature expansion in tabular medical data. MedKGTab seeks to infer uncollected biomedical features from available ones by exploiting their inherent statistical dependencies and established medical correlations. By employing a row-column dual-attention mechanism, MedKGTab operates directly on raw structured tabular data, inherently capturing exact numerical distributions without the structural loss caused by tokenization. Crucially, MedKGTab integrates data-driven statistical priors with the SPOKE biomedical knowledge graph, achieving an optimal synergy between the data and knowledge channels. Within this synergy, the representations derived from the data channel are modulated by the injected biomedical knowledge, ensuring the final generated data are grounded in empirical medical research. Experimental results demonstrate that MedKGTab achieves high data fidelity and realistic data representation in cross-domain feature expansion. It outperforms both SOTA medical large models (e.g., Baichuan M3-plus) and specialized tabular models designed for medical data generation. Furthermore, MedKGTab consistently delivers superior performance across various data generation scenarios, whether inferring missing features within the same dataset or generalizing across different medical cohorts.
Summary
MedKGTab tackles data scarcity in biomedical research by generating missing cross-domain features in tabular medical datasets. Rather than collecting expensive new profiles, the framework infers uncollected variables from existing ones through a combination of observed statistical dependencies and established biomedical relationships encoded in external knowledge.
The model processes raw tabular records directly with a row-column dual-attention mechanism. This design preserves exact numerical distributions and avoids the structural degradation that occurs when tabular values are tokenized for language-model pipelines. At the same time, MedKGTab injects feature-level priors derived from the SPOKE biomedical knowledge graph into the attention layers as a controllable bias. The resulting representations therefore remain anchored in empirical data statistics while being modulated by medically grounded associations, preventing the dilution of knowledge that affects simpler graph-injection approaches.
Experiments show that these mechanisms produce higher-fidelity synthetic profiles than both specialized medical large language models, such as Baichuan M3-plus, and existing tabular generators. Performance gains hold when expanding features within a single cohort and when generalizing across different medical datasets, indicating that the synergy between statistical priors and graph knowledge improves reliability under limited sample sizes and domain shifts.
Why it matters
This research is highly relevant for Dutch AI researchers and health-tech enterprises dealing with electronic health records and medical data scarcity. By leveraging knowledge graphs to expand tabular data, it offers a robust methodology to enhance predictive modeling while navigating the strict data collection constraints typical in the EU.
