Reproduce “Identification of Pediatric Respiratory Diseases Using Fine-grained Diagnosis System” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Gang Yu, Zhongzhi Yu, Yemin Shi, Yingshuo Wang, Xiaoqing Liu, Zheming Li, Yonggen Zhao, Fenglei Sun and 2 more

Researchers from Zhejiang University School of Medicine and Deepwise AI Lab developed a two-stage deep learning system to identify pediatric respiratory diseases using only unstructured clinical notes and routine test results. This system achieved a mean Average Precision (mAP) of 0.819 in differentiating pneumonia, RTI, bronchitis, and asthma, surpassing baseline models including BERT and other multi-modal fusion approaches.

Key insights

  • Unstructured clinical notes and routine test results, when effectively processed, can serve as sufficient input for accurate pediatric respiratory disease diagnosis in resource-constrained environments.
  • Adaptive Feature Infusion (AFI) and Multi-Modal Attentive Fusion modules are essential for synergistically combining heterogeneous textual and numerical clinical data, yielding superior diagnostic performance.
  • Domain-specific deep learning architectures, such as those incorporating ResBlocks and targeted attention mechanisms, demonstrate advantages over general-purpose large language models like BERT for the specific language and data characteristics of clinical notes.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.