Summary of Global Contrastive Training For Multimodal Electronic Health Records with Language Supervision, by Yingbo Ma et al.
Global Contrastive Training for Multimodal Electronic Health Records with Language Supervision
by Yingbo Ma, Suraj Kolla, Zhenhong Hu, Dhruv Kaliraman, Victoria Nolan, Ziyuan Guan, Yuanfang Ren, Brooke Armfield, Tezcan Ozrazgat-Baslanti, Jeremy A. Balch, Tyler J. Loftus, Parisa Rashidi, Azra Bihorac, Benjamin Shickel
First submitted to arxiv on: 10 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computation and Language (cs.CL)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel framework for leveraging electronic health records (EHRs) to track personalized patient health trajectories through sequential deep learning. The challenge lies in effectively utilizing multiple modalities from EHRs, which possess complex characteristics such as high dimensionality, multimodality, sparsity, and temporal irregularities. To address this issue, the authors introduce a multimodal contrastive learning framework that integrates temporal cross-attention transformers with a dynamic embedding and tokenization scheme to learn multimodal feature representations. The framework also incorporates a global contrastive loss to align a patient’s multimodal feature representations with corresponding discharge summaries, enabling machine learning models to learn discriminative multimodal features via global contrasting. Experiments using a real-world EHR dataset demonstrate that the proposed framework outperforms state-of-the-art approaches on predicting postoperative complications. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps doctors use electronic health records (EHRs) to understand how patients get better or worse over time. The big challenge is combining all the different kinds of data in EHRs, like medical notes and test results, into something useful for making predictions. To solve this problem, researchers developed a new way to analyze EHRs using special computer algorithms that can handle the complexity of these records. They tested their method on a large dataset from a hospital system and found it was better at predicting when patients might develop complications after surgery than other methods. |
Keywords
» Artificial intelligence » Contrastive loss » Cross attention » Deep learning » Embedding » Machine learning » Tokenization
