Summary of Temporal Cross-attention For Dynamic Embedding and Tokenization Of Multimodal Electronic Health Records, by Yingbo Ma et al.
Temporal Cross-Attention for Dynamic Embedding and Tokenization of Multimodal Electronic Health Records
by Yingbo Ma, Suraj Kolla, Dhruv Kaliraman, Victoria Nolan, Zhenhong Hu, Ziyuan Guan, Yuanfang Ren, Brooke Armfield, Tezcan Ozrazgat-Baslanti, Tyler J. Loftus, Parisa Rashidi, Azra Bihorac, Benjamin Shickel
First submitted to arxiv on: 6 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 The paper presents a new approach to estimating personalized patient health trajectories using electronic health records (EHRs). The authors acknowledge that EHRs are high-dimensional, sparse, and multimodal, making it challenging to learn useful representations. To address this, they introduce a dynamic embedding and tokenization framework that incorporates novel methods for encoding time and sequential position with temporal cross-attention. This framework is integrated into a multitask transformer classifier with sliding window attention and outperforms baseline approaches on the task of predicting postoperative complications using multimodal data from three hospitals. The authors use EHRs to predict the occurrence of nine postoperative complications in over 120,000 major surgeries. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about using electronic health records (EHRs) to predict how patients will do after surgery. Right now, it’s hard to make good predictions because EHRs are very complex and have a lot of different types of information. The authors created a new way to understand this information better by combining time-based and sequential data with attention mechanisms. This approach was tested on real patient data from three hospitals and did much better than other methods at predicting postoperative complications. |
Keywords
* Artificial intelligence * Attention * Cross attention * Embedding * Tokenization * Transformer
