[CS 598 DL4H] Generating Faithful and High Quality Patient Summaries with LLMs

[caiqile]

Contributor: Ethan Cai (epcai2), Abrar Ahmed (abrara3), Carina Zhang (xz125)

Contribution Type: Dataset + Task

Original Paper:
Hegselmann, S.; Shen, S. Z.; Gierse, F.; Agrawal, M.; Son-
tag, D.; and Jiang, X. 2024. A Data-Centric Approach To
Generate Faithful and High Quality Patient Summaries with
Large Language Models. In Proceedings of the Conference
on Health, Inference, and Learning (CHIL), volume 248 of
Proceedings of Machine Learning Research, 339–379.
https://arxiv.org/pdf/2402.15422

Description:
Implements MIMIC4NoteExtDIBHCDataset, a dataset that includes the Brief Hospital Course and Discharge Instructions columns to represent relevant information from MIMIC4Note that can be used as input to LLM for summary generation.

Implements BHCSummarizationTask to input Brief Hospital Course into an LLM and ask for a summary.

Implements HallucinationDetectionTask to automatically detect the number of hallucinations in an AI generated summary.

Files to Review:

• pyhealth/tasks/mimic4_note_ext_dibhc_tasks.py - Task Implementation
• tests/core/test_mimic4_note_ext_dibhc_tasks.py - Task Testcases
• pyhealth/datasets/mimic4noteextdibhc.py - Dataset Implementation
• tests/core/test_mimic4noteextdibhc.py - Dataset Testcases
• examples/mimic4_note_ext_dibhc_hallucination_detection_logreg.py - Example usage + Ablation study

[Jathurshan0330]

Good PR

[github-actions]

This PR has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs.