ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
PyHealth contribution documentation:
2
Why should you contribute?We have a wide variety of projects that need help, and are actively looking for researchers. The goal of these projects is to produce high quality research towards improving healthcare outcomes. In the past, we have published papers around various important healthcare topics. These are great ways to highlight to employers new skills, get recommendation letters, etc. towards those that want to work at the intersection of AI and health.
Note: We're actively adding new projects to this page
3
How can I join?Please make a high quality pull request, and join the discord server (documentation above)
4
5
Project TitleDescription/AbstractContact (s)Project Details Link Existing PyHealth Researchers
6
Rare Disease Differential DiagnosisDespite rare diseases affecting 1 in 10 Americans, their differential diagnosis remains challenging. Due to their impressive recall abilities, large language models (LLMs) have been recently explored for differential diagnosis. Existing approaches to evaluating LLM-based rare disease diagnosis suffer from two critical limitations: they rely on idealized clinical case studies that fail to capture real-world clinical complexity, or they use ICD codes as disease labels, which significantly undercounts rare diseases since many lack direct mappings to comprehensive rare disease databases like Orphanet. To address these limitations, we explore MIMIC-RD, a rare disease differential diagnosis bench mark constructed by directly mapping clinical text entities to Orphanet. Our methodology involved an initial LLM-based mining process followed by validation from four medical annotators to confirm identified entities were genuine rare diseases. We evaluated various models on our dataset of 145 patients and found that current state-of-the-art LLMs perform poorly onrare disease differential diagnosis, highlighting the substantial gap between existing capabilities and clinical needs. From our findings, we outline several future steps towards improving differential diagnosis of rare diseases.John Wu johnwu3@illinois.edu, Adam CrossContinuing on an earlier link. Please see the precursor preliminary paper https://openreview.net/forum?id=CrZyNUfWHpZilal Eiz Aldin
7
Conformal EEG for Dynamic Patient DistributionsJathurshan Pradeepkumar jp65@illinois.edu
8
Visual Medical Reasoning
9
Clinical Reasoning Step by Step
10
Prostate Cancer Clinical Validation ProjectAbraham Arellano, Umesh KumarAbraham Arellano, Umesh Kumar
11
 Pediatric Tracheitis Detection & Antibiotic ResponseDevelop models to detect pediatric bacterial tracheitis from radiology reports and culture data, and predict which antibiotics will work best. Students will work on EHR labeling, predictive modeling, and treatment response analysis.Adam Cross
12
Clinical Trial Similarity SearchMultifaceted need of trial similarity search: 1) trial design, 2) patient-to-trial matching, etc. How to handle diifferent views of trial similarity? One trial representation might not be good enough for different needs. Trisha Das trishad@illinois.edu
13
Diabetic Ketoacidosis (DKA) Early PredictionDevelop AI models to predict the onset of DKA in children, integrating labs, vitals, and clinical notes. Students will help design multimodal models, test generalizability, and handle multimodal EHR data + glucometer data.Dr. Adam Cross (arcross@uic.edu)
14
Asthma Management & AdherenceUse novel data sources (EHR + medication adherence data) to predict asthma exacerbations and personalize management. Students will extract data, analyze adherence patterns, and build predictive models.Adam Cross
15
Retinopathy of Prematurity (ROP)Work with retinal images and EHR data to detect and stratify risk of ROP in premature infants. Students will gain experience in image analysis, multimodal learning, and clinical decision support.Xu Cao
16
Rural health and SDOHOne of the largest challenge in rural health is sparsity in data. In other words, more signal is needed in order to understand a patient's healthcare journey. One idea would be to use clinical notes to mine for sDOH factors, with a focus on rural health. Another idea is to attempt to generate synthetic clinical notes (probably much trickier) that can hopefully represent a population that is older and more resistant to care. William Pang (williampangbest1@gmail.com)
17
Frequency and temporal position prediction based SSL for biosignalsTBDJathurshan Pradeepkumar jp65@illinois.edu
18
Adaptive Mixture-of-Experts for Cross-Modal Biosignal Unified RepresentationTBDJathurshan Pradeepkumar jp65@illinois.edu
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100