Create README
Browse files
README.md
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: "en"
|
3 |
+
tags:
|
4 |
+
- bert
|
5 |
+
- medical
|
6 |
+
- clinical
|
7 |
+
- mortality
|
8 |
+
thumbnail: "https://core.app.datexis.com/static/paper.png"
|
9 |
+
---
|
10 |
+
|
11 |
+
# CORe Model - Clinical Mortality Risk Prediction
|
12 |
+
|
13 |
+
## Model description
|
14 |
+
|
15 |
+
The CORe (_Clinical Outcome Representations_) model is introduced in the paper [Clinical Outcome Predictions from Admission Notes using Self-Supervised Knowledge Integration](https://www.aclweb.org/anthology/2021.eacl-main.75.pdf).
|
16 |
+
It is based on BioBERT and further pre-trained on clinical notes, disease descriptions and medical articles with a specialised _Clinical Outcome Pre-Training_ objective.
|
17 |
+
|
18 |
+
This model checkpoint is **fine-tuned on the task of mortality risk prediction**.
|
19 |
+
The model expects patient admission notes as input and outputs the predicted risk of in-hospital mortality.
|
20 |
+
|
21 |
+
#### How to use CORe Diagnosis Prediction
|
22 |
+
|
23 |
+
You can load the model via the transformers library:
|
24 |
+
```
|
25 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
26 |
+
tokenizer = AutoTokenizer.from_pretrained("bvanaken/CORe-clinical-mortality-prediction")
|
27 |
+
model = AutoModelForSequenceClassification.from_pretrained("bvanaken/CORe-clinical-mortality-prediction")
|
28 |
+
```
|
29 |
+
|
30 |
+
The following code shows an inference example:
|
31 |
+
|
32 |
+
```
|
33 |
+
input = "CHIEF COMPLAINT: Headaches\n\nPRESENT ILLNESS: 58yo man w/ hx of hypertension, AFib on coumadin presented to ED with the worst headache of his life."
|
34 |
+
|
35 |
+
tokenized_input = tokenizer(input, return_tensors="pt")
|
36 |
+
output = model(**tokenized_input)
|
37 |
+
|
38 |
+
import torch
|
39 |
+
predictions = torch.softmax(output.logits.detach(), dim=1)
|
40 |
+
mortality_risk_prediction = predictions[0][1].item()
|
41 |
+
```
|
42 |
+
|
43 |
+
|
44 |
+
### More Information
|
45 |
+
|
46 |
+
For all the details about CORe and contact info, please visit [CORe.app.datexis.com](http://core.app.datexis.com/).
|
47 |
+
|
48 |
+
### Cite
|
49 |
+
|
50 |
+
```bibtex
|
51 |
+
@inproceedings{vanaken21,
|
52 |
+
author = {Betty van Aken and
|
53 |
+
Jens-Michalis Papaioannou and
|
54 |
+
Manuel Mayrdorfer and
|
55 |
+
Klemens Budde and
|
56 |
+
Felix A. Gers and
|
57 |
+
Alexander Löser},
|
58 |
+
title = {Clinical Outcome Prediction from Admission Notes using Self-Supervised
|
59 |
+
Knowledge Integration},
|
60 |
+
booktitle = {Proceedings of the 16th Conference of the European Chapter of the
|
61 |
+
Association for Computational Linguistics: Main Volume, {EACL} 2021,
|
62 |
+
Online, April 19 - 23, 2021},
|
63 |
+
publisher = {Association for Computational Linguistics},
|
64 |
+
year = {2021},
|
65 |
+
}
|
66 |
+
```
|