# PromptEHR
[](https://badge.fury.io/py/promptehr)
[](https://pepy.tech/project/promptehr)
Wang, Zifeng and Sun, Jimeng. (2022). PromptEHR: Conditional Electronic Healthcare Records Generation with Prompt Learning. EMNLP'22.
# News
- [2023/01/08] `PromptEHR` is now integrated into [`PyTrial`](https://github.com/RyanWangZf/PyTrial) with a complete [documentation](https://pytrial.readthedocs.io/en/latest/trial_simulation/sequence/promptehr.html) and [example](https://colab.research.google.com/drive/1EbzLdSwTrbgsEgz8z70qzTLQWiPWlyRm?usp=sharing), please check! New version with bugs fixed is also released!
# Usage
Get pretrained PromptEHR model (learned on MIMIC-III sequence EHRs) in three lines:
```python
from promptehr import PromptEHR
model = PromptEHR()
model.from_pretrained()
```
A jupyter example is available at https://github.com/RyanWangZf/PromptEHR/blob/main/example/demo_promptehr.ipynb.
# How to install
Install the correct `PyTorch` version by referring to https://pytorch.org/get-started/locally/.
Then try to install `PromptEHR` by
```bash
pip install git+https://github.com/RyanWangZf/PromptEHR.git
```
or
```bash
pip install promptehr
```
# Load demo synthetic EHRs (generated by PromptEHR)
```python
from promptehr import load_synthetic_data
data = load_synthetic_data()
```
# Use PromptEHR for generation
```python
from promptehr import SequencePatient
from promptehr import load_synthetic_data
from promptehr import PromptEHR
# init model
model = PromptEHR()
model.from_pretrained()
# load input data
demo = load_synthetic_data(n_sample=1000) # we have 10,000 samples in total
# build the standard input data for train or test PromptEHR models
seqdata = SequencePatient(data={'v':demo['visit'], 'y':demo['y'], 'x':demo['feature'],},
metadata={
'visit':{'mode':'dense'},
'label':{'mode':'tensor'},
'voc':demo['voc'],
'max_visit':20,
}
)
# you can try to fit on this data by
# model.fit(seqdata)
# start generate
# n: the target total number of samples to generate
# n_per_sample: based on each sample, how many fake samples will be generated
# the output will have the same format of `SequencePatient`
fake_data = model.predict(seqdata, n=1000, n_per_sample=10)
```
Raw data
{
"_id": null,
"home_page": "https://github.com/RyanWangZf/PromptEHR",
"name": "PromptEHR",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "synthetic data,healthcare,EHR,deep learning,AI",
"author": "Zifeng Wang",
"author_email": "zifengw2@illinois.edu",
"download_url": "https://files.pythonhosted.org/packages/ad/bb/9aadbcb822c0d5c358c49b9d4486639821bb389c8ac43f30eb2260611dca/PromptEHR-0.0.6.tar.gz",
"platform": null,
"description": "# PromptEHR\n[](https://badge.fury.io/py/promptehr)\n[](https://pepy.tech/project/promptehr)\n\n\n\nWang, Zifeng and Sun, Jimeng. (2022). PromptEHR: Conditional Electronic Healthcare Records Generation with Prompt Learning. EMNLP'22.\n\n# News\n- [2023/01/08] `PromptEHR` is now integrated into [`PyTrial`](https://github.com/RyanWangZf/PyTrial) with a complete [documentation](https://pytrial.readthedocs.io/en/latest/trial_simulation/sequence/promptehr.html) and [example](https://colab.research.google.com/drive/1EbzLdSwTrbgsEgz8z70qzTLQWiPWlyRm?usp=sharing), please check! New version with bugs fixed is also released!\n\n\n# Usage\n\nGet pretrained PromptEHR model (learned on MIMIC-III sequence EHRs) in three lines:\n\n```python\nfrom promptehr import PromptEHR\n\nmodel = PromptEHR()\n\nmodel.from_pretrained()\n```\n\nA jupyter example is available at https://github.com/RyanWangZf/PromptEHR/blob/main/example/demo_promptehr.ipynb.\n\n\n\n# How to install\n\nInstall the correct `PyTorch` version by referring to https://pytorch.org/get-started/locally/.\n\nThen try to install `PromptEHR` by\n\n```bash\npip install git+https://github.com/RyanWangZf/PromptEHR.git\n```\n\nor\n\n```bash\npip install promptehr\n```\n\n\n\n# Load demo synthetic EHRs (generated by PromptEHR)\n\n```python\nfrom promptehr import load_synthetic_data\ndata = load_synthetic_data()\n```\n\n\n\n# Use PromptEHR for generation\n\n```python\nfrom promptehr import SequencePatient\nfrom promptehr import load_synthetic_data\nfrom promptehr import PromptEHR\n\n# init model\nmodel = PromptEHR()\nmodel.from_pretrained()\n\n# load input data\ndemo = load_synthetic_data(n_sample=1000) # we have 10,000 samples in total\n\n# build the standard input data for train or test PromptEHR models\nseqdata = SequencePatient(data={'v':demo['visit'], 'y':demo['y'], 'x':demo['feature'],},\n metadata={\n 'visit':{'mode':'dense'},\n 'label':{'mode':'tensor'}, \n 'voc':demo['voc'],\n 'max_visit':20,\n }\n )\n# you can try to fit on this data by\n# model.fit(seqdata)\n\n# start generate\n# n: the target total number of samples to generate\n# n_per_sample: based on each sample, how many fake samples will be generated\n# the output will have the same format of `SequencePatient`\nfake_data = model.predict(seqdata, n=1000, n_per_sample=10)\n```\n\n",
"bugtrack_url": null,
"license": "",
"summary": "Sequence patient electronic healthcare record generation with large language models (LLMs) as the neural database.",
"version": "0.0.6",
"project_urls": {
"Homepage": "https://github.com/RyanWangZf/PromptEHR"
},
"split_keywords": [
"synthetic data",
"healthcare",
"ehr",
"deep learning",
"ai"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "bd71067429415052340eba76d82a0adf9135b59b8cf4781703a45e123b1372cd",
"md5": "d303d6162a1150b811c1df1918e4c90f",
"sha256": "a2eae33478b51b3a84ad8b9d676dd9acf439250dfe1908f07c0257749654e775"
},
"downloads": -1,
"filename": "PromptEHR-0.0.6-py3-none-any.whl",
"has_sig": false,
"md5_digest": "d303d6162a1150b811c1df1918e4c90f",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 312862,
"upload_time": "2023-06-08T05:26:16",
"upload_time_iso_8601": "2023-06-08T05:26:16.423154Z",
"url": "https://files.pythonhosted.org/packages/bd/71/067429415052340eba76d82a0adf9135b59b8cf4781703a45e123b1372cd/PromptEHR-0.0.6-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "adbb9aadbcb822c0d5c358c49b9d4486639821bb389c8ac43f30eb2260611dca",
"md5": "f11f68f386c719591c21580f9969f260",
"sha256": "31638dbdf274b35a37b8930cc47619b6467ef689cd9205204a04b01c83f20fe6"
},
"downloads": -1,
"filename": "PromptEHR-0.0.6.tar.gz",
"has_sig": false,
"md5_digest": "f11f68f386c719591c21580f9969f260",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 230161,
"upload_time": "2023-06-08T05:26:18",
"upload_time_iso_8601": "2023-06-08T05:26:18.176011Z",
"url": "https://files.pythonhosted.org/packages/ad/bb/9aadbcb822c0d5c358c49b9d4486639821bb389c8ac43f30eb2260611dca/PromptEHR-0.0.6.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-06-08 05:26:18",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "RyanWangZf",
"github_project": "PromptEHR",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [],
"lcname": "promptehr"
}
JSON
