retuve
The first Open, Collaborative Framework for Infant Hip Health using AI
Retuve (from the scottish gaelic Ri taobh
meaning beside
) is a framework for analysing infant hips. It is designed to be a flexible and extensible framework that can be used by developers, AI researchers and clinicians.
It takes in raw hip Ultrasound and X-Ray images, and outputs a report with the labelled images, and the results, exactly as a clinician would.
Attribution of the above Ultrasound Images: Case courtesy of Ryan Thibodeau from https://radiopaedia.org 172535 (https://radiopaedia.org/cases/172535)
Attribution of the above X-Ray Images: Fraiwan, Mohammad; Al-Kofahi, Noran; Hanatleh, Omar; ibnian, ali (2022), “A dataset of DDH x-ray images”, Mendeley Data, V2, doi: 10.17632/jf3pv98m9g.2
Docs
See our docs online at https://radoss-org.github.io/retuve/retuve.html
License
Retuve is licensed under the Apache 2.0 License. This means that you can use it for commercial purposes, and modify it as you see fit. The only requirement is that you must provide the license with any distribution of the software.
Quickstart
Please Note that the 2D Sweep and 3D DICOMs are sythetically stitched together from 2D Graf Hips, so do not expect accurate results. We are working on getting a 3DUS Case on a Creative Commons License to have as an example.
To get started with Retuve, you can install it via pip:
pip install git+https://github.com/radoss-org/retuve.git
You can then run the following code to get a basic report:
WARNING: Before running this script, please make sure you have read the data disclaimer at DATA_DISCLAIMER.md
.
import pydicom
from radstract.data.dicom import convert_dicom_to_images
from retuve.defaults.hip_configs import default_US
from retuve.defaults.manual_seg import manual_predict_us
from retuve.testdata import Cases, download_case
from retuve.funcs import analyse_hip_2DUS
# Example usage
dcm_file, seg_file = download_case(Cases.ULTRASOUND_DICOM)
dcm = pydicom.dcmread(dcm_file)
images = convert_dicom_to_images(dcm)
hip_data, img, dev_metrics = analyse_hip_2DUS(
images[0],
keyphrase=default_US,
modes_func=manual_predict_us,
modes_func_kwargs_dict={"seg": seg_file},
)
img.save("2dus.png")
Attribution of the above Ultrasound Images: Case courtesy of Ryan Thibodeau from https://radiopaedia.org 172535 (https://radiopaedia.org/cases/172535)
Features
- pip installable (easy to intergrate with you existing systems)
- Apache 2.0 Licensed
- AI is fully pluggable/modular
- Basic Web Interface bundled (through Retuve Trak)
- CLI Interface
- Swagger API Provided (through Retuve Trak)
Examples
Examples can be found at https://github.com/radoss-org/retuve/tree/main/examples
Docs
We have separate documentation for different use cases. Please see them all here: retuve.docs.usecases
We provide high level overviews for different types of users. This includes a tailored description of Retuve, and some highlighted features:
- For Developers:
retuve.docs.overviews.developers
- For AI Researchers:
retuve.docs.overviews.ai_researchers
- For Clinicians:
retuve.docs.overviews.clinicians
Modalities
Retuve can analyse Hips for:
- Ultrasound:
retuve.docs.modalities.ultrasound
- X-Ray:
retuve.docs.modalities.xray
Developer Guide
You can clone the repository and install the dependencies with the following command:
git clone https://github.com/radoss-org/retuve.git
You can then install retuve with uv, and then run the tests:
NOTE: These tests are about testing consistency between changes, and not directly testing the accuracy of the AI. See changenotes
for tracking.
cd retuve
# Install uv if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh
# Install dependencies
uv sync
# Generate the test data
uv run poe testgen
# Run all tests, including examples.
uv run poe test_all
# Get info on all other dev scripts
uv run poe help
1# Copyright 2024 Adam McArthur 2# 3# Licensed under the Apache License, Version 2.0 (the "License"); 4# you may not use this file except in compliance with the License. 5# You may obtain a copy of the License at 6# 7# http://www.apache.org/licenses/LICENSE-2.0 8# 9# Unless required by applicable law or agreed to in writing, software 10# distributed under the License is distributed on an "AS IS" BASIS, 11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12# See the License for the specific language governing permissions and 13# limitations under the License. 14 15""" 16# The first Open, Collaborative Framework for Infant Hip Health using AI 17 18 19Retuve (from the scottish gaelic `Ri taobh` meaning `beside`) is a framework for analysing infant hips. It is designed to be a flexible and extensible framework that can be used by developers, AI researchers and clinicians. 20 21It takes in raw hip Ultrasound and X-Ray images, and outputs a report with the labelled images, and the results, exactly as a clinician would. 22 23<img src="https://raw.githubusercontent.com/radoss-org/radoss-creative-commons/main/other/224_ddh_115_%26_172535_0_diagram.jpg" alt="drawing" width="500"/> 24 25Attribution of the above Ultrasound Images: Case courtesy of Ryan Thibodeau from https://radiopaedia.org 172535 (https://radiopaedia.org/cases/172535) 26 27Attribution of the above X-Ray Images: Fraiwan, Mohammad; Al-Kofahi, Noran; Hanatleh, Omar; ibnian, ali (2022), “A dataset of DDH x-ray images”, Mendeley Data, V2, doi: 10.17632/jf3pv98m9g.2 28 29# Docs 30 31See our docs online at [https://radoss-org.github.io/retuve/retuve.html](https://radoss-org.github.io/retuve/retuve.html) 32 33# License 34 35Retuve is licensed under the Apache 2.0 License. This means that you can use it for commercial purposes, and modify it as you see fit. The only requirement is that you must provide the license with any distribution of the software. 36 37# Quickstart 38 39**Please Note that the 2D Sweep and 3D DICOMs are sythetically stitched together from 2D Graf Hips, so do not expect accurate results. We are working on getting a 3DUS Case on a Creative Commons License to have as an example.** 40 41To get started with Retuve, you can install it via pip: 42 43```bash 44pip install git+https://github.com/radoss-org/retuve.git 45``` 46 47You can then run the following code to get a basic report: 48 49**WARNING: Before running this script, please make sure you have read the data disclaimer at `DATA_DISCLAIMER.md`.** 50 51 52```python 53import pydicom 54from radstract.data.dicom import convert_dicom_to_images 55from retuve.defaults.hip_configs import default_US 56from retuve.defaults.manual_seg import manual_predict_us 57from retuve.testdata import Cases, download_case 58from retuve.funcs import analyse_hip_2DUS 59 60# Example usage 61dcm_file, seg_file = download_case(Cases.ULTRASOUND_DICOM) 62 63dcm = pydicom.dcmread(dcm_file) 64images = convert_dicom_to_images(dcm) 65 66hip_data, img, dev_metrics = analyse_hip_2DUS( 67 images[0], 68 keyphrase=default_US, 69 modes_func=manual_predict_us, 70 modes_func_kwargs_dict={"seg": seg_file}, 71) 72 73img.save("2dus.png") 74``` 75<img src="https://raw.githubusercontent.com/radoss-org/radoss-creative-commons/main/other/ultrasound/172535_0_processed.png" alt="drawing" width="500"/> 76 77Attribution of the above Ultrasound Images: Case courtesy of Ryan Thibodeau from https://radiopaedia.org 172535 (https://radiopaedia.org/cases/172535) 78 79# Features 80 81- pip installable (easy to intergrate with you existing systems) 82- Apache 2.0 Licensed 83- AI is fully pluggable/modular 84- Basic Web Interface bundled (through Retuve Trak) 85- CLI Interface 86- Swagger API Provided (through Retuve Trak) 87 88# Examples 89 90Examples can be found at https://github.com/radoss-org/retuve/tree/main/examples 91 92# Docs 93 94We have separate documentation for different use cases. Please see them all here: `retuve.docs.usecases` 95 96We provide high level overviews for different types of users. This includes a tailored description of Retuve, and some highlighted features: 97 98- For Developers: `retuve.docs.overviews.developers` 99- For AI Researchers: `retuve.docs.overviews.ai_researchers` 100- For Clinicians: `retuve.docs.overviews.clinicians` 101 102# Modalities 103 104Retuve can analyse Hips for: 105 106- Ultrasound: `retuve.docs.modalities.ultrasound` 107- X-Ray: `retuve.docs.modalities.xray` 108 109 110# Developer Guide 111 112You can clone the repository and install the dependencies with the following command: 113 114```bash 115git clone https://github.com/radoss-org/retuve.git 116``` 117 118You can then install retuve with uv, and then run the tests: 119 120**NOTE: These tests are about testing consistency between changes, and not directly testing the accuracy of the AI. See `changenotes` for tracking.** 121 122```bash 123cd retuve 124 125# Install uv if you haven't already 126curl -LsSf https://astral.sh/uv/install.sh | sh 127 128# Install dependencies 129uv sync 130 131# Generate the test data 132uv run poe testgen 133 134# Run all tests, including examples. 135uv run poe test_all 136 137# Get info on all other dev scripts 138uv run poe help 139``` 140""" 141 142from importlib.metadata import version 143 144__version__ = version("retuve") 145 146 147def _print_disclaimer(): 148 message = """ 149**Disclaimer:** 150 151This software is an experimental tool for **Research Use Only**. 152It is intended solely for research purposes related to the diagnosis of 153Developmental Dysplasia of the Hip (DDH). 154 155This software has **NOT** been approved as a medical device by any 156regulatory agency in any country. It is not intended for clinical use, 157diagnosis, or treatment. 158 159The software is provided "as is" without any warranty, express or implied. 160Use of this software is at your own risk. The developers and contributors are 161not responsible for any consequences resulting from the use or misuse of this 162software. 163 164Please consult with qualified healthcare professionals for any medical advice 165or diagnosis. 166""" 167 print(message) 168 169 170_print_disclaimer()