|
|
--- |
|
|
license: mit |
|
|
|
|
|
extra_gated_heading: "Accept OAI terms & created account." |
|
|
extra_gated_description: "Acknowledge that you will abide by the rules set out by the OAI and have registered here: https://nda.nih.gov/oai" |
|
|
extra_gated_fields: |
|
|
I have signed up for an OAI account and accept their terms: checkbox |
|
|
--- |
|
|
|
|
|
<table style="width: 500px;"> |
|
|
<tr style="background-color: #fec44f;"> |
|
|
<th>Resource</th> |
|
|
<th>Link</th> |
|
|
</tr> |
|
|
<tr style="background-color: #fff7bc;"> |
|
|
<td>Model</td> |
|
|
<td><a href="https://huggingface.co/aagatti/ShapeMedKnee/">ShapeMedKnee Model</a></td> |
|
|
</tr> |
|
|
<tr style="background-color: #fff7bc;"> |
|
|
<td>NSM Model Code</td> |
|
|
<td><a href="https://github.com/gattia/nsm">GitHub Repository for NSM</a></td> |
|
|
</tr> |
|
|
<tr style="background-color: #fff7bc;"> |
|
|
<td>Example Implementation</td> |
|
|
<td><a href="https://github.com/gattia/shapemedknee">GitHub Repository for ShapeMedKnee Example</a></td> |
|
|
</tr> |
|
|
<tr style="background-color: #fff7bc;"> |
|
|
<td>Paper</td> |
|
|
<td><a href="https://www.medrxiv.org/content/10.1101/2024.05.06.24306965v1">ShapeMedKnee medRxiv Paper</td> |
|
|
</tr> |
|
|
</table> |
|
|
|
|
|
|
|
|
<img src="https://cdn-uploads.huggingface.co/production/uploads/6395556bb6aa39d7cdb5db57/faIuhic4DMD7rMlAxajMX.png" alt="Description of image" style="width: 750px;"> |
|
|
|
|
|
This dataset was created to enable the development and testing of |
|
|
models of 3D anatomy. Baseline DESS knee MRIs from the Osteoarthritis |
|
|
Initiative (OAI) were autosegmented using a previously developed |
|
|
algorithm [Gatti & Maly 2021](https://doi.org/10.1007/s10334-021-00934-z). |
|
|
These segmentations were post-processed to create 3D models of the |
|
|
femur bone and cartilage using [pyMSKT](https://github.com/gattia/pymskt). |
|
|
|
|
|
Instructions on how to download the data, |
|
|
fit a neural shape model (NSM) with the data, do inference |
|
|
using a [shared model](https://huggingface.co/aagatti/ShapeMedKnee), and |
|
|
perform tests are provided at: https://github.com/gattia/shapemedknee. |
|
|
|
|
|
Check back soon for the associated publication. |
|
|
|
|
|
|
|
|
The data are structured as: |
|
|
``` |
|
|
/meshes |
|
|
/train |
|
|
/subfolder_X |
|
|
subjectid_LEG_fem_cart.vtk |
|
|
subjectid_LEG_femur.vtk |
|
|
... |
|
|
/val |
|
|
/subfolder_X |
|
|
subjectid_LEG_fem_cart.vtk |
|
|
subjectid_LEG_femur.vtk |
|
|
... |
|
|
/test |
|
|
/subfolder_X |
|
|
subjectid_LEG_fem_cart.vtk |
|
|
subjectid_LEG_femur.vtk |
|
|
... |
|
|
/segs |
|
|
subjectid_LEG-labl.nii.gz |
|
|
prediction_dataset.csv |
|
|
``` |
|
|
|
|
|
The `/meshes` & `/segs` folders include the data used to train and test |
|
|
surface mesch reconstructions. The prediction_dataset.csv includes |
|
|
information needed for clinical prediction tasks (e.g., osteoarthritis |
|
|
grading, future knee replacement prediction). |
|
|
|
