Added first working FMCIB model container #73
Conversation
models/fmcib/utils/FMCIBRunner.py
Outdated
| with torch.no_grad(): | ||
| features = model(image) | ||
|
|
||
| feature_dict = {f"feature_{idx}": feature for idx, feature in enumerate(features.flatten().tolist())} |
There was a problem hiding this comment.
Our logical data wrappers could be a good fit here.
Although we can (and should) return the original output, this, besides adding verbosity, allows us to bring the output into every format using e.g. our ReportExporter right from the config.
| ARG MHUB_MODELS_REPO | ||
| # Add pull models repo command here after local testingRUN | ||
| RUN buildutils/import_mhub_model.sh fmcib_radiomics ${MHUB_MODELS_REPO} | ||
| RUN wget https://zenodo.org/records/10528450/files/model_weights.torch?download=1 -O /app/model_weights.torch |
There was a problem hiding this comment.
Maybe we choose a distinct file name or place the weights inside a folder to keep us the option to add then later successive models and have all their weights organized. Let's spend a minute thinking about this.
| "contrast": true | ||
| }, | ||
| { | ||
| "label": "Center of mass", |
There was a problem hiding this comment.
This doesn't match with the current default workflow but we can address this later and decide which of the two (json / mask) we want to keep as the default. I'd choose whatever is the easiest and most standard one.
There was a problem hiding this comment.
I provided an alternative workflow that can start from the JSON file. I've some ideas on how we could provide a workflow starting from Dicom (which should be the default then) that I want to discuss later!
|
I tested the default workflow, made some small fixes and it all looks good. Also tested the slicer workflow on a folder structure that looks like this Using the For the testing, |
There was a problem hiding this comment.
This is here for now just to show you what I did with the slicer stuff, we can just restructure this later as you suggest
|
Hmm, I'm not too sure how you'd want the format to be for these on the release (since we don't have any DICOM + json workflows yet and I am not too sure how to do that myself without spending a lot of time 😓 ) So I've attached both the features.json and the centroids.json for the IDC DICOM scan that I linked above. Maybe you have an idea of how to do this best? |
|
That's great! I'll implement the dicom workflow later today! |
|
@LennyN95 Here is another sample. Image: |
|
/test sample:
idc_version: 17.0
data:
- SeriesInstanceUID: 1.3.6.1.4.1.32722.99.99.298991776521342375010861296712563382046
aws_url: s3://idc-open-data-cr/40db0070-9773-413a-8d7e-c91c37c0006b/*
path: dicom
- url: https://github.com/MHubAI/models/files/14520712/centroids.json
path: 1.3.6.1.4.1.32722.99.99.298991776521342375010861296712563382046.json
- SeriesInstanceUID: 1.3.6.1.4.1.32722.99.99.309695689942006699082558764031786785731
aws_url: s3://idc-open-data-cr/18b83380-0f27-42f2-af7c-65e2d23668ac/*
path: dicom
- url: https://github.com/MHubAI/models/files/14528031/LUNG1-076_centroids.json
path: 1.3.6.1.4.1.32722.99.99.309695689942006699082558764031786785731.json
reference:
url: https://github.com/MHubAI/models/files/14612917/fmcib_mhub_default_output.zipTest Results (24.03.15_11.57.01_d91gDb1QPt)id: dbce25f8-eaa8-411f-a4ec-d56e6e9f7df6
date: '2024-03-15 11:57:56'
checked_files:
- file: features.json
path: /app/test/src/1.3.6.1.4.1.32722.99.99.309695689942006699082558764031786785731/features.json
checks:
- checker: DataFileCheck
notes:
- label: Value Match
description: These keys have identical values
info: feature_0,feature_1,feature_2,feature_3,feature_4,feature_5,feature_6,feature_7,feature_8,feature_9,feature_10,feature_11,feature_12,feature_13,feature_14,feature_15,feature_16,feature_17,feature_18,feature_19
(+ 4076 more)
- checker: SizeCheck
- file: features.json
path: /app/test/src/1.3.6.1.4.1.32722.99.99.298991776521342375010861296712563382046/features.json
checks:
- checker: DataFileCheck
notes:
- label: Value Match
description: These keys have identical values
info: feature_0,feature_1,feature_2,feature_3,feature_4,feature_5,feature_6,feature_7,feature_8,feature_9,feature_10,feature_11,feature_12,feature_13,feature_14,feature_15,feature_16,feature_17,feature_18,feature_19
(+ 4076 more)
- checker: SizeCheck
summary:
files_missing: 0
files_extra: 0
checks:
DataFileCheck:
files: 2
SizeCheck:
files: 2
conclusion: true |
github-actions
bot
added
INVALID TEST REQUEST
No description provided.