Unable to save annotation ROI

[winston52]

Hi Authors,

Thank you for sharing this tool. I have successfully built Slim on my local machine by following your instructions using the Docker Compose configuration.

I was able to store a DICOM file using your command:

dicomweb_client -vv --url http://localhost:8008/dcm4chee-arc/aets/DCM4CHEE/rs store instances -h path_to_dicom

Then I made some annotations in my browser, but I don't know how to retrieve these annotations. Could you please advise me on how to get these annotations?

Could I get the annotation files directly from the browser's console? or use the dicomweb_client command to get the annotation files?

Thank you for your help.

[winston52]

Hi Authors,

Here are some additional information I'd like to provide.

I reviewed the repo and the paper again, and understand the annotation file is saved in dicom format. Could you please provide some guidence on how to get the annotation files?

Here is what I did:

Step1: upload a sample based on your provided dicomweb_client command
Step2: click the ROI annotation button (ALT + D)
Step3: after completing the annotation, click the save button (ALT + S), and get the ROI_UID from browser
Step4: try to get the annotation file through URL address of the DICOMweb server

I tried to get the annotation file from this path:

http://localhost:8008/dcm4chee-arc/aets/DCM4CHEE/rs/studies/study_UID/series/series_UID/instances/ROI_UID

But i got this error

{"errorMessage":"No matches found."}

if I use this address http://localhost:8008/dcm4chee-arc/aets/DCM4CHEE/rs/studies/study_UID/series/series_UID

to get the series file, it works

also use this address : http://localhost:8008/dcm4chee-arc/aets/DCM4CHEE/rs/studies/study_UID

to get the studies file, it also works.

Could you share some guidence on how to get the annotation dicom files?

Thank you very much for your assitance.

[fedorov]

Have you tried listing all series available in the study after creating the ROI, and see if there is a new series for the annotation?

The UID you see in the UI might not be the UID of the series.

[winston52]

Thank you for your reply. I followed your suggestions and tried to list all series based on the following command.

I used the command: dicomweb_client --url http://localhost:8008/dcm4chee-arc/aets/DCM4CHEE/rs search series --study study_UID to get all the series under this study_UID

and then used the command: dicomweb_client --url http://localhost:8008/dcm4chee-arc/aets/DCM4CHEE/rs search instances --study study_UID --series series_UID to get all the instances under this series(whole slide image).

But there are only 5 instances (5 dicom files under different resolution), which means there is no annotation files saved. I suspected the annotation files was not saved successuflly.

Then I opened the console of the browser, and when I tried to click the "save" button, it will get an error, and I attach the screenshot of the error infomation here:

It shows: no user information available, and error for reading 'SpecimenUID'.

I am wondering if there is any part missing in the metadata header of my dicom files.

Thank you very much for your quick reply and patience.

[fedorov]

Interesting. I have to say, I don't know. This would require debugging.

Unfortunately, our resources for maintaining slim are very limited, and at the moment we focus on support of visualization features in slim, and not annotation generation. I am not sure when we might look into this. But if you debug and find an issue - you are welcome to submit a fix!

[winston52]

Thank you for your clarification, I will try it.

[fedorov]

@igoroctaviano @pieper I had some further communication with @winston52 and was able to reproduce the error saving annotation created in Slim UI.

Do you have any idea why this error in dcmjs is triggered?

https://github.com/dcmjs-org/dcmjs/blob/158c396f1852399ad50f2bb7fe7bc44e8db34baa/src/sr/templates.js#L904-L908

trackingIdentifier is invoked here and has 2 parameters:

slim/src/components/SlideViewer.tsx

Lines 2389 to 2392 in 808a62e

	trackingIdentifier: new dcmjs.sr.templates.TrackingIdentifier({
	uid: roi.properties.trackingUID ?? roi.uid,
	identifier: `ROI #${i + 1}`
	}),

[fedorov]

@winston52 as pointed out by @pieper, in the snippet below, the parameter passed to the constructor is a structure, not array. Do you have any experience with JS/TypeScript? Can you experiment to investigate/debug this issue and work on a fix?

slim/src/components/SlideViewer.tsx

Lines 2389 to 2392 in 808a62e

	trackingIdentifier: new dcmjs.sr.templates.TrackingIdentifier({
	uid: roi.properties.trackingUID ?? roi.uid,
	identifier: `ROI #${i + 1}`
	}),

[winston52]

Thank you very much for your help @fedorov @pieper.

I am not familiar with JS, so I only set some breakpoints in the console of the browser to track whether the annotation is captured by the slim. Specially I draw a ROI on the same public CPTAC-BRCA dicom file and then click the save button.

It seems the code from the snippet section is correct because it successfully initializes the trackingIdentifier. I've attached a screenshot here for your reference, showing that the drawn annotation is indeed captured. However, when I continue running the code step by step, it it jumps from line 2407 back to line 2388, triggering the same trackingIdentifier error as before. In other words, it works correctly for the first iteration but then automatically jumps back to re-initialize the tracking identifier, causing the error.

Could you please check this issue and perhaps provide further guidance on debugging?

slim/src/components/SlideViewer.tsx

Lines 2387 to 2410 in 808a62e

	findingType = findingType as dcmjs.sr.valueTypes.CodeContentItem
	const group = new dcmjs.sr.templates.PlanarROIMeasurementsAndQualitativeEvaluations({
	trackingIdentifier: new dcmjs.sr.templates.TrackingIdentifier({
	uid: roi.properties.trackingUID ?? roi.uid,
	identifier: `ROI #${i + 1}`
	}),
	referencedRegion: new dcmjs.sr.contentItems.ImageRegion3D({
	graphicType: roi.scoord3d.graphicType,
	graphicData: roi.scoord3d.graphicData,
	frameOfReferenceUID: roi.scoord3d.frameOfReferenceUID
	}),
	findingType: new dcmjs.sr.coding.CodedConcept({
	value: findingType.ConceptCodeSequence[0].CodeValue,
	schemeDesignator:
	findingType.ConceptCodeSequence[0].CodingSchemeDesignator,
	meaning: findingType.ConceptCodeSequence[0].CodeMeaning
	}),
	qualitativeEvaluations: roi.evaluations.filter(
	(item: dcmjs.sr.valueTypes.ContentItem) => {
	return item.ConceptNameCodeSequence[0].CodeValue !== '121071'
	}
	),
	measurements: roi.measurements
	})

Additionally, I noticed there's code filtering the ROI by the CodeValue at line 2406:

CodeValue !== '121071'

However, in two other parts (lines 1608 and 2375 in the slideViewer.tsx file), it checks the condition using:

CodeValue === '121071'

Could you please clarify why there is this difference?

Thank you again for your quick response and assiatance.

[pieper]

I'm afraid you are digging into some code that neither @fedorov nor I have much familiarity with. Our funded use case for Slim doesn't include making annotations and the original author of that code is no longer available.

The code jumping around in the debugger does sound very odd, maybe due to some async behavior or something. I do try to keep dcmjs code in good shape for IDC and other projects so if you find issues with that code it's easier to help with (and easier to test since it's not event driven).

[winston52]

Thank you for the clarification. That makes sense. I’ll take a closer look at the dcmjs part and see whether the issue might be isolated there.

[fedorov]

@winston52 we discussed this issue briefly with @igoroctaviano today.

We believe the underlying issue is still in this condition that fails:

https://github.com/dcmjs-org/dcmjs/blob/158c396f1852399ad50f2bb7fe7bc44e8db34baa/src/sr/templates.js#L904-L908

        if (options.trackingIdentifier.length !== 2) {
            throw new Error(
                "Option 'trackingIdentifier' must include a human readable tracking " +
                    "identifier and a tracking unique identifier."
            );
        }

The problem is not that it is not initialized, but that the length is not 2, and we believe it is not 2 because TrackingIdentifier extends Template, and there are some variables inherited from the parent that make length larger than 2. Maybe you can debug that?

[fedorov]

@winston52 in fact you can confirm the above by doing this in the dev console, which will yield 3.

const test = new dcmjs.sr.templates.TrackingIdentifier({
  uid: roi.properties.trackingUID ?? roi.uid,
  identifier: `ROI #${i + 1}`
});
console.log(test.length);

I am not a JS developer, so I am just learning some of those tricks....