It should work to support drop in vscode, since it works for the niivue instance inside the webview

• Reduce required performance
• Switching the underlying surface
• Save/Load view settings

Loading multiple large images in compare view can take several seconds.
Displaying them one at a time might make the wait time more bearable and transparent.

The amount of time is roughly the same for .nii and .nii.gz (with 15% of the size). Therefore, the time for loading is probably not vscode related but the data loading / preparation in NiiVue. <- Need to investigate this more closely to find out if a speedup is possible

After loading 8 images (224x224x208), the last 7 images were loading very slowly. (niivue-compare)

Functionalty

• Open remote files
• click on links to images (remote and local)
• Open clicked web urls for supported image types (currently web urls are sent to the browser) (solved with hover)
• Mosaic view when selecting multiple files
• File Picker to select files
• DICOM folders
• Not reading the full file for 4D images
• Overlays

UI

• Add decorators (highlights) for links of images
• Add brain-icon to supported file types
• Add NiiVue icon for extension
• Buttons
• Box to enter Link (via command Niivue: Open Image)
• Show NIfTI info
• Show scale
• Change color scale

This is just brainstorming, some features might be simple, some impossible.

An X button in the corner to remove an image should be easy to add

A drag-bar could be at the top, which when clicked would move the image.
The insertion logic at a new place could be swap or insert. Probably insert is more intuitive

It makes probably sense to adapt the license to match the other repositories in the niivue organization

Show a nice layout when no image is loaded.

Maybe instructions about Niivue-ify and other projects

See discussion in:
niivue/niivue#692

maybe this is #32

We could create a command to send code from an open editor to the niivue extension.
For example files ending in .niivue could execute in the extension, producing a niivue output.

Currently debugging the webview layout is quite challenging in the vscode rendering. It doesn't give access to console or debug view.

Making the same html with the main.js file run in plain html (with mockup data or drag'n drop) would make it easier to debug/develop

currently the webpanel shortly freezes after typing in the first number

It would be possible with virtual webgl, see:
niivue/niivue#693

https://vite-pwa-org.netlify.app/assets-generator/

All images would be rendered above each other to switch between the images with identical pixel locations

Currently the matrix size and voxelsize is only displayed once. This might be confusing if mutliple images are loaded

It seems like it could work in the change viewer of vscode source control, but it only sometimes renders.

Currently, there are different modes. For the compareView, a NvArray is used and for the normal view a nv object.
Better combining both modes and using the same display and resize logic (with special case if it's only one nv in the array) would make the functionality for the different views more consistent.

Only applying the scale on click would solve two issues

• loading additional images and reapplying the same scale
• avoid long waiting time for many large images between scaling min and max

The web app needs a button to reset (F5 works currently)

Would be great to be able to add overlays, that's it.

This extension is awesome! Keep up the great work :)

Feautre request: Display metadata (dimensions, data resolution, value)

Hi, thanks for this effort on niivue, it's amazing. It's especially useful when working in remote servers, to visualize images without downloading them locally.

However: I use vscode-remote, images are opened quite well in niivue, but to add an overlay I first have to download it because when I click add overlay, it opens a file browser to open a local file. I would like to be able to open files from vscode file browser (maybe via drag and drop?).

• change origin
• change voxel size
• zoom
• set scaling like on console (scale applies to selected images)

Simon:

• some images don't load, for reasons that don't have anything to do with the size, but maybe due to eccentric/broken headeres (in the example I sent you, if I multiply Julia_t2star_centred.nii by 1 with fslmaths it works)?
• currently only seems to view in triplanar in the stand-alone implementation - (axial/sag/cor option doesn't seem to work)
• maybe make a clearer space to drop new images (e.g. when the display is full)?
• middle button image shift is nice - could there be a reset (like "home zoom/pan" in Siemens Syngo)?
• what about adding blue box highlight round 'active' images (the ones current operations e.g. windowing/zoom are applied to)? ctr-a to highlight all?
• for multi-echo/multi- time points, could you add the option to pick an echo/tp?
• will it be possible to write out images as pngs, etc?
• Linux thing 1: what would the bash command be to open images?

The drop area doesn't fill the whole window but only where the box is

results/rsos we have

1. ICE_t2star.nii
2. Julia_t2star_centred.nii
3. Julia_t2star.nii

Only 3) loaded. These might be good images for debugging the load issue. (FYI I notice that when I multiply Julia_t2star_centred.nii by 1 with fslmaths it works - so maybe an excentric/broken header?)

One suggestion is to embed this information on each image canvas. Currently, it is hard to tell from which loaded image that information comes from.

The overlays of different meshes seem to share the same ranges

create a view button, that when clicked disables the over-volume menus

When building tests to check whether surfaces were appropriately loaded, we noticed that it takes a little while to detect them. Currently, we have to load two surfaces to change the state of whether a surface was loaded and then be able to test for the loading by checking the output about the number of points/nodes as in:

${isMesh.value? getNumberOfPoints(nvArraySelected.value[0]): 'No mesh'}

The same problem happens with the surface overlay.

The name of the web app changed to "niivue-vscode web app"

The extension consists of two parts, the vs code part (simple logic) and the webview app (most logic)

1. Test communication: vs code <-> app
2. Test app logic. Are the correct elements shown for a given task (probably use Jest, maybe RTL/enzyme in addition)

cal_min and cal_max are reset after colormap change
Maybe this is an issue of niivue
=> investigate and possibly create issue on niivue

This is used in the demo:
alert(nv1.volumes[0].hdr.toFormattedString());