Feature request: Image reader: read multiple series

Could you please allow the Image Reader Node to read multiple series at once, just like all the other dimensions (xyzct)?

Thanks!

Hi Lorenz,

nice idea: I wrote it down on our list: https://github.com/knime-ip/knip/issues/41.

Cheers,

Christian

 

 

 

One more thing about the image reader: it would be very helpful if the image cells that it generates would remember which was the Image Subset Selection. Currently, it's always a bit tricky to get that part of the metadata back into the workflow. My best solution for now is to modify the image Name and to append all the information needed there.

Somehow I would expect the "Offset" Property of  the image should help here. But so far it's implemented to just give an offset of the dimensions that the image still has, e.g. an XY image has the offsets [0,0], but it is also a Subset from zct [7,4,10].

Later in my workflow, this missing feature translates to the fact that my Image Segments all think their "Centroid T" was 0.

Hi Lorenz,

is there a reason, that you split of the images into individual time-points? Because if you would have a real "video", the Image Segment will be able to return the true "Centroid T".

Additionally, I don't understand how this relates to the Image Reader "Cropping Property", or are you sequentially reading the images in, T by T?

Cheers,

Christian

 

Maybe I am doing something wrong with the Labelings. The reason why I am splitting the image and labelings is that I want to get a measure, e.g. area, for each individual timepoint.

The "Cropping Property" would help to get the timepoints / channels back in the table as an own column. I guess the Centroid T/C/... could most often do the job - if they are not 0.

The Image Segment Feature node calculates features per label. So if you have an object with the same label over several dimensions (e.g. 3d object in Z Stack) it will calculate the properties (e.g. mean) for the complete object. If you only have objects which lie in a plane, you will calculate it for those 2d objects and so on. So you can directly extract the features you need from a n-dimensional object. No need to split them if you already represent the objects in two dimensions.

Anyway, if you have a 3D Segmentation and you want to calculate the features plane-wise, you can follow the "Slice Loop Start" example workflow on our page.

Hope this helps,

Christian