Відео

@@chinmayganguly457 Hi, could you post details to forum.image.sc/

@@ajaxel done, just posted. I just found out about this software. I'm trying to understand how to use U-Net for deconvolution of 3D microscopy images. is it possible?

@@chinmayganguly457 no, you can only do semantic segmentation. As I understand, deconvolution is done by applying PSF to datasets

@@ajaxel yes. But I already have deconvolved images. Can't I train U-Net to estimate the PSF from that and perform blind deconvolution?

@@chinmayganguly457 @chinmayganguly457 I would not call that deconvolution. It is rather denoising. You need to check something like CARE or noise2void for that. I unfortunately did not have time to implement that in MIB

good point, I've never tested that. I will add it to the next update, meanwhile if you are on Matlab version, you easily enable that by modifying "MIB/Classes/@mibController/mibGUI_WindowButtonDownFcn.m", where in line 323 change the last parameter from 0 to 1 as: "[w, h, z, t] = obj.mibModel.convertMouseToDataCoordinates(xy(1,1), xy(1,2), 'shown', 1);"

Hi, do you have GPU? Send me a message with description and we can check. Also, we have personal zoom meetings available, you can check calendar on mib.helsinki.fi/

yes, it should work there as well

thanks, if you can put a message on forum.image.sc with *mib* tag, I can check that

Glad you liked it! :)

Thank you, Alex! it is definitely possible, but in this case it may be better to start with some simpler segmentations as membranes. I guess the typical workflow for neuroscience is to assemble several networks starting from segmentation of membranes and continue after that to segment cells and their organelles

all details for datasets are in the supporting information at journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1008374 direct link: journals.plos.org/ploscompbiol/article/file?type=supplementary&id=info:doi/10.1371/journal.pcbi.1008374.s002

@@ajaxel Hi Ilya. I trained the images I obtained from the normal microscope with U-net and Segnet. I used 15 images for the tutorial. I could not get results after training in U-net. but I got a segmented result. Why didn't U-net produce results? I can send images and training results to your e-mail address.

I got a SegNet result

@@aslanturkhan1825 sure, I am at work this week, so we can chat about it if you want

we have not put this model anywhere. I need to check, probably we can do that

@@ajaxel where can i find cell models online to look at, doesnt have to be yours just good models?

@@Acuraintegraman1 One of small tutorial datasets is available from mib.helsinki.fi/tutorials_segmentation.html with the upper HTML link. For mitochondria in 3D you can check this: www.epfl.ch/labs/cvlab/data/data-em. If you want something on large scale you can check this one: www.nature.com/articles/s42003-021-01699-w#data-availability In general, it is possible to find those in many places across the net

Ilya Belevich thank you 🙏 this is truly amazing technology!

good point, thank you Download link: http:\\mib.helsinki.fi\downloads.html

ohh, very sorry, I missed the comment... preparation of ground truth datasets may be done in multiple ways depending on your particular data. It is possible to do that with MIB, for example, you can check tutorials from here: mib.helsinki.fi/tutorials_segmentation.html, for our data graphcut segmentation gives most bang for the buck. The other good suggestion to use machine-learning approaches, for example, image classifiers. You can check www.ilastik.org/ for a user-friendly approach.

Hi, I use createDataset function from IceImarisConnector, which creates a dataset with this command: iDataset = this.mImarisApplication.GetFactory().CreateDataSet(); iDataset.Create(classDataSet, sizeX, sizeY, sizeZ, sizeC, sizeT);

Thanks. I got it. This is the correct way to de novo create dataset. By the way, may I ask another two questions about Imaris, Thanks. It is very lucky for me to have Imaris coding expert to discuss with. (1)vApplication = vImarisApplication.GetFactory.CreateApplication; This command create a vApplication, but it seems that it is directed towards the current vImarisApplicatio instance itself. In your opinion what's the potential utility of this command? (2) vImarisApplication.FileSave('D:\QMDownload\a b\Exp.tif','writer="OmeTiff"');. This command is to save the current file as OmeTiff. There are flexible XYZCT configuration and strip/tile setting in format settings of Imaris when clicking SaveAs. I noticed that this code always follow the most recent settings. So I have to change the setting manually. However, what I really need is to use this command to save as OmeTiff with flexible settings automatically in code. Do you know if there is another way/varargin to specify these settings? I tried to search the ImarisLib.jar and found a hidden API as below. It seems that there is a Map object as third argument inside this jar file. As I'm not familiar with java, based on these information, do you think a equivalent Map object could be sent from Matlab to java as the third argument? Or is there other ways to specify it? Thanks. =========== public abstract void FileSave(String paramString1, String paramString2) throws Error; public abstract void FileSave(String paramString1, String paramString2, Map<String, String> paramMap) throws Error; public abstract AsyncResult begin_FileSave(String paramString1, String paramString2); public abstract AsyncResult begin_FileSave(String paramString1, String paramString2, Map<String, String> paramMap); public abstract AsyncResult begin_FileSave(String paramString1, String paramString2, Callback paramCallback); public abstract AsyncResult begin_FileSave(String paramString1, String paramString2, Map<String, String> paramMap, Callback paramCallback); public abstract AsyncResult begin_FileSave(String paramString1, String paramString2, Callback_IApplication_FileSave paramCallback_IApplication_FileSave); public abstract AsyncResult begin_FileSave(String paramString1, String paramString2, Map<String, String> paramMap, Callback_IApplication_FileSave paramCallback_IApplication_FileSave); public abstract void end_FileSave(AsyncResult paramAsyncResult) throws Error;

public void FileSave(String paramString1, String paramString2) throws Error { FileSave(paramString1, paramString2, null, false); } public void FileSave(String paramString1, String paramString2, Map<String, String> paramMap) throws Error { FileSave(paramString1, paramString2, paramMap, true); } The FileSave(String paramString1, String paramString2) was reloaded to FileSave(paramString1, paramString2, null, false), while FileSave(String paramString1, String paramString2, Map<String, String> paramMap) was reloaded to FileSave(paramString1, paramString2, paramMap, true). There should be a third argument. I cannot find detailed document about the third argument, and do not know how to send Map<String, String> paramMap to java. I do not know the supposed content of Map itself. Could you help me, Thanks.

I'm not sure if java.util.HashMap would work.

The following code worked!!!!!!!!! a = java.util.HashMap(); a.put('wefew','wfewe'); vImarisApplication.FileSave('D:\QMDownload\a b\Exp.tif','writer="OmeTiff"',a); This command does not throw error, means that there is a way to indicate 3rd argument! Wow! I got a breakthrough now! Then next problem is the content of the Map.