matnwb: DynamicTable error in PlaneSegmentation object

[Icardij]

Hello,

I am very new to working with the NWB data format, and am currently having problems using matnwb. The long and short of it is when I run my script to create, and write the NWB file, it prints out the below error.

I am not entirely sure on what a dynamic table is, though I do understand that the script seems to be crashing as it tries to create a PlaneSegmentation object that is later used to create a dynamic table. (crash location posted below)

This code is a near exact copy of the Ophys tutorial base, except without the behavior section since we collect data from non-behaving subjects.
I went ahead and tried to use a exact copy of the Ophys tutorial (without behavior) and the same error appeared. If anyone has a idea of why this error is occurring, whether it's a incorrect argument in my script, a error in the install, or something else entirely I would appreciate the help.

Thanks

[oruebel]

I am not entirely sure on what a dynamic table ...

NWB uses tables in a number of places to store metadata. DynamicTable is the base type that NWB uses to store tables. In a DynamicTable the data is stored in a column-based fashion, i.e., each column of the table is stored as a separate dataset/array. All DynamicTable objects have an id column and can otherwise contain an arbitrary number of columns and rows. The tables are called "Dynamic" because they allow addition of custom columns to include user-defined data without having to create an extension to the NWB data standard schema.

The PlaneSegmention type inherits from DynamicTable, i.e., PlaneSegmention is itself a table that defines a number of additional predefined columns.

If anyone has a idea of why this error is occurring, ...

I'm not an expert in MatNWB, but based on the error message it appears that there is a mismatch between the id column and the rest of the columns in the PlaneSegmention table. Since PlaneSegmention is a table, all columns are required to have the same length. Based on the error, it seems that your id has length 1 while other columns are length 0. Is it possible that maybe your value for RG is wrong when you create the ElementIdentifiers for the table or maybe (since the table appears to be empty) you are missing some calls to add your data to the table?

CC @bendichter

[bendichter]

@Icardij could you copy/paste the entire command?

[Icardij]

Sure thing, There are a few variables that we define in our data, such as Im_both which is a 4D matrix of our raw images. (xpixel X ypixel X depth X channel) We also have image_mask, which is our segmented ROIs organized by (frame X cellnumber X pixellist) and is shared across both channels, and finally we have RG which is a ratio calculated between the values of the two channels. Not sure if any of these would cause problems.

nwb = NwbFile( ...
    'session_description', 'mouse brain imaging',...        
    'identifier', 'hSyn-AAV2/1-Addgene', ...                
    'session_start_time', datetime(2022, 1, 27), ...        
    'general_experimenter', 'Dan M', ...                       
    'general_institution', 'Lerner Research Institute', ...  
    'general_related_publications', 'TBD');                 

subject = types.core.Subject( ...
    'subject_id', 'D20', ...           
    'age', 'P90D', ...                 
    'description', 'AAV2/1 mouse', ...  
    'species', 'Mus musculus', ...      
    'sex', 'M' ...                      
);
nwb.general_subject = subject;

optical_channel = types.core.OpticalChannel( ...
    'description', 'two channels green/red', ...  
    'emission_lambda', 550);   

device = types.core.Device();
nwb.general_devices.set('Device', device); 

imaging_plane_name = 'imaging_plane';
imaging_plane = types.core.ImagingPlane( ...
    'optical_channel', optical_channel, ...  
    'description', 'Photoconversion of Neurons', ...  
    'device', types.untyped.SoftLink(device), ... 
    'excitation_lambda', 1100, ...  
    'imaging_rate', 0, ...  
    'indicator', 'CaMPARI', ...  
    'location', 'Visual cortex');  
nwb.general_optophysiology.set(imaging_plane_name, imaging_plane);

image_series = types.core.TwoPhotonSeries( ...
    'imaging_plane', types.untyped.SoftLink(imaging_plane), ... 
    'starting_time', 0.0, ... 
    'starting_time_rate', 3.0, ...
    'data', Im_both, ...                       
    'data_unit', 'unit16',...
    'description', 'Frames X Depth X Channel');
nwb.acquisition.set('TwoPhotonSeries', image_series);

plane_segmentation = types.core.PlaneSegmentation( ... 
    'colnames', {'image_mask'}, ...  
    'description', 'Cells segmented from visual cortex, Frames X cellnumber X pixellist', ...  
    'image_mask', types.hdmf_common.VectorData('data', image_mask, 'description', 'image masks'), ...
    'id', types.hdmf_common.ElementIdentifiers('data', int64(1:2000)), ...
    'imaging_plane', types.untyped.SoftLink(imaging_plane)); 
% plane_segmentation.image_mask = types.hdmf_common.VectorData( ...
%     'data', image_mask, 'description', 'image masks');                                          

img_seg = types.core.ImageSegmentation();
img_seg.planesegmentation.set('PlaneSegmentation', plane_segmentation);

ophys_module = types.core.ProcessingModule( ...
    'description',  'contains optical physiology data'); 
ophys_module.nwbdatainterface.set('ImageSegmentation', img_seg);
nwb.processing.set('ophys', ophys_module);

% Fluorescence and dF/F responces
roi_table_region = types.hdmf_common.DynamicTableRegion( ...
    'table', types.untyped.ObjectView(plane_segmentation), ...
    'description', 'Red to Green ratio from all ROIs', ...
    'data', RG);
roi_response_series = types.core.RoiResponseSeries( ...
    'rois', roi_table_region, ...
    'data', NaN(10, 100), ...
    'data_unit', 'Not Relevant', ...
    'description', 'Not Relevant',...
    'starting_time_rate', 3.0, ...
    'starting_time', 0.0);

fluorescence = types.core.Fluorescence();
fluorescence.roiresponseseries.set('RoiResponseSeries', roi_response_series);
ophys_module.nwbdatainterface.set('Fluorescence', fluorescence);

nwb.processing.set('ophys', ophys_module);

nwbExport(nwb, 'test.nwb');

[bendichter]

An "image_mask" should be a matrix that is n_rois * x * y. Could you reshape your data into that and try that? We also have the option of a "pixel mask" which is similar to what you are describing.

[Icardij]

I have changed the "image_mask" variable to what you said, as well as changing the 'data' argument in "roi_table_region" and "roi_response_series" so that the code now looks like below.

roi_table_region = types.hdmf_common.DynamicTableRegion( ...
    'table', types.untyped.ObjectView(plane_segmentation), ...
    'description', 'all ROIs', ...
    'data', (1:length(RG))');
roi_response_series = types.core.RoiResponseSeries( ...
    'rois', roi_table_region, ...
    'data', RG', ...
    'data_unit', 'red to green ratio', ...
    'description', 'red to green ratio from same cell',...
    'starting_time_rate', 0.0, ...
    'starting_time', 0.0);

The script seems to run in its entirety now without kicking back any errors, and I wanted to check that this looks appropriate before I go ahead with actually upload our data.

[bendichter]

You should use 0-indexing for the types.hdmf_common.DynamicTableRegion: 'data', (0:length(RG)-1)'