wiheto / netplotbrain Goto Github PK

Two small improvements needed.

• Tests sometimes fail after small changes (currently after nodealpha was implemented baseline figures and test figures do not always match). Why this is the case is unknown at the moment. Perhaps some different matplotlib version weirdness. Needs to be investigated. This seems to include titles on figure sometimes becoming too close to the figure in one version.
• One-panel figures with node property legends (see the tutorial in the documentation) have overlapping text/figures despite tight_layout on. Not sure why this is happening.

Both above issues should be fixed.

Hi!

I'm trying to plot multiple plots in a figure, but I encounted an error when I try to pass the parameter ax. Here below is a example:

import netplotbrain
import pandas as pd
import matplotlib.pyplot as plt

import netplotbrain
import pandas as pd

# Define the nodes
nodes_df = pd.DataFrame({'x': [40, 10, 30, -15, -25], 
                         'y': [50, 40, -10, -20, 20], 
                         'z': [20, 30, -10, -15, 30]})
# Call netplotbrain to plot

fig, ax = plt.subplots(1,1,figsize=(5,5))
netplotbrain.plot(ax = ax, nodes=nodes_df, arrowaxis=None)

The error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
File [d:\Softwares\Mambaforge\envs\gl\Lib\site-packages\netplotbrain\plot.py:200](file:///D:/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py:200), in plot(nodes, fig, ax, view, edge_weights, frames, edges, template, network, edge_color, node_size, node_color, node_type, hemisphere, highlight_nodes, highlight_edges, **kwargs)
    [198](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=197) affine = None
    [199](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=198) if template is not None and viewtype[fi]=='b':
--> [200](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=199)     affine = _plot_template(ax, template_style_frame, template,
    [201](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=200)                             hemisphere=hemi_frame,
    [202](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=201)                             azim=azim[fi], elev=elev[fi],
    [203](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=202)                             **profile)
    [205](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=204) # Template voxels will have origin at 0,0,0
    [206](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=205) # It is easier to scale the nodes from the image affine
    [207](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=206) # Then to rescale the ax.voxels function
    [208](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=207) # So if affine is not None, nodes get scaled in relation to origin and voxelsize,
    [209](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=208) # If node coords are derived from nodeimg, this has already been taken care of.
    [210](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/netplotbrain/plot.py?line=209) if nodes is not None and nodeimg is None and viewtype[fi]=='b' and scaled_nodes == False:
...
   [1448](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/matplotlib/__init__.py?line=1447)     bound = new_sig.bind(ax, *args, **kwargs)
   [1449](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/matplotlib/__init__.py?line=1448)     auto_label = (bound.arguments.get(label_namer)
   [1450](file:///d%3A/Softwares/Mambaforge/envs/gl/Lib/site-packages/matplotlib/__init__.py?line=1449)                   or bound.kwargs.get(label_namer))

TypeError: Axes.scatter() got multiple values for argument 's'

I'm not sure if I am doing it wrong. Thanks a lot!

For the templateflow, some templatestyles (surface) can be ready made and packaged to save on time.

An additional flag called can be added to use cached templates

Hi,

I would like to ask if this is possible to modify the edge_color input to get the list or NumPy array that represents each edge color?
I think these is only two lines of modifications where:
if isinstance(edgecolor, np.ndarray): if edgecolor.shape[0] == len(edges): ec = edgecolor[eci, :]

can be changed to:
if isinstance(edgecolor, np.ndarray) or isinstance(edgecolor, list): if len(edgecolor) == len(edges): ec = edgecolor[eci]

When I am using the current code and I want to pass the NumPy array with the colors for each edge then I have an error:
ValueError: array(['#6c1fff', '#68fcc1', '#1dd1e2', '#149df1', '#08bee9', '#2adddd', '#4a53fb', '#08bee9', '#4a53fb', '#622ffe', '#149df1', '#622ffe', '#6c1fff', '#68fcc1', '#1dd1e2', '#2adddd'], dtype='<U7') is not a valid value for color

I would like to have the color for each edge based on the connectivity values.

Best,
Marta

Subtitles still remain when setting title to None.

The following code results with a subtitle superior.

import netplotbrain
netplotbrain.plot(template='MNI152NLin6Asym', view='S', title=None)

This is not the expected behaivour as title=None should suppress the automatic subtitle behaviour (as discussed previously). @silviafan since it was you who added the subtitle behaivour, can you look into this?

Current state: At the moment, you can specify a view as a multi letter argument, e.g. view='LR' and frames=5. This will produce 5 frames going from L to R.

Problem: Currently it is not possible to specify 360 degree rotation without some subpotimal hacking of arguments (which may even be broken).

Desired implementation: view='L', frames=5, will plot 5 figures going from L view around and excluding L view.

Potential problems Rotation can be along the rotation axis xy, xz or yz axes. I think 360 arguments only makes sense for L, R, A, and P along the x.y axis. Maybe the arguments has to be "L360", "R360", "A360", "P360"

Difficulty: maybe not the easiest to do. But relatively easy. Good first issue for someone with programming experience wanting to learn more about the package.

I would like to ask whether the coordinates XYZ in the node represent the index of the image, similar to IJK, or represent the world coordinate system, such as the RAS coordinate system or the LPS coordinate system. The situation is not explained in the document

Some string argument to views that specify a default multi panel layout.

For example: view=”sides”

may great a 2x2 figure where you see:

row1: L, R

row 2: R (left hemisphere only), L (right hemisphere only)

motivation: otherwise to plot this type of figure you need to type

views=[[“L” ,”R”], [“L”, “R”]]

hemi=[[“” ,””], [“R”, “L”]]

A selection of different default view combinations can then be created (Eg 6 and 8 panel figures) with a single string.

difficulty to implement: easy to moderate.

Observation: Saving a 3-panel svg results in large files being created (60mb).

What is needed:

• dpi kwarg that is included when figure is set up
• all templates get rasterized automatically or allow option for it (need to check how this works for 3D axis)

Check template compatibility with all atlases. If not, make json for special settings for each

• glass
• cloudy
• surface
• filled

Note: heavily edited comment

Not sure when this happened or why but the previous code is no longer run. Should be fixed asap

Instead of default strings that map to certain arguments, allow for user to supply an entire dictionary.

Difficulty: Easy, (if templateflow.api.get() is known)

• Direction of rotation RP+ or RP- should go different ways around the brain.
• Add lists of directions ['AP+', 'LR-']

Based on the templatestyle the number of padded zeros around images changes, which changes figure size.

Possible solution: if MNI is provided, rescale XYZ to the size of the input MNI at the end of netplotbrain.plot.

The following example

atlasinfo = tf.get(template='MNI152NLin2009cAsym',
       atlas='Schaefer2018',
       desc='400Parcels7Networks',
       extension='.tsv')
atlasinfo = pd.read_csv(atlasinfo, sep='\t')
networks = list(map(lambda x: x.split('_')[2], atlasinfo.name.values))
atlasinfo['yeo7networks'] = networks

netplotbrain.plot(nodes=atlasinfo,
                  nodeimg={'atlas': 'Schaefer2018',
                           'desc': '400Parcels7Networks',
                           'resolution': 1},
                  template='MNI152NLin2009cAsym',
                  templatestyle=None,
                  view=['LSA'],
                  nodetype='parcels',
                  nodecmap='tab20b',
                  nodecolorby='yeo7networks',
                  title='',
                  arrowaxis=[],
                  nodealpha = 0.8,
                  highlightnodes={'yeo7networks': 'Cont'},
                  highlightlevel=0.95,
                  nodecolorlegend=True,
                  nodesizelegend=False)

results in the following figure and cluttered legend

Have option to only plot in single hemisphere.

Following code

import netplotbrain 
# Get 100 node atlas from templateflow
node_dict = {'atlas':       'Schaefer2018', 
             'desc':        '100Parcels7Networks',
             'extension':   '.nii.gz',
             'resolution':  1}
netplotbrain.plot(template = 'MNI152NLin2009cAsym',
                  nodes =     node_dict,
                  nodetype = 'parcels',
                  view =      'S',
                  hemisphere = 'L')

Leads to

RuntimeError: No surface found at the given iso value.

Problem seems to be that some voxels that some voxels for R hemisphere nodes are still included and too small to plot as parcel surfaces.

Solution: Need to add checks about parcel center's hemisphere when getting single hemisphere. OR different non-surface method to plot parcels (just voxels takes too long)

Difficulty to solve: medium to hard

Problem with set_rasterized and alpha interacting. Works when template_style='surface' but not with glass.

Gifs can only go 180 degrees at the moment. Should allow 360 through one argument

Add bctpy's NBS implementation as possible edge input that interacts with highlightnodes.

• codacy
• readthedocs pdocs
• coveralls
• travis circleci
• zenodo

From showcase_image.py:

netplotbrain.plot(template='MNI152NLin2009cAsym',
                  templatestyle='surface',
                  view='L',
                  nodes='./examples/example_nodes.tsv',
                  hemisphere=['R'],
                  nodescale=5,
                  edges='./examples/example_edges.tsv',
                  templatealpha=0.05,
                  title='Highlight results',
                  templatecolor='gray',
                  highlightlevel=0.9,
                  highlightnodes=[6, 8, 25, 29],
                  fig=fig, ax=ax,
                  nodecolorlegend=False,
                  nodesizelegend=False,)

In general, this happens when nodetype is 'circles' or 'spheres'.

• fix subplot title. Can intersect with figure when no template argument is given
• Legend font often too small
• Legend title often intersecting with figure,
• showcase image title too large
• fix documentation and examples after subtitle/title change
• if title is list, make it none and subtitles=title (will also fix old behaviour and make documentation correct)

Following code

import netplotbrain
import pandas as pd

edges = pd.read_csv('./examples/example_edges.tsv', sep='\t', index_col=0)
nodes = pd.read_csv('./examples/example_nodes.tsv', sep='\t', index_col=0)

netplotbrain.plot(template='MNI152NLin2009cAsym', view='LS', templatestyle='glass', nodes=nodes, edges=edges, nodecolor='community')

Leads to an error message:

TypeError                                 Traceback (most recent call last)
Input In [11], in <cell line: 7>()
      4 edges = pd.read_csv('./examples/example_edges.tsv', sep='\t', index_col=0)
      5 nodes = pd.read_csv('./examples/example_nodes.tsv', sep='\t', index_col=0)
----> 7 netplotbrain.plot(template='MNI152NLin2009cAsym', view='LS', templatestyle='glass', nodes=nodes, edges=edges, nodecolor='community')

File ~/packages/netplotbrain/netplotbrain/plot.py:261, in plot(nodes, fig, ax, view, frames, edges, template, templatestyle, network, arrowaxis, arroworigin, edgecolor, nodesize, nodecolor, nodetype, nodecmap, edgecmap, edgeweights, nodes_df, hemisphere, title, highlightnodes, highlightedges, showlegend, **kwargs)
    259 else:
    260     legend_subplotp_colind = int(np.round(spind / 2) - 1)
--> 261 ax = fig.add_subplot(gridspec[nrows + li, legend_subplotp_colind])
    262 if legend == 'nodesize':
    263     ax = _add_nodesize_legend(ax, nodes, nodesize, **profile)

File ~/anaconda3/lib/python3.9/site-packages/matplotlib/gridspec.py:257, in GridSpecBase.__getitem__(self, key)
    254     except ValueError as err:
    255         raise ValueError("Unrecognized subplot spec") from err
    256     num1, num2 = np.ravel_multi_index(
--> 257         [_normalize(k1, nrows, 0), _normalize(k2, ncols, 1)],
    258         (nrows, ncols))
    259 else:  # Single key
    260     num1, num2 = _normalize(key, nrows * ncols, None)

File ~/anaconda3/lib/python3.9/site-packages/matplotlib/gridspec.py:240, in GridSpecBase.__getitem__.<locals>._normalize(key, size, axis)
    237     raise IndexError("GridSpec slice would result in no space "
    238                      "allocated for subplot")
    239 else:
--> 240     if key < 0:
    241         key = key + size
    242     if 0 <= key < size:

TypeError: '<' not supported between instances of 'list' and 'int'

No error if nodecolor is not specified or string to views is 1 or 3 in length.

Includes:

• Dynamically place origin of arrows based on view. (i.e. make sure the arrows are always in front of brain)
• Dynamically choose which arrows are shown (e.g. calculate the 6 letter placements in relation to view, and remove letters with smallest distance)
• Fix rotation bug

Allow users to specify colors from dataframe column or matplotlib color.

the data is below:

I have reinstalled the package netplotbrain and matplotlib, but it still show me error.

All examples in

Should have explanations to how to achieve the effects in a specific page in the documentation where each the unique concept in each plot is explained as a tutorial.

This is a collection of the long term roadmap (will start to use projects if more people contribute to this)

Released in #60

• Improve svg saving for all templates (ensure all template backgrounds are rasteriszed)
• More legend freedom (increase sizes) (see #57) -> currently on develop branch
• Link published article in docs
• Plot from bids
• Include connectivity matrices (like spring plots) that can be specified in views
• Experienced some a bug with pybids layout.get_entities() sometimes failing. Need to investigate/create minimal example of to report/find workaround. -> unsure why but was caused but getting unique entities from the main folder and not the derivative folder was causing the error

Bug fixes

• #52 Parcel plots with single hemisphere
  ... And any others found

Enhancements

• Better arrows for directions. Consider no arrows but just lines or lines with sphere (see #7)
• Improve gif plots (see #29)
• Improve continuous colorbar use/flexibility (specify cbar tick range instead of just being extremes).
• If node_color organizes connectivity matrix, then add color lines to connectivity matrix diagonal

New features

• Include surface plots - no reasons they can't be included #49
• Allow to shuffle parcel id numbers to improve colorbar distirbution (e.g. sequential)
• More BIDSoperations

Improved Documentation

• Add BIDS and CM feature in README/Main page as a major feature.
• Tutorial: BIDS
• Gallery example of BIDS
• Gallery example of legend_span
• Gallery example, rotated cm

Not really a bug, but kinda. Title is currently one of the variables that has the behaviour that is asked for in #27 however it is the one variable where this behaviour is not wanted to be repeated if given a singular string.

When specifying title in a multi-axis figure it has to be something like (here 3 panels):

['', 'This is the title', '']

For it to only appear once. Otherwise, if:

title='This is the title'

is given to netplotbrain.plot, it will generate:

['This is the title', 'This is the title', 'This is the title']

when the desired behaviours is going to generally be:

['', 'This is the title', '']

Additionally, we would probably want a titlealign argument where, if "left" will make

['This is the title', '', '']

And so on.

For the user's sake we should have consistent input for view and hemisphere.

Currently working on #31 and noticed some weirdness in user interface. For a 2x2 view plotting L/R hemispheres form L/R views.

At the moment:

view = [['L' 'R'], ['R', 'L']]
hemisphere = [['L', 'R'], ['L', 'R']]

and

view = [['LR'], ['RL']]
hemisphere = [['LR'], ['LR']]

will both lead to an error

At present what is needed to get the above to work is:

view = [['LR'], ['RL']]
hemisphere = [['L', 'R'], ['L', 'R']]

Desired outcome:

Both the invalid examples should work.

Function that needs modifying:

_get_frame_input

Difficulty:

moderate (cause modifications to _get_frame_input could change other functions)

Provided template files are not load.
ImageFileError: Cannot work out file type of ".cache\templateflow\tpl-MNI152NLin2009cAsym\tpl-MNI152NLin2009cAsym_res-01_desc-brain_T1w.nii.gz"

Code that I am using:
netplotbrain.plot(template='MNI152NLin2009cAsym', templatestyle='surface', nodes=new_nodes, edges=edges, templatealpha=1, nodecolor='blue', templatecolor='gray', view='L') plt.show()

I have tried also to download the file and load this as a template but then the surface is not showing.

Take examples (or amended examples) from the examples in README.md or ./examples/showcase.py and move them to new documentation.

See how examples in docs/gallary/*.md and docs/gallary.md are written for inspiration.

Desired outcome: Allow nodes and edges to be plotted from a networkx network.

Proposed implementation: Add argument network to plot with a function that converts networkx.Graph to create (or add to) nodes and edges dataframes at the start of netplotbrain.plot.

Difficulty: Easy (some networkx knowledge probably needed)

This looks like a great python alternative to other options (e.g., BrainNet Viewer). One feature that is helpful in BrainNet Viewer is the ability to add a text label to certain nodes in the figure. Is this a feature that could be added (or is this already implemented and I have missed it)?

Thanks so much.

Make it possible for the discrete legend to be bigger.

Some ideas:

Allow legend_size = ‘large’ to span all subplots. Alternatively, allow the legend to be longer

Have an argument like legend_discrete_Boer row to allow multiple rows and set it to 3 or 4

Make default text higher

Try and detect when text overlaps.

None of these are decided, just writing notes to myself (and allowing others to contribute idea about how to make them better)

Probabilistic atlases (e.g., difumo) are not piecewise smooth.

It would be interesting to weigh the colors of the ROIs and modify the alpha channel to make them transparent as you move away from the ROI and enter a different one.

From the example currently in the readme:

fig = plt.figure()
ax_m2 = fig.add_subplot(121, projection='3d')
netplotbrain.plot(template='MNI152NLin2009cAsym',
                  templatestyle='surface',
                  fig = fig,
                  ax = ax_m2,
                  view='A',
                  nodes=nodes,
                  nodesize='centrality_measure1',
                  nodecolor='red',
                  edges=edges)

ax_m2 = fig.add_subplot(122, projection='3d')
netplotbrain.plot(template='MNI152NLin2009cAsym',
                  templatestyle='surface',
                  fig = fig,
                  ax = ax_m2,
                  view='A',
                  nodes=nodes,
                  nodesize='centrality_measure2',
                  nodecolor='blue',
                  edges=edges)
plt.show()

We could make arguments be lists across views. So the above would be

netplotbrain.plot(template='MNI152NLin2009cAsym',
                  templatestyle='surface',
                  view=['A', 'A'],
                  nodes=nodes,
                  nodesize=['centrality_measure1', 'centrality_measure2'],
                  nodecolor=['red', 'blue'],
                  edges=edges)
plt.show()

Benfitis:

• Less code for user to write
• Less figure management by user.

Negatives:

• All input arguments would have to be checked for this behaviour. (So quite a bit of work. Although the function get_frame_input could do most of the work, if modified.)
• Breaks legend behaviour if multiple legends are needed (e.g. example above) unless there is some way to detect when a legend would needed to be repeated or not.

The following minimal example

import netplotbrain
netplotbrain.plot(nodes=None,
                  edges=None,
                  template=None,
                  view=['SS'])

results in the following error:

~/packages/netplotbrain/netplotbrain/plot.py in plot(nodes, fig, ax, view, frames, edges, template, templatestyle, templatevoxsize, arrowaxis, arroworigin, edgecolor, nodesize, nodecolor, nodetype, nodecolorby, nodecmap, edgeweights, nodecols, nodeimg, hemisphere, title, highlightnodes, showlegend, **kwargs)
    244     for ri in range(nrows):
    245         # Get the azim, elev and arrowaxis for each row
--> 246         azim, elev, arrowaxis_row = _get_view(
    247             view[ri], frames, arrowaxis=arrowaxis)
    248         for fi in range(frames):

~/packages/netplotbrain/netplotbrain/utils/plot_utils.py in _get_view(views, frames, arrowaxis)
     78         for n in range(frames):
     79             if direction == '+':
---> 80                 vx.append(v1x + n * ((v2x - v1x) / (frames - 1)))
     81                 vy.append(v1y + n * ((v2y - v1y) / (frames - 1)))
     82             elif direction == '-':

ZeroDivisionError: division by zero

Improvement to make input more unified.

• nodesimg should be specified using nodes. If nodes is a dictionary, valid path, or nifti, assume it is nodesimg.
• nodes_df should be allowed to supplement a nodsimg file for colour/size/highlighting. Implementation: Depends how the above is implemented
• edges_df should also be allowed for when the input is a numpy array. Implementation: if edges is numpy array, edges_df outer merges with it after it is converted edges.merge(edges_df, how='outer').

Hello,

I'd like to visualise a weighted network (weight varies from 0-1) by colour. With the following code:

netplotbrain.plot(
template=template
nodes=nodes
edges=data_array,
edge_cmap='viridis',
edge_colorvminvmax=[0,1])

I get a network with only black edges. Please let me know what I'm doing wrong! Thanks very much for your time.

Currently gif duration is in ms regardless of number of figures. Since gif will generally be combined with the frames argument, a gifframerate to replace gifduration is probably better.

Or at least as a possible alternative. So that if gifframerate is not None, gifduration = (framerate * 1000) * n_frames.

I think multiply with a 1000 cause framerate is natural in seconds and gifduration is currently in ms.

Theoretically should not be hard to do.

• Add gifti/cifti loading of the file.
• check each templatestyles or throw errors for templatestyles that are not compatible. Perhaps only allow surface as a templatestyle for surface images.
• plot surfaces in resource effective way.