feat(plotting): add plot_matrix for connectivity/correlation matrices#243
Open
sdiebolt wants to merge 14 commits into
Open
feat(plotting): add plot_matrix for connectivity/correlation matrices#243sdiebolt wants to merge 14 commits into
sdiebolt wants to merge 14 commits into
Conversation
…ines The recipe only regenerated TestPlotVolume/TestPlotContours baselines, so running `just generate-baselines` silently deleted the carpet, composite, and (new) matrix baseline PNGs without recreating them.
Adds plot_matrix, inspired by nilearn.plotting.plot_matrix, for displaying 2D matrices (numpy.ndarray or xarray.DataArray) with optional lower/diag triangle masking, grid lines, and a symmetric diverging colormap. Also adds a groups feature (no nilearn equivalent): contiguous label runs are annotated with colored rectangle strips alongside the matrix, useful for marking anatomical groupings (e.g. cortex, thalamus) when there are too many individual labels to read. Moves the color/fontsize helpers shared between image.py and the new matrix.py into the module-private plotting/_utils.py. Refs #241
Contributor
|
📖 Doc preview: https://confusius.tools/pr-preview/pr-243/ |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
Brings plot_matrix to 100% coverage: missing-coordinate/empty-labels defaults, all three tri grid variants, show_group_labels=False, caller-provided ax, and cbar_label. codecov/patch was failing on the uncovered lines.
… labels The group-strip axes share their perpendicular axis with the main axes (sharex/sharey) so the strips line up with rows/columns. Clearing that shared axis's ticks with set_xticks([])/set_yticks([]) propagated through the shared locator and wiped out the main axes' own row/column labels too. Use tick_params to hide the strip's ticks/labels instead of resetting the shared locator, and widen the row-strip pad so labels have room to render next to it instead of being covered by the opaque strip. Found while building a real-world plot_matrix example (fUSI connectivity matrix with anatomical region groups) where both labels and groups were used together for the first time.
Demonstrates an end-to-end regional connectivity analysis: register a single-slice fUSI recording (Nunez-Elizalde 2022, CR022) to the Pepe-Mariani 2026 template with a rigid transform, resample the Allen Mouse Brain Atlas onto the recording's native grid, extract region signals for a handful of cortical/hippocampal/thalamic/hypothalamic ROIs, and plot their correlation matrix with plot_matrix's new groups annotation. Wires the new section into the gallery nav, prefetches the Nunez-Elizalde recording, the Pepe-Mariani template, and the allen_mouse_100um BrainGlobe atlas in CI, and adds matching actions/cache steps for the template and BrainGlobe atlas downloads (the Nunez-Elizalde recording reuses the existing dataset cache step).
The row group-strip axis was inserted with a fixed pad, which only left enough room for short labels; longer labels (e.g. region acronyms with a hemisphere suffix) got visually covered by the opaque strip even though the label text itself was set correctly. Measure the rendered y tick label width (Nilearn's _fit_axes technique) and size the pad from it instead. Found while building a real-world connectivity-matrix example with longer region labels.
- Use cf.<module>.<func> consistently instead of importing individual functions, applied across the example gallery for consistency. - Resample the recording to a uniform time grid before any time-domain processing (its raw timestamps are not perfectly uniform). - Extract left/right hemispheres separately via get_masks(sides=...) and order each area's regions medial-to-lateral, mirrored across the midline (lateral-to-medial on the left, medial-to-lateral on the right), so each area block reads as one continuous sweep across the slice with homotopic pairs meeting in the middle. - Clean the averaged region signals (high-pass filter + white-matter aCompCor regression) before computing the correlation matrix.
Renames "diag" to "diag_lower" and adds "diag_upper" and "upper", mirroring "lower"/"diag_lower" across the diagonal. Together with the existing ax= support, this lets two matrices be overlaid on the same axes (e.g. an estimate on the lower triangle and its significance mask on the upper), by calling plot_matrix twice with complementary tri values and show_colorbar=False on the second call.
…ample The narrative said "rigid transform" while the code already used transform_type="affine". Also note the visible improvement after registration, most noticeable around the anterior choroidal arteries in the bottom of the field of view.
…zing _y_tick_label_width_inches falls back to 0.0 when get_tightbbox returns None, which happens for real (not just hypothetically) when labels are all empty strings: they have no rendered extent. codecov/patch flagged this line as uncovered.
get_masks accepts sides as a per-region list, so both hemispheres can be requested in a single call (region acronyms duplicated, sides list matching) instead of two separate calls concatenated together. Each layer still needs a unique id reassigned afterward since get_masks reuses the same region id for both hemispheres of a region.
Member
Author
|
@FelipeCybis Ready to review! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
plot_matrix, inspired bynilearn.plotting.plot_matrix, for plotting 2D matrices (numpy.ndarrayorxarray.DataArray) with lower/diag triangle masking, grid lines, and a symmetric diverging colormap.groupsfeature (no Nilearn equivalent): contiguous label runs get colored rectangle strips alongside the matrix, useful for marking anatomical groupings (e.g. cortex, thalamus) when there are too many individual labels to read.image.pyand the newmatrix.pyinto module-privateplotting/_utils.py.just generate-baselines, which only regenerated two of the visual-regression test classes and was silently deleting the rest.Closes #241