Spaces:
Running on Zero
Running on Zero
| """The page's views on constructed inputs (no GPU, no layout): the FOD peaks of a delta FOD are its direction, the slice | |
| viewer draws with and without the overlay, the 3-D views hold the sample they say, the timings rows are the stages.""" | |
| import numpy as np | |
| from space import pipeline as P | |
| from space import viewers as V | |
| CFG = P.config() | |
| def test_the_peak_of_a_delta_fod_is_its_direction(): | |
| from dmipy_tract import sh_matrix | |
| d = np.array([[0.6, 0.0, 0.8], [0.0, 1.0, 0.0]]) | |
| sh = np.zeros((2, 1, 1, 45)); sh[0, 0, 0] = sh_matrix(8, d[:1])[0]; sh[1, 0, 0] = sh_matrix(8, d[1:])[0] | |
| peaks, amp = P.peaks(sh) | |
| assert peaks.shape == (2, 1, 1, 3) and (amp > 0).all() | |
| for k in range(2): | |
| assert abs(abs(peaks[k, 0, 0] @ d[k]) - 1.0) < 0.01 # within the hemisphere's angular resolution | |
| zero = np.zeros((1, 1, 1, 45)); zero[..., 0] = np.nan | |
| p0, a0 = P.peaks(zero) | |
| assert (p0 == 0).all() and (a0 == 0).all() | |
| def _result(): | |
| p = P.Protocol((P.Shell("d12-D24", 1000, 6),), n_b0=1) | |
| m = P.measurements(p, CFG["shapes"]) | |
| rng = np.random.default_rng(0) | |
| dwi = rng.random((6, 6, 6, 7)).astype(np.float32); dwi[0, 0] = np.nan | |
| sh = np.zeros((6, 6, 6, 45)); sh[..., 0] = 1.0 | |
| peaks, amp = P.peaks(sh) | |
| return dwi, m, peaks, amp | |
| def test_the_slice_viewer_draws_with_and_without_the_overlay(): | |
| dwi, m, peaks, amp = _result() | |
| a = V.dwi_slice(dwi, m, 3, 2, peaks=peaks, peak_amp=amp, overlay=True) | |
| b = V.dwi_slice(dwi, m, 3, 0) | |
| assert a.size[0] > 100 and b.size[0] > 100 | |
| def test_the_3d_views_hold_their_sample(): | |
| import nibabel as nib | |
| import os | |
| rois = np.asarray(nib.load(os.path.join(P.DATA_DIR, "DiSCo_ROIs.nii.gz")).dataobj).astype(np.int32) | |
| strands = [np.cumsum(np.random.default_rng(i).normal(size=(20, 3)), 0) + 20 for i in range(50)] | |
| regions = V.region_markers(rois) | |
| assert regions.mode == "markers+text" and len(regions.x) == 16 | |
| fig = V.strands3d(strands, np.full(50, 2e-6), regions, rois.shape, n=30) | |
| lines = [t for t in fig.data if t.mode == "lines"] | |
| assert sum(int(np.isnan(np.asarray(t.x, float)).sum()) for t in lines) == 30 # one NaN gap per path | |
| assert "2.0-2.0 µm" in fig.layout.title.text | |
| from dmipy_tract import Tractogram | |
| pts = np.concatenate(strands[:10]).astype(np.float32); offsets = np.r_[0, np.cumsum([len(s) for s in strands[:10]])] | |
| tg = Tractogram(pts, offsets, np.arange(10), np.zeros((10, 2), np.int8)) | |
| fig2 = V.tractogram3d(tg, regions, rois.shape, n=4, total=12345) | |
| assert sum(int(np.isnan(np.asarray(t.x, float)).sum()) for t in fig2.data if t.mode == "lines") == 4 | |
| assert "4 of 12,345" in fig2.layout.title.text | |
| assert V.ground_truth_matrix(np.eye(16) * 0 + np.triu(np.ones((16, 16)), 1), np.triu(np.ones((16, 16)), 1) * 0.5, 120).size[0] > 100 | |
| def test_the_timings_rows_are_the_stages(): | |
| rows = V.timings_rows(dict(replay=1.0, csd=2.0, total=3.0), 10.0) | |
| assert rows[0][0].startswith("source") and [r[0] for r in rows[1:]] == ["replay", "csd", "total"] | |
| def test_the_floor_slice_draws_the_positive_voxels(): | |
| floor = np.zeros((8, 8, 4), np.float32); floor[2:6, 2:6, :] = 0.01; floor[3, 3, 1] = 0.05 | |
| img = V.floor_slice(floor, 1, 0.01, label="A: ") | |
| assert img.size[0] > 100 and img.size[1] > 100 | |
| def test_the_spread_matrices_draw(): | |
| M = np.zeros((16, 16)); M[0, 1] = M[1, 0] = 10 | |
| sp = P.pair_spread([M, M * 1.2], [dict(pearson_count=0.8), dict(pearson_count=0.82)], key="pearson_count") | |
| assert V.spread_matrices(sp).size[0] > 100 | |
| def test_the_map_slice_draws_plain_and_diverging(): | |
| vol = np.random.default_rng(0).normal(size=(6, 6, 3)); vol[0, 0] = np.nan | |
| assert V.map_slice(vol, 1, "t").size[0] > 100 and V.map_slice(vol, 1, "d", symmetric=True).size[0] > 100 | |
| assert V.map_slice(np.full((6, 6, 3), np.nan), 1, "empty", symmetric=True).size[0] > 100 | |