NF: Adds overlay option in OrthoSlicer3D

Beginning of functionality to allow overlays in OrthoSlicer3D
nipy · rmarkello · Dec 9, 2019 · Apr 9, 2020 · Apr 9, 2020 · Jul 27, 2020
commit 989ac26e6d60962a4f8ac8ce077d5d2181311842
diff --git a/nibabel/viewers.py b/nibabel/viewers.py
@@ -69,6 +69,9 @@ def __init__(self, data, affine=None, axes=None, title=None):
         self._title = title
         self._closed = False
         self._cross = True
+        self._overlay = None
+        self._threshold = None
+        self._alpha = 1
 
         data = np.asanyarray(data)
         if data.ndim < 3:
@@ -285,6 +288,111 @@ def clim(self, clim):
         self._clim = tuple(clim)
         self.draw()
 
+    @property
+    def overlay(self):
+        """The current overlay """
+        return self._overlay
+
+    @property
+    def threshold(self):
+        """The current data display threshold  """
+        return self._threshold
+
+    @threshold.setter
+    def threshold(self, threshold):
+        # mask data array
+        if threshold is not None:
+            self._data = np.ma.masked_array(np.asarray(self._data),
+                                            np.asarray(self._data) <= threshold)
+            self._threshold = float(threshold)
+        else:
+            self._data = np.asarray(self._data)
+            self._threshold = threshold
+
+        # update current volume data w/masked array and re-draw everything
+        if self._data.ndim > 3:
+            self._current_vol_data = self._data[..., self._data_idx[3]]
+        else:
+            self._current_vol_data = self._data
+        self._set_position(None, None, None, notify=False)
+
+    @property
+    def alpha(self):
+        """ The current alpha (transparency) value """
+        return self._alpha
+
+    @alpha.setter
+    def alpha(self, alpha):
+        alpha = float(alpha)
+        if alpha > 1 or alpha < 0:
+            raise ValueError('alpha must be between 0 and 1')
+        for im in self._ims:
+            im.set_alpha(alpha)
+        self._alpha = alpha
+        self.draw()
+
+    def set_overlay(self, data, affine=None, threshold=None, cmap='viridis'):
+        if affine is None:
+            try:  # did we get an image?
+                affine = data.affine
+                data = data.dataobj
+            except AttributeError:
+                pass
+
+        # check that we have sufficient information to match the overlays
+        if affine is None and data.shape[:3] != self._data.shape[:3]:
+            raise ValueError('Provided `data` do not match shape of '
+                             'underlay and no `affine` matrix was '
+                             'provided. Please provide an `affine` matrix '
+                             'or resample first three dims of `data` to {}'
+                             .format(self._data.shape[:3]))
+
+        # we need to resample the provided data to the already-plotted data
+        if not np.allclose(affine, self._affine):
+            from .processing import resample_from_to
+            from .nifti1 import Nifti1Image
+            target_shape = self._data.shape[:3] + data.shape[3:]
+            # we can't just use SpatialImage because we need an image type
+            # where the spatial axes are _always_ first
+            data = resample_from_to(Nifti1Image(data, affine),
+                                    (target_shape, self._affine)).dataobj
+            affine = self._affine
+
+        if self._overlay is not None:
+            # remove all images + cross hair lines
+            for nn, im in enumerate(self._overlay._ims):
+                im.remove()
+                for line in self._overlay._crosshairs[nn].values():
+                    line.remove()
+            # remove the fourth axis, if it was created for the overlay
+            if (self._overlay.n_volumes > 1 and len(self._overlay._axes) > 3
+               and self.n_volumes == 1):
+                a = self._axes.pop(-1)
+                a.remove()
+
+        # create an axis if we have a 4D overlay (vs a 3D underlay)
+        axes = self._axes
+        o_n_volumes = int(np.prod(data.shape[3:]))
+        if o_n_volumes > self.n_volumes:
+            axes += [axes[0].figure.add_subplot(224)]
+        elif o_n_volumes < self.n_volumes:
+            axes = axes[:-1]
+
+        # mask array for provided threshold
+        self._overlay = self.__class__(data, affine=affine, axes=axes)
+        self._overlay.threshold = threshold
+
+        # set transparency and new cmap
+        self._overlay.cmap = cmap
+        for im in self._overlay._ims:
+            im.set_alpha(0.7)
+
+        # no double cross-hairs (they get confused when we have linked orthos)
+        for cross in self._overlay._crosshairs:
+            cross['horiz'].set_visible(False)
+            cross['vert'].set_visible(False)
+        self._overlay._draw()
+
     def link_to(self, other):
         """Link positional changes between two canvases
 
@@ -412,7 +520,7 @@ def _set_position(self, x, y, z, notify=True):
             idx = [slice(None)] * len(self._axes)
             for ii in range(3):
                 idx[self._order[ii]] = self._data_idx[ii]
-            vdata = self._data[tuple(idx)].ravel()
+            vdata = np.asarray(self._data[tuple(idx)].ravel())
             vdata = np.concatenate((vdata, [vdata[-1]]))
             self._volume_ax_objs['patch'].se
42F5
t_x(self._data_idx[3] - 0.5)
             self._volume_ax_objs['step'].set_ydata(vdata)