Deselect clusters

napari_clusters_plotter/_Qt_code.py

@@ -12,8 +12,8 @@
 from matplotlib.path import Path
 from matplotlib.widgets import LassoSelector, RectangleSelector, SpanSelector
 from napari.layers import Image, Layer
-from qtpy.QtCore import QRect
-from qtpy.QtGui import QIcon
+from qtpy.QtCore import QRect, Qt
+from qtpy.QtGui import QGuiApplication, QIcon
 from qtpy.QtWidgets import (
     QAbstractItemView,
     QHBoxLayout,
@@ -390,24 +390,54 @@ def create_options_dropdown(name: str, value, options: dict, label: str):
 
 
 class SelectFrom2DHistogram:
-    def __init__(self, parent, ax, full_data):
+    def __init__(self, parent, ax, full_data, histogram):
         self.parent = parent
         self.ax = ax
         self.canvas = ax.figure.canvas
         self.xys = full_data
+        self.cluster_id_histo_overlay = None
+        self.histogram = histogram
 
         self.lasso = LassoSelector(ax, onselect=self.onselect)
         self.ind = []
         self.ind_mask = []
 
+    def vert_to_coord(self, vert):
+        """
+        Converts verticis to histogram coordinates in pixels
+        """
+
+        # I tried to solve it with self.ax.transData.transform... but it did not work...
+        xrange = self.histogram[1][-1] - self.histogram[1][0]
+        yrange = self.histogram[2][-1] - self.histogram[2][0]
+        v = (
+            (vert[0] - self.histogram[1][0]) / (xrange) * self.histogram[0].shape[0],
+            (vert[1] - self.histogram[2][0]) / (yrange) * self.histogram[0].shape[1],
+        )
+
+        coord = tuple([int(c) for c in v])
+        return coord
+
     def onselect(self, verts):
-        path = Path(verts)
+        if self.parent.manual_clustering_method is None:
+            return
 
-        self.ind_mask = path.contains_points(self.xys)
-        self.ind = np.nonzero(self.ind_mask)[0]
+        modifiers = QGuiApplication.keyboardModifiers()
 
-        if self.parent.manual_clustering_method is not None:
-            self.parent.manual_clustering_method(self.ind_mask)
+        if (
+            modifiers == Qt.ControlModifier and len(verts) == 2
+        ):  # has len of 2 when single click was done
+            coord_click = self.vert_to_coord(verts[0])
+            cluster_id_to_delete = self.cluster_id_histo_overlay[coord_click[::-1]][0]
+            if cluster_id_to_delete > 0:
+                self.parent.manual_clustering_method(
+                    np.zeros(shape=self.xys.shape), delete_cluster=cluster_id_to_delete
+                )
+                return
+
+        path = Path(verts)
+        self.ind_mask = path.contains_points(self.xys)
+        self.parent.manual_clustering_method(self.ind_mask)
 
     def disconnect(self):
         self.lasso.disconnect_events()
@@ -521,6 +551,7 @@ class MplCanvas(FigureCanvas):
     def __init__(self, parent=None, width=7, height=4, manual_clustering_method=None):
         self.fig = Figure(figsize=(width, height), constrained_layout=True)
         self.manual_clustering_method = manual_clustering_method
+        self.parent = parent
 
         self.axes = self.fig.add_subplot(111)
         self.histogram = None
@@ -551,6 +582,9 @@ def __init__(self, parent=None, width=7, height=4, manual_clustering_method=None
 
         self.reset()
 
+    def set_selector_cluster_id_overlay(self, overlay: np.array):
+        self.selector.cluster_id_histo_overlay = overlay
+
     def reset_zoom(self):
         if self.xylim:
             self.axes.set_xlim(self.xylim[0])
@@ -617,7 +651,9 @@ def make_2d_histogram(
 
         full_data = pd.concat([pd.DataFrame(data_x), pd.DataFrame(data_y)], axis=1)
         self.selector.disconnect()
-        self.selector = SelectFrom2DHistogram(self, self.axes, full_data)
+        self.selector = SelectFrom2DHistogram(
+            self, self.axes, full_data, self.histogram
+        )
         self.axes.figure.canvas.draw_idle()
 
     def make_1d_histogram(

napari_clusters_plotter/_plotter.py

@@ -90,8 +90,9 @@ def __init__(self, napari_viewer):
 
         self.analysed_layer = None
         self.visualized_layer = None
+        self.cluster_id_histo_overlay = None
 
-        def manual_clustering_method(inside):
+        def manual_clustering_method(inside, **kwargs):
             inside = np.array(inside)  # leads to errors sometimes otherwise
 
             if self.analysed_layer is None or len(inside) == 0:
@@ -101,7 +102,14 @@ def manual_clustering_method(inside):
             features = get_layer_tabular_data(self.analysed_layer)
 
             modifiers = QGuiApplication.keyboardModifiers()
-            if modifiers == Qt.ShiftModifier and clustering_ID in features.keys():
+            if "delete_cluster" in kwargs:
+                features[clustering_ID].mask(
+                    features[clustering_ID] == kwargs["delete_cluster"],
+                    other=-1,
+                    inplace=True,
+                )
+
+            elif modifiers == Qt.ShiftModifier and clustering_ID in features.keys():
                 features[clustering_ID].mask(
                     inside, other=features[clustering_ID].max() + 1, inplace=True
                 )
@@ -677,7 +685,7 @@ def run(
                         log_scale=self.log_scale.isChecked(),
                     )
 
-                    rgb_img = make_cluster_overlay_img(
+                    rgb_img, self.cluster_id_histo_overlay = make_cluster_overlay_img(
                         cluster_id=plot_cluster_name,
                         features=features,
                         feature_x=self.plot_x_axis_name,
@@ -686,6 +694,9 @@ def run(
                         histogram_data=self.graphics_widget.histogram,
                         hide_first_cluster=self.plot_hide_non_selected.isChecked(),
                     )
+                    self.graphics_widget.set_selector_cluster_id_overlay(
+                        self.cluster_id_histo_overlay
+                    )
                     xedges = self.graphics_widget.histogram[1]
                     yedges = self.graphics_widget.histogram[2]
 

napari_clusters_plotter/_plotter_utilities.py

@@ -549,6 +549,7 @@ def make_cluster_overlay_img(
         ]
 
     cluster_overlay_rgba = np.zeros((*h.shape, 4), dtype=float)
+    cluster_overlay_cluster_id = np.zeros((*h.shape, 1), dtype=int)
     output_max = np.zeros(h.shape, dtype=float)
 
     for cluster, entries in relevant_entries.groupby(cluster_id):
@@ -565,5 +566,8 @@ def make_cluster_overlay_img(
         ]
         rgba.append(0.9)
         cluster_overlay_rgba[mask] = rgba
+        cluster_overlay_cluster_id[mask] = cluster
 
-    return cluster_overlay_rgba.swapaxes(0, 1)
+    return cluster_overlay_rgba.swapaxes(0, 1), cluster_overlay_cluster_id.swapaxes(
+        0, 1
+    )