nettracer3d 0.5.4__tar.gz → 0.5.6__tar.gz

{nettracer3d-0.5.4/src/nettracer3d.egg-info → nettracer3d-0.5.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: nettracer3d
-Version: 0.5.4
+Version: 0.5.6
 Summary: Scripts for intializing and analyzing networks from segmentations of three dimensional images.
 Author-email: Liam McLaughlin <mclaughlinliam99@gmail.com>
 Project-URL: User_Tutorial, https://www.youtube.com/watch?v=cRatn5VTWDY
@@ -44,10 +44,6 @@ NetTracer3D is free to use/fork for academic/nonprofit use so long as citation i
 
 NetTracer3D was developed by Liam McLaughlin while working under Dr. Sanjay Jain at Washington University School of Medicine.
 
--- Version 0.5.4 updates --
+-- Version 0.5.6 updates --
 
-1. Added new function to GUI in image -> overlays -> color nodes/edges. Generates a rgb array corresponding to the nodes/edge labels where each node/edge (depending which array is selected) is randomly assigned a unique rgb color in an overlay channel. This can be used, for example, to color code labeled branches for easy identification of which branch is which.
-
-2. Improved highlight overlay general functionality (for selecting nodes/edges). Previously selecting a node/edge had the program attempting to create an equal sized array as an overlay, find all objects corresponding to the selected ones, fill those into the new highlight overlay, then overlay that image. This was understandably quite slow in big arrays where the system was wasting a lot of time searching the entire array every time something was selected. New version retains this functionality for arrays below 125 million voxels, since search time is rather manageable at that size. For larger arrays, it instead draws the highlight for the selected objects only into the current slice, rendering a new slice whenever the user scrolls in the stack (although the entire highlight overlay is still initialized as a placeholder). Functions that require the use of the entire highlight overlay (such as masking) are correspondingly updated to draw the entirety of the highlight overlay before executing (when the system has up until that point been drawing slices one at a time). This will likely be the retained behavior moving forward, although to eliminate this behavior, one can open nettracer_gui.py and set self.mini_thresh to some comically large value. The new highlight overlay seems to work effectively the same but faster in my testing although it is possible a bug slipped through, which I will fix if informed about (or if I find it myself).
-
-3. For the machine learning segmenter, changed the system to attempt to segment the image by chunking the array into the largest possible chunks that can be divided across all CPU cores. Previously the system split the array into 64^3 voxel sized chunks and passed those to the CPU cores until everything was processed. I am not sure which version is more efficient/faster so this is somewhat of a test. In theory the new behavior could be faster because it asking Python to interpret less stuff. 
+1. Minor change - Updated some of the overhead and post-random forrest voxel assignment for the segmenter to be faster. The major bottleneck for this is still querying sklearn which seems a bit harder to work around.

nettracer3d-0.5.6/README.md

@@ -0,0 +1,13 @@
+NetTracer3D is a python package developed for both 2D and 3D analysis of microscopic images in the .tif file format. It supports generation of 3D networks showing the relationships between objects (or nodes) in three dimensional space, either based on their own proximity or connectivity via connecting objects such as nerves or blood vessels. In addition to these functionalities are several advanced 3D data processing algorithms, such as labeling of branched structures or abstraction of branched structures into networks. Note that nettracer3d uses segmented data, which can be segmented from other softwares such as ImageJ and imported into NetTracer3D, although it does offer its own segmentation via intensity and volumetric thresholding, or random forest machine learning segmentation. NetTracer3D currently has a fully functional GUI. To use the GUI, after installing the nettracer3d package via pip, enter the command 'nettracer3d' in your command prompt:
+
+
+This gui is built from the PyQt6 package and therefore may not function on dockers or virtual envs that are unable to support PyQt6 displays. More advanced documentation is coming down the line, but for now please see: https://www.youtube.com/watch?v=cRatn5VTWDY
+for a video tutorial on using the GUI.
+
+NetTracer3D is free to use/fork for academic/nonprofit use so long as citation is provided, and is available for commercial use at a fee (see license file for information).
+
+NetTracer3D was developed by Liam McLaughlin while working under Dr. Sanjay Jain at Washington University School of Medicine.
+
+-- Version 0.5.6 updates --
+
+1. Minor change - Updated some of the overhead and post-random forrest voxel assignment for the segmenter to be faster. The major bottleneck for this is still querying sklearn which seems a bit harder to work around.

{nettracer3d-0.5.4 → nettracer3d-0.5.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "nettracer3d"
-version = "0.5.4"
+version = "0.5.6"
 authors = [
   { name="Liam McLaughlin", email="mclaughlinliam99@gmail.com" },
 ]

{nettracer3d-0.5.4 → nettracer3d-0.5.6}/src/nettracer3d/nettracer_gui.py

@@ -398,6 +398,7 @@ class ImageViewerWindow(QMainWindow):
         self.highlight_overlay = None
         self.highlight_bounds = None  # Store bounds for positioning
         self.mini_overlay = False # If the program is currently drawing the overlay by frame this will be true
+        self.mini_overlay_data = None #Actual data for mini overlay
         self.mini_thresh = (500*500*500) # Array volume to start using mini overlays for
 
     def start_left_scroll(self):
@@ -446,6 +447,7 @@ class ImageViewerWindow(QMainWindow):
         """
 
         self.mini_overlay = False #If this method is ever being called, it means we are rendering the entire overlay so mini overlay needs to reset.
+        self.mini_overlay_data = None
 
         def process_chunk(chunk_data, indices_to_check):
             """Process a single chunk of the array to create highlight mask"""
@@ -562,6 +564,8 @@ class ImageViewerWindow(QMainWindow):
 
     def create_highlight_overlay_slice(self, indices, bounds = False):
 
+        """Highlight overlay generation method specific for the segmenter interactive mode"""
+
 
         def process_chunk_bounds(chunk_data, indices_to_check):
             """Process a single chunk of the array to create highlight mask"""
@@ -638,7 +642,7 @@ class ImageViewerWindow(QMainWindow):
     def create_mini_overlay(self, node_indices = None, edge_indices = None):
 
         """
-        Create a binary overlay highlighting specific nodes and/or edges using parallel processing, one slice at a time for efficiency.
+        Create a highlight overlay one slice at a time.
         
         Args:
             node_indices (list): List of node indices to highlight
@@ -668,28 +672,29 @@ class ImageViewerWindow(QMainWindow):
         current_xlim = self.ax.get_xlim() if hasattr(self, 'ax') and self.ax.get_xlim() != (0, 1) else None
         current_ylim = self.ax.get_ylim() if hasattr(self, 'ax') and self.ax.get_ylim() != (0, 1) else None
         
-        if not node_indices and not edge_indices:
-            self.highlight_overlay = None
+        if not node_indices and not edge_indices: #Theoretically this can't be called because it uses full highlight overlay method for empty clicks
+            self.mini_overlay_data = None
+            self.mini_overlay = False
             self.update_display(preserve_zoom=(current_xlim, current_ylim))
             return
             
-        # Get the shape of the full array from any existing channel
+        # Get the shape of the mini array from any existing channel
         for channel in self.channel_data:
             if channel is not None:
                 full_shape = channel.shape
+                full_shape = (full_shape[1], full_shape[2]) #Just get (Y, X) shape
                 break
         else:
             return  # No valid channels to get shape from
             
         # Initialize full-size overlay
-        if self.highlight_overlay is None:
-            self.highlight_overlay = np.zeros(full_shape, dtype=np.uint8)
+        self.mini_overlay_data = np.zeros(full_shape, dtype=np.uint8)
         
         # Get number of CPU cores
         num_cores = mp.cpu_count()
         
         # Calculate chunk size along y-axis
-        chunk_size = full_shape[1] // num_cores
+        chunk_size = full_shape[0] // num_cores
         if chunk_size < 1:
             chunk_size = 1
         
@@ -699,8 +704,8 @@ class ImageViewerWindow(QMainWindow):
                 
             # Create chunks
             chunks = []
-            for i in range(0, array_shape[1], chunk_size):
-                end = min(i + chunk_size, array_shape[1])
+            for i in range(0, array_shape[0], chunk_size):
+                end = min(i + chunk_size, array_shape[0])
                 chunks.append(channel_data[i:end, :])
                 
             # Process chunks in parallel using ThreadPoolExecutor
@@ -729,11 +734,10 @@ class ImageViewerWindow(QMainWindow):
                 edge_overlay = None
                         
         # Combine results
-        self.highlight_overlay[self.current_slice, :, :] = np.zeros_like(self.highlight_overlay[self.current_slice, :, :])
         if node_overlay is not None:
-            self.highlight_overlay[self.current_slice, :, :] = np.maximum(self.highlight_overlay[self.current_slice, :, :], node_overlay)
+            self.mini_overlay_data = np.maximum(self.mini_overlay_data, node_overlay)
         if edge_overlay is not None:
-            self.highlight_overlay[self.current_slice, :, :] = np.maximum(self.highlight_overlay[self.current_slice, :, :], edge_overlay)
+            self.mini_overlay_data = np.maximum(self.mini_overlay_data, edge_overlay)
 
                 
         # Update display
@@ -3488,6 +3492,8 @@ class ImageViewerWindow(QMainWindow):
             if self.mini_overlay == True: #If we are rendering the highlight overlay for selected values one at a time.
                 self.create_mini_overlay(node_indices = self.clicked_values['nodes'], edge_indices = self.clicked_values['edges'])
             self.update_display(preserve_zoom=view_settings)
+            if self.machine_window is not None:
+                self.machine_window.poke_segmenter()
             self.pending_slice = None
 
     def update_brightness(self, channel_index, values):
@@ -3623,7 +3629,15 @@ class ImageViewerWindow(QMainWindow):
             self.create_highlight_overlay_slice(self.targs, bounds = self.bounds)
 
         # Add highlight overlay if it exists
-        if self.highlight_overlay is not None and self.highlight and self.machine_window is None:
+        if self.mini_overlay and self.highlight and self.machine_window is None:
+            highlight_cmap = LinearSegmentedColormap.from_list(
+                'highlight',
+                [(0, 0, 0, 0), (1, 1, 0, 1)]  # yellow
+            )
+            self.ax.imshow(self.mini_overlay_data,
+                         cmap=highlight_cmap,
+                         alpha=0.5)
+        elif self.highlight_overlay is not None and self.highlight and self.machine_window is None:
             highlight_slice = self.highlight_overlay[self.current_slice]
             highlight_cmap = LinearSegmentedColormap.from_list(
                 'highlight',
@@ -6202,7 +6216,18 @@ class MachineWindow(QMainWindow):
         self.GPU.setChecked(False)
         self.GPU.clicked.connect(self.toggle_GPU)
         self.use_gpu = False
+        self.two = QPushButton("Train By 2D Slice Patterns (Cheaper - CPU only)")
+        self.two.setCheckable(True)
+        self.two.setChecked(True)
+        self.two.clicked.connect(self.toggle_two)
+        self.use_two = True
+        self.three = QPushButton("Train by 3D Patterns")
+        self.three.setCheckable(True)
+        self.three.setChecked(False)
+        self.three.clicked.connect(self.toggle_three)
         processing_layout.addWidget(self.GPU)
+        processing_layout.addWidget(self.two)
+        processing_layout.addWidget(self.three)
         processing_group.setLayout(processing_layout)
 
         # Group 3: Training Options
@@ -6220,6 +6245,7 @@ class MachineWindow(QMainWindow):
         segmentation_group = QGroupBox("Segmentation")
         segmentation_layout = QVBoxLayout()
         seg_button = QPushButton("Preview Segment")
+        self.seg_button = seg_button
         seg_button.clicked.connect(self.start_segmentation)
         full_button = QPushButton("Segment All")
         full_button.clicked.connect(self.segment)
@@ -6237,9 +6263,12 @@ class MachineWindow(QMainWindow):
         self.setCentralWidget(main_widget)
 
         self.trained = False
+        self.previewing = False
 
 
         self.segmenter = segmenter.InteractiveSegmenter(active_data, use_gpu=True)
+        self.segmentation_worker = None
+
 
 
     def toggle_GPU(self):
@@ -6247,6 +6276,27 @@ class MachineWindow(QMainWindow):
 
         self.use_gpu = self.GPU.isChecked()
 
+    def toggle_two(self):
+        if self.two.isChecked():
+            # If button two is checked, ensure button three is unchecked
+            self.three.setChecked(False)
+            self.use_two = True
+        else:
+            # If button three is checked, ensure button two is unchecked
+            self.three.setChecked(True)
+            self.use_two = False
+
+    def toggle_three(self):
+        if self.three.isChecked():
+            # If button two is checked, ensure button three is unchecked
+            self.two.setChecked(False)
+            self.use_two = False
+        else:
+            # If button three is checked, ensure button two is unchecked
+            self.two.setChecked(True)
+            self.use_two = True
+
+
 
     def toggle_foreground(self):
 
@@ -6295,15 +6345,29 @@ class MachineWindow(QMainWindow):
         self.kill_segmentation()
         # Wait a bit for cleanup
         time.sleep(0.1)
-        self.segmenter.train_batch(self.parent().channel_data[2], speed = speed, use_gpu = self.use_gpu)
-        self.trained = True
+        if not self.use_two:
+            self.previewing = False
+        try:
+            self.segmenter.train_batch(self.parent().channel_data[2], speed = speed, use_gpu = self.use_gpu, use_two = self.use_two)
+            self.trained = True
+        except MemoryError:
+            QMessageBox.critical(
+                self,
+                "Alert",
+                "Out of memory computing feature maps. Note these for 3D require 7x the RAM of the active image (or 9x for the detailed map).\n Please use 2D slice models if you do not have enough RAM."
+            )
+
 
     def start_segmentation(self):
 
         self.kill_segmentation()
         time.sleep(0.1)
 
-        print("Beginning new segmentation...")
+        if self.use_two:
+            self.previewing = True
+        else:
+            print("Beginning new segmentation...")
+
 
         if self.parent().active_channel == 0:
             if self.parent().channel_data[0] is not None:
@@ -6317,7 +6381,7 @@ class MachineWindow(QMainWindow):
         if not self.trained:
             return
         else:
-            self.segmentation_worker = SegmentationWorker(self.parent().highlight_overlay, self.segmenter, self.use_gpu)
+            self.segmentation_worker = SegmentationWorker(self.parent().highlight_overlay, self.segmenter, self.use_gpu, self.use_two, self.previewing, self)
             self.segmentation_worker.chunk_processed.connect(self.update_display)  # Just update display
             self.segmentation_worker.finished.connect(self.segmentation_finished)
             current_xlim = self.parent().ax.get_xlim()
@@ -6345,10 +6409,42 @@ class MachineWindow(QMainWindow):
         msg.setStandardButtons(QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No)
         return msg.exec() == QMessageBox.StandardButton.Yes
 
+
+
+    def check_for_z_change(self):
+        current_z = self.parent().current_slice
+        if not hasattr(self, '_last_z'):
+            self._last_z = current_z
+            return False
+        
+        changed = (self._last_z != current_z)
+        self._last_z = current_z
+        
+        if changed and self.previewing and self.segmentation_worker is not None:
+            self.segmentation_worker.stop()
+            time.sleep(0.1)
+            
+            # Force regeneration of chunks
+            self.segmenter.realtimechunks = None
+            
+            # Restart the worker
+            self.start_segmentation()
+            
+        return changed
+
     def update_display(self):
         if not hasattr(self, '_last_update'):
             self._last_update = 0
         
+        current_z = self.parent().current_slice
+        if not hasattr(self, '_last_z'):
+            self._last_z = current_z
+
+        self._last_z = current_z
+
+        if self.previewing:
+            changed = self.check_for_z_change()
+
         current_time = time.time()
         if current_time - self._last_update >= 1:  # Match worker's interval
             try:
@@ -6366,19 +6462,84 @@ class MachineWindow(QMainWindow):
             except Exception as e:
                 print(f"Display update error: {e}")
 
+    def poke_segmenter(self):
+        if self.use_two and self.previewing:
+            try:
+                # Clear any processing flags in the segmenter
+                if hasattr(self.segmenter, '_currently_processing'):
+                    self.segmenter._currently_processing = None
+                    
+                # Force regenerating the worker
+                if self.segmentation_worker is not None:
+                    self.kill_segmentation()
+                    
+                time.sleep(0.2)
+                self.start_segmentation()
+                
+            except Exception as e:
+                print(f"Error in poke_segmenter: {e}")
+                import traceback
+                traceback.print_exc()
+
     def segmentation_finished(self):
-        print("Segmentation completed")
+        if not self.use_two:
+            print("Segmentation completed")
+        
         current_xlim = self.parent().ax.get_xlim()
         current_ylim = self.parent().ax.get_ylim()
         self.parent().update_display(preserve_zoom=(current_xlim, current_ylim))
+        
+        # Store the current z position before killing the worker
+        current_z = self.parent().current_slice
+        
+        # Clean up the worker
         self.kill_segmentation()
+        self.segmentation_worker = None
         time.sleep(0.1)
+        
+        # Auto-restart for 2D preview mode only if certain conditions are met
+        if self.previewing and self.use_two:
+            # Track when this slice was last processed
+            if not hasattr(self, '_processed_slices'):
+                self._processed_slices = {}
+                
+            current_time = time.time()
+            
+            # Check if we've recently tried to process this slice (to prevent loops)
+            recently_processed = False
+            if current_z in self._processed_slices:
+                time_since_last_attempt = current_time - self._processed_slices[current_z]
+                recently_processed = time_since_last_attempt < 5.0  # 5 second cooldown
+                
+            if not recently_processed:
+                self._processed_slices[current_z] = current_time
+                
+                # Reset any processing flags in the segmenter
+                if hasattr(self.segmenter, '_currently_processing'):
+                    self.segmenter._currently_processing = None
+                    
+                # Create a new worker after a brief delay
+                QTimer.singleShot(500, self.start_segmentation)
+
 
 
     def kill_segmentation(self):
-        if hasattr(self, 'segmentation_worker'):
+        if hasattr(self, 'segmentation_worker') and self.segmentation_worker is not None:
+            # Signal the thread to stop
             self.segmentation_worker.stop()
-            del self.segmentation_worker  # Clean up reference
+            
+            # Wait for the thread to finish
+            if self.segmentation_worker.isRunning():
+                self.segmentation_worker.wait(1000)  # Wait up to 1 second
+                
+                # If thread is still running after timeout, try to force termination
+                if self.segmentation_worker.isRunning():
+                    self.segmentation_worker.terminate()
+                    self.segmentation_worker.wait()  # Wait for it to be terminated
+            
+            # Now safe to delete
+            del self.segmentation_worker
+            self.segmentation_worker = None
 
 
     def segment(self):
@@ -6391,6 +6552,8 @@ class MachineWindow(QMainWindow):
             self.kill_segmentation()
             time.sleep(0.1)
 
+            self.previewing = False
+
             if self.parent().active_channel == 0:
                 if self.parent().channel_data[0] is not None:
                     active_data = self.parent().channel_data[0]
@@ -6406,8 +6569,12 @@ class MachineWindow(QMainWindow):
             # Clean up when done
             self.segmenter.cleanup()
 
-        for z,y,x in foreground_coords:
-            self.parent().highlight_overlay[z,y,x] = 255
+        fg_array = np.array(list(foreground_coords))
+        if len(fg_array) > 0:  # Check if we have any foreground coordinates
+            # Unpack into separate coordinate arrays
+            z_coords, y_coords, x_coords = fg_array[:, 0], fg_array[:, 1], fg_array[:, 2]
+            # Assign values in a single vectorized operation
+            self.parent().highlight_overlay[z_coords, y_coords, x_coords] = 255
 
         self.parent().load_channel(3, self.parent().highlight_overlay, True)
 
@@ -6420,20 +6587,26 @@ class MachineWindow(QMainWindow):
 
         self.parent().update_display()
 
+        self.previewing = False
+
         print("Finished segmentation moved to Overlay 2. Use File -> Save(As) for disk saving.")
 
     def closeEvent(self, event):
         if self.parent().isVisible():
             if self.confirm_close_dialog():
-
+                # Clean up resources before closing
                 if self.brush_button.isChecked():
                     self.silence_button()
                     self.toggle_brush_mode()
+                
                 self.parent().pen_button.setEnabled(True)
                 self.parent().brush_mode = False
-                self.parent().machine_window = None
+                
+                # Kill the segmentation thread and wait for it to finish
                 self.kill_segmentation()
-                time.sleep(0.1)
+                time.sleep(0.2)  # Give additional time for cleanup
+                
+                self.parent().machine_window = None
             else:
                 event.ignore()
 
@@ -6444,48 +6617,84 @@ class SegmentationWorker(QThread):
     finished = pyqtSignal()
     chunk_processed = pyqtSignal()
     
-    def __init__(self, highlight_overlay, segmenter, use_gpu):
+    def __init__(self, highlight_overlay, segmenter, use_gpu, use_two, previewing, machine_window):
         super().__init__()
         self.overlay = highlight_overlay
         self.segmenter = segmenter
         self.use_gpu = use_gpu
+        self.use_two = use_two
+        self.previewing = previewing
+        self.machine_window = machine_window
         self._stop = False
         self.update_interval = 1  # Increased to 500ms
         self.chunks_since_update = 0
         self.chunks_per_update = 5  # Only update every 5 chunks
+        self.poked = False # If it should wake up or not
         self.last_update = time.time()
         
     def stop(self):
         self._stop = True
+
+    def get_poked(self):
+        self.poked = True
         
     def run(self):
         try:
             self.overlay.fill(False)
             
-            for foreground_coords, background_coords in self.segmenter.segment_volume_realtime(gpu = self.use_gpu):
-                if self._stop:
-                    break
+            # Remember the starting z position
+            self.starting_z = self.segmenter.current_z
+            
+            if self.previewing and self.use_two:
+                # Process current z-slice in chunks
+                current_z = self.segmenter.current_z
+                
+                # Process the slice with chunked generator
+                for foreground, background in self.segmenter.segment_slice_chunked(current_z):
+                    if self._stop:
+                        break
                     
-                for z,y,x in foreground_coords:
-                    self.overlay[z,y,x] = 1
-                for z,y,x in background_coords:
-                    self.overlay[z,y,x] = 2
-                
-                # Update only after several chunks AND minimum time interval
-                self.chunks_since_update += 1
-                current_time = time.time()
-                if (self.chunks_since_update >= self.chunks_per_update and 
-                    current_time - self.last_update >= self.update_interval):
-                    self.chunk_processed.emit()
-                    self.chunks_since_update = 0
-                    self.last_update = current_time
-
+                    # Update the overlay
+                    for z,y,x in foreground:
+                        self.overlay[z,y,x] = 1
+                    for z,y,x in background:
+                        self.overlay[z,y,x] = 2
+                    
+                    # Signal update after each chunk
+                    self.chunks_since_update += 1
+                    current_time = time.time()
+                    if (self.chunks_since_update >= self.chunks_per_update and 
+                        current_time - self.last_update >= self.update_interval):
+                        self.chunk_processed.emit()
+                        self.chunks_since_update = 0
+                        self.last_update = current_time
+                
+            else:
+                # Original 3D approach
+                for foreground_coords, background_coords in self.segmenter.segment_volume_realtime(gpu=self.use_gpu):
+                    if self._stop:
+                        break
+                    
+                    for z,y,x in foreground_coords:
+                        self.overlay[z,y,x] = 1
+                    for z,y,x in background_coords:
+                        self.overlay[z,y,x] = 2
+                    
+                    self.chunks_since_update += 1
+                    current_time = time.time()
+                    if (self.chunks_since_update >= self.chunks_per_update and 
+                        current_time - self.last_update >= self.update_interval):
+                        self.chunk_processed.emit()
+                        self.chunks_since_update = 0
+                        self.last_update = current_time
+            
             self.finished.emit()
             
         except Exception as e:
             print(f"Error in segmentation: {e}")
-            raise
-        
+            import traceback
+            traceback.print_exc()
+
     def run_batch(self):
         try:
             foreground_coords, _ = self.segmenter.segment_volume()