Merge pull request #12 from adamltyson/injectionsite

Refactor injection site finder

.gitignore

@@ -117,3 +117,7 @@ venv.bak/
 .idea/
 .vs/
 *.~lock.*
+
+
+# macOS
+*.DS_Store

neuro/injection_finder/extraction.py

@@ -1,50 +1,54 @@
 import os
 import numpy as np
+from pathlib import Path
 
 from skimage.filters import gaussian as gaussian_filter
 from skimage.filters import threshold_otsu
 from skimage import measure
 
 from brainio import brainio
+from imlib.IO.surfaces import marching_cubes_to_obj
+from imlib.image.orient import reorient_image
 
-from registration import get_registered_image
-from utils import reorient_image, marching_cubes_to_obj, get_largest_component
-from parsers import extraction_parser
+from neuro.injection_finder.registration import get_registered_image
+from neuro.injection_finder.parsers import extraction_parser
 
-# For logging
+import neuro as package_for_log
 import logging
 from fancylog import fancylog
-import fancylog as package
 
 
 class Extractor:
     def __init__(
         self,
         img_filepath,
         registration_folder,
-        logging,
         overwrite=False,
         gaussian_kernel=2,
         percentile_threshold=99.95,
         threshold_type="otsu",
         obj_path=None,
         overwrite_registration=False,
     ):
-
         """
-            Extractor processes a downsampled.nii image to extract the location of the injection site.
-            This is done by registering the image to the allen CCF, blurring, thresholding and finally a 
-            marching cube algorithm to extract the surface of the injection site. 
-
-            :param img_filepath: str, path to .nii file
-            :param registration_folder: str, path to the registration folder [from cellfinder or amap]
-            :param logging: instance of fancylog logger
-            :param overwrite: bool, if False it will avoid overwriting files
-            :gaussian_kernel: float, size of kernel used for smoothing
-            :param percentile_threshold: float, in range [0, 1] percentile to use for thresholding
-            :param threshold_type: str, either ['otsu', 'percentile'], type of threshold used
-            :param obj_path: path to .obj file destination. 
-            :param overwrite_registration: if false doesn't overwrite the registration step
+        Extractor processes a downsampled.nii image to extract the location of
+        the injection site.
+        This is done by registering the image to the allen CCF, blurring,
+        thresholding and finally a marching cube algorithm to extract the
+        surface of the injection site.
+
+        :param img_filepath: str, path to .nii file
+        :param registration_folder: str, path to the registration folder
+        [from cellfinder or amap]
+        :param overwrite: bool, if False it will avoid overwriting files
+        :gaussian_kernel: float, size of kernel used for smoothing
+        :param percentile_threshold: float, in range [0, 1] percentile to use
+        for thresholding
+        :param threshold_type: str, either ['otsu', 'percentile'],
+        type of threshold used
+        :param obj_path: path to .obj file destination.
+        :param overwrite_registration: if false doesn't overwrite the
+        registration step
         """
 
         # Get arguments
@@ -72,14 +76,14 @@ def setup(self):
             self.img_filepath.split(".")[0] + "_thresholded.nii"
         )
 
-        # Get path to obj file and check if it existsts
+        # Get path to obj file and check if it exists
         if self.obj_path is None:
             self.obj_path = self.img_filepath.split(".")[0] + ".obj"
 
         if os.path.isfile(self.obj_path) and not self.overwrite:
             self.logging.warning(
-                "A file exists already at {}. \
-                        Analysis will not run as overwrite is set disabled".format(
+                "A file exists already at {}."
+                "Analysis will not run as overwrite is set disabled".format(
                     self.obj_path
                 )
             )
@@ -88,7 +92,6 @@ def setup(self):
         image = get_registered_image(
             self.img_filepath,
             self.registration_folder,
-            self.logging,
             overwrite=self.overwrite_registration,
         )
         return image
@@ -103,12 +106,12 @@ def extract(self, image, voxel_size=10):
 
         # Gaussian filter
         kernel_shape = [self.gaussian_kernel, self.gaussian_kernel, 6]
-        filtered = gaussian_filter(image, kernel_shape)
+        image = gaussian_filter(image, kernel_shape)
         self.logging.info("Filtering completed")
 
         # Thresholding
         if self.threshold_type.lower() == "otsu":
-            thresh = threshold_otsu(filtered)
+            thresh = threshold_otsu(image)
             self.logging.info(
                 "Thresholding with {} threshold type".format(
                     self.threshold_type
@@ -119,21 +122,20 @@ def extract(self, image, voxel_size=10):
             self.threshold_type.lower() == "percentile"
             or self.threshold_type.lower() == "perc"
         ):
-            thresh = np.percentile(filtered.ravel(), self.percentile_threshold)
+            thresh = np.percentile(image.ravel(), self.percentile_threshold)
             self.logging.info(
-                "Thresholding with {} threshold type. {}th percentile [{}]".format(
+                "Thresholding with {} threshold type. "
+                "{}th percentile [{}]".format(
                     self.threshold_type, self.percentile_threshold, thresh
                 )
             )
         else:
-            raise valueError(
+            raise ValueError(
                 "Unrecognised thresholding type: " + self.threshold_type
             )
 
-        binary = filtered > thresh
-        oriented_binary = reorient_image(
-            binary, invert_axes=[2,], orientation="coronal"
-        )
+        binary = image > thresh
+        binary = keep_n_largest_objects(binary)
 
         # Save thresholded image
         if not os.path.isfile(self.thresholded_savepath) or self.overwrite:
@@ -144,10 +146,14 @@ def extract(self, image, voxel_size=10):
             )
             brainio.to_nii(binary.astype(np.int16), self.thresholded_savepath)
 
+        binary = reorient_image(
+            binary, invert_axes=[2,], orientation="coronal"
+        )
+
         # apply marching cubes
         self.logging.info("Extracting surface from thresholded image")
         verts, faces, normals, values = measure.marching_cubes_lewiner(
-            oriented_binary, 0, step_size=1
+            binary, 0, step_size=1
         )
 
         # Scale to atlas spacing
@@ -159,32 +165,73 @@ def extract(self, image, voxel_size=10):
         faces = faces + 1
         marching_cubes_to_obj((verts, faces, normals, values), self.obj_path)
 
-        # Keep only the largest connected component
-        get_largest_component(self.obj_path)
+
+def keep_n_largest_objects(numpy_array, n=1, connectivity=None):
+    """
+    Given an input binary numpy array, return a "clean" array with only the
+    n largest connected components remaining
+
+    Inspired by stackoverflow.com/questions/47540926
+
+    TODO: optimise
+
+    :param numpy_array: Binary numpy array
+    :param n: How many objects to keep
+    :param connectivity: Labelling connectivity (see skimage.measure.label)
+    :return: "Clean" numpy array with n largest objects
+    """
+
+    labels = measure.label(numpy_array, connectivity=connectivity)
+    assert labels.max() != 0  # assume at least 1 CC
+    n_largest_objects = get_largest_non_zero_object(labels)
+    if n > 1:
+        i = 1
+        while i < n:
+            labels[n_largest_objects] = 0
+            n_largest_objects += get_largest_non_zero_object(labels)
+            i += 1
+    return n_largest_objects
+
+
+def get_largest_non_zero_object(label_image):
+    """
+    In a labelled (each object assigned an int) numpy array. Return the
+    largest object with a value >= 1.
+    :param label_image: Output of skimage.measure.label
+    :return: Boolean numpy array or largest object
+    """
+    return label_image == np.argmax(np.bincount(label_image.flat)[1:]) + 1
 
 
 def main():
     args = extraction_parser().parse_args()
 
     # Get output directory
     if args.output_directory is None:
-        outdir = os.get_cwd()
+        outdir = os.getcwd()
     elif not os.path.isdir(args.output_directory):
         raise ValueError("Output directory invalid")
     else:
         outdir = args.output_directory
 
+    if args.obj_path is None:
+        args.obj_path = Path(args.img_filepath).with_suffix(".obj")
+    else:
+        args.obj_path = Path(args.obj_path)
+
     # Start log
-    log_name = "injection_finder_{}".format(
-        os.path.split(args.registration_folder)[-1]
+    fancylog.start_logging(
+        outdir,
+        package_for_log,
+        filename="injection_finder",
+        verbose=args.debug,
+        log_to_file=args.save_log,
     )
-    fancylog.start_logging(outdir, package, filename=log_name, verbose=True)
 
     # Start extraction
     Extractor(
         args.img_filepath,
         args.registration_folder,
-        logging,
         overwrite=args.overwrite,
         gaussian_kernel=args.gaussian_kernel,
         percentile_threshold=args.percentile_threshold,

neuro/injection_finder/parsers.py

@@ -38,8 +38,8 @@ def extraction_parser():
         "--obj-path",
         dest="obj_path",
         type=str,
-        default=False,
-        help="Path to output .obj file. Optional.",
+        default=None,
+        help="Path to output .obj file. Will default to the image directory.",
     )
 
     parser.add_argument(
@@ -57,7 +57,8 @@ def extraction_parser():
         dest="percentile_threshold",
         type=float,
         default=99.995,
-        help="Float in range [0, 100]. The percentile number of pixel intensity values for tresholding",
+        help="Float in range [0, 100]. The percentile number of pixel "
+        "intensity values for tresholding",
     )
 
     parser.add_argument(
@@ -66,7 +67,8 @@ def extraction_parser():
         dest="threshold_type",
         type=str,
         default="otsu",
-        help="'otsu' or 'percentile'. Determines how the threshold value is computed",
+        help="'otsu' or 'percentile'. Determines how the threshold "
+        "value is computed",
     )
 
     parser.add_argument(
@@ -77,4 +79,17 @@ def extraction_parser():
         default="False",
         help="If false skip running again the registration",
     )
+    parser.add_argument(
+        "--debug",
+        dest="debug",
+        action="store_true",
+        help="Debug mode. Will increase verbosity of logging and save all "
+        "intermediate files for diagnosis of software issues.",
+    )
+    parser.add_argument(
+        "--save-log",
+        dest="save_log",
+        action="store_true",
+        help="Save logging to file (in addition to logging to terminal).",
+    )
     return parser

neuro/injection_finder/registration.py

@@ -1,55 +1,25 @@
 import os
+import logging
 
 from brainio import brainio
-from amap.tools import source_files
-from amap.config.config import get_binary
-from amap.tools.exceptions import RegistrationError
+from imlib.register.niftyreg.transform import run_transform
 
 
-from cellfinder.tools.system import (
-    safe_execute_command,
-    SafeExecuteCommandError,
-)
-
-PROGRAM_NAME = "reg_resample"
 DEFAULT_CONTROL_POINT_FILE = "inverse_control_point_file.nii"
 default_atlas_name = "brain_filtered.nii"
 
 
-def prepare_segmentation_cmd(
-    program_path,
-    floating_image_path,
-    output_file_name,
-    destination_image_filename,
-    control_point_file,
-):
-    cmd = "{} -cpp {} -flo {} -ref {} -res {}".format(
-        program_path,
-        control_point_file,
-        floating_image_path,
-        destination_image_filename,
-        output_file_name,
-    )
-    return cmd
-
-
-def get_registered_image(nii_path, registration_dir, logging, overwrite=False):
-    # get binaries
-    nifty_reg_binaries_folder = source_files.get_niftyreg_binaries()
-    program_path = get_binary(nifty_reg_binaries_folder, PROGRAM_NAME)
-
+def get_registered_image(nii_path, registration_dir, overwrite=False):
     # get file paths
     basedir = os.path.split(nii_path)[0]
     output_filename = os.path.join(
         basedir,
         "{}_transformed.nii".format(os.path.split(nii_path)[1].split(".")[0]),
     )
     if os.path.isfile(output_filename) and not overwrite:
-        run = False
+        logging.info("Skipping registration as output file already exists")
     else:
-        run = True
 
-    if run:
         destination_image = os.path.join(registration_dir, default_atlas_name)
         control_point_file = os.path.join(
             registration_dir, DEFAULT_CONTROL_POINT_FILE
@@ -58,19 +28,14 @@ def get_registered_image(nii_path, registration_dir, logging, overwrite=False):
         log_file_path = os.path.join(basedir, "registration_log.txt")
         error_file_path = os.path.join(basedir, "registration_err.txt")
 
-        reg_cmd = prepare_segmentation_cmd(
-            program_path,
+        logging.info("Running registration")
+        run_transform(
             nii_path,
             output_filename,
             destination_image,
             control_point_file,
+            log_file_path,
+            error_file_path,
         )
-        logging.info("Running registration")
-        try:
-            safe_execute_command(reg_cmd, log_file_path, error_file_path)
-        except SafeExecuteCommandError as err:
-            raise RegistrationError("Registration failed; {}".format(err))
-    else:
-        logging.info("Skipping registration as output file already exists")
 
     return brainio.load_any(output_filename)