Merge branch 'master' into prob-unet

.github/workflows/package.yml

@@ -9,6 +9,8 @@ jobs:
   docker:
     runs-on: ubuntu-latest
     steps:
+      - name: Delete huge unnecessary tools folder
+        run: rm -rf /opt/hostedtoolcache
       - name: Checkout
         uses: actions/checkout@v2
       - name: Set up QEMU
@@ -124,3 +126,13 @@ jobs:
           tags: |
             platipy/platipy:nnunet
             ghcr.io/pyplati/platipy:nnunet
+
+      - name: Build and Push TotalSegmentator Image
+        uses: docker/build-push-action@v2
+        with:
+          context: services/totalsegmentator
+          platforms: linux/amd64
+          push: true
+          tags: |
+            platipy/platipy:totalsegmentator
+            ghcr.io/pyplati/platipy:totalsegmentator

CITATION.cff

@@ -1,46 +1,34 @@
-cff-version: 1.2.0
-title: "PlatiPy: Processing Library and Analysis Toolkit for Medical Imaging in Python"
-message: >-
-  If you find the PlatiPy library useful in your research, please consider citing it using these
-  metadata.
-abstract: >-
-  PlatiPy is a library of tools developed in Python for image processing and analysis of medical
-  images. A broad scope of functionality is provided including image conversion, registration
-  and segmentation is provided as well as computation of similarity and radiotherapy dosimtery
-  metrics and much more.
-type: software
+cff-version: "1.2.0"
 authors:
-  - given-names: Phillip
-    family-names: Chlap
-    email: [email protected]
-    affiliation: University of New South Wales
+- family-names: Chlap
+  given-names: Phillip
+  orcid: "https://orcid.org/0000-0002-6517-8745"
+- family-names: Finnegan
+  given-names: Robert N.
+  orcid: "https://orcid.org/0000-0003-4728-8462"
+doi: 10.5281/zenodo.8032858
+message: If you use this software, please cite our article in the
+  Journal of Open Source Software.
+preferred-citation:
+  authors:
+  - family-names: Chlap
+    given-names: Phillip
     orcid: "https://orcid.org/0000-0002-6517-8745"
-  - given-names: Robert
-    family-names: Finnegan
-    email: [email protected]
-    affiliation: University of Sydney
+  - family-names: Finnegan
+    given-names: Robert N.
     orcid: "https://orcid.org/0000-0003-4728-8462"
-license: Apache-2.0
-repository-code: "https://github.com/pyplati/platipy"
-url: "https://pyplati.github.io/platipy/"
-keywords:
-  - "medical image analysis"
-  - "auto-segmentation"
-  - "image registration"
-references:
-  - authors:
-      - family-names: Lowekamp
-        given-names: Bradley C
-      - family-names: Chen
-        given-names: David T
-      - family-names: Ibáñez
-        given-names: Luis
-      - family-names: Blezek
-        given-names: Daniel
-    doi: 10.3389/fninf.2013.00045
-    issn: 16625196
-    journal: "Frontiers in Neuroinformatics"
-    title: "The design of simpleITK"
-    type: article
-    volume: 7
-    year: 2013
+  date-published: 2023-06-26
+  doi: 10.21105/joss.05374
+  issn: 2475-9066
+  issue: 86
+  journal: Journal of Open Source Software
+  publisher:
+    name: Open Journals
+  start: 5374
+  title: "PlatiPy: Processing Library and Analysis Toolkit for Medical
+    Imaging in Python"
+  type: article
+  url: "https://joss.theoj.org/papers/10.21105/joss.05374"
+  volume: 8
+title: "PlatiPy: Processing Library and Analysis Toolkit for Medical
+  Imaging in Python"

README.md

@@ -1,12 +1,15 @@
-# PlatiPy 
+# PlatiPy
+
+[![DOI](https://joss.theoj.org/papers/10.21105/joss.05374/status.svg)](https://doi.org/10.21105/joss.05374)
+
 ## Processing Library and Analysis Toolkit for Medical Imaging in Python
 
 PlatiPy is a library of **amazing** tools for image processing and analysis - designed specifically
-for medical imaging! 
+for medical imaging!
 
 Check out the [PlatiPy documentation](https://pyplati.github.io/platipy/) for more info.
 
-This project was motivated by the need for a simple way to use, visualise, process, and analyse 
+This project was motivated by the need for a simple way to use, visualise, process, and analyse
 medical images. Many of the tools and algorithms are designed in the context of radiation therapy,
 although they are more widely applicable to other fields that use 2D, 3D, or 4D imaging.
 
@@ -16,31 +19,33 @@ We welcome feedback and contributions from the community (yes, you!) and you can
 information about contributing [here](https://pyplati.github.io/platipy/contributing.html).
 
 ## What can I do with **platipy**?
-A lot! A good place to start is by looking in the 
+
+A lot! A good place to start is by looking in the
 [examples directory](https://github.com/pyplati/platipy/tree/master/examples).
 
 Some examples of what PlatiPy can do:
- - DICOM organising and converting:
-    * Bulk convert from multiple series and studies with a single function
-    * Convert DICOM-RT structure and dose filesto NIfTI images
-    * Create DICOM-RT structure files from binary masks e.g. from automatic contouring algorithms
- - Image registration
-    * Register images and transform labels with a few lines of code
-    * Linear transformations: rigid, affine, similarity
-    * Non-linear deformable transformations: demons, b-splines
-    * Multiple metrics for optimisation
- - Atlas-based segmentation
-    * A suite of tools that can be used out-of-the-box
-    * Includes advanced algorithms for 
-      [iterative atlas selection](https://doi.org/10.1088/1361-6560/ab652a/) and 
+
+- DICOM organising and converting:
+  - Bulk convert from multiple series and studies with a single function
+  - Convert DICOM-RT structure and dose files to NIfTI images
+  - Create DICOM-RT structure files from binary masks e.g. from automatic contouring algorithms
+- Image registration
+  - Register images and transform labels with a few lines of code
+  - Linear transformations: rigid, affine, similarity
+  - Non-linear deformable transformations: demons, b-splines
+  - Multiple metrics for optimisation
+- Atlas-based segmentation
+  - A suite of tools that can be used out-of-the-box
+  - Includes advanced algorithms for
+      [iterative atlas selection](https://doi.org/10.1088/1361-6560/ab652a/) and
       [vessel splining](https://doi.org/10.1088/1361-6560/abcb1d/)
- - Synthetic deformation field generation
-    * Simulate anatomically realistic shifts, expansions, and bending
-    * Compare DIR results from clinical systems
- - Basic tools for image processing and analysis
-    * Computing label similarity metrics: DSC, mean distance to agreement, Hausdorff distance, and more
-    * Cropping images to a region of interest
-    * Rotate images and generate maximum/mean intensity projections (beams eye view modelling)
+- Synthetic deformation field generation
+  - Simulate anatomically realistic shifts, expansions, and bending
+  - Compare DIR results from clinical systems
+- Basic tools for image processing and analysis
+  - Computing label similarity metrics: DSC, mean distance to agreement, Hausdorff distance, and more
+  - Cropping images to a region of interest
+  - Rotate images and generate maximum/mean intensity projections (beams eye view modelling)
 
 A major part of this package is **visualisation**, and some examples are shown below!
 
@@ -74,7 +79,7 @@ fig = vis.show()
 ![Figure 2](assets/figure_2.png)
 
 #### Calculate deformation vector fields
-    
+
 ```python
 from platipy.imaging.registration.deformable import fast_symmetric_forces_demons_registration
 
@@ -99,6 +104,7 @@ fig = vis.show()
 ![Figure 3](assets/figure_3.png)
 
 ## Getting started
+
 There aren't many requirements, just an installed Python interpreter (3.7 or greater). PlatiPy can
 be installed with **pip**:
 
@@ -117,5 +123,5 @@ pip install platipy[backend]
 
 ## Authors
 
-* **Phillip Chlap** - [[email protected]]([email protected])
-* **Robert Finnegan** - [[email protected]]([email protected])
+- **Phillip Chlap** - [[email protected]]([email protected])
+- **Robert Finnegan** - [[email protected]]([email protected])

examples/dvh_analysis.ipynb

@@ -23,6 +23,7 @@
     "    !pip install platipy\n",
     "    import platipy\n",
     "\n",
+    "import matplotlib\n",
     "import matplotlib.pyplot as plt\n",
     "import SimpleITK as sitk\n",
     "\n",
@@ -103,7 +104,7 @@
    "source": [
     "vis = ImageVisualiser(ct_image, cut=get_com(structures[\"PTV60\"]))\n",
     "\n",
-    "vis.add_scalar_overlay(dose, discrete_levels=20, colormap=plt.cm.get_cmap(\"inferno\"), name=\"Dose (Gy)\")\n",
+    "vis.add_scalar_overlay(dose, discrete_levels=20, colormap=matplotlib.colormaps.get_cmap(\"inferno\"), name=\"Dose (Gy)\")\n",
     "vis.add_contour(structures)\n",
     "\n",
     "fig = vis.show()"
@@ -160,7 +161,7 @@
     "\n",
     "# Plot the DVH\n",
     "fig, ax = plt.subplots()\n",
-    "plt_dvh.plot(ax=ax, kind=\"line\", colormap=plt.cm.get_cmap(\"rainbow\"), legend=False)\n",
+    "plt_dvh.plot(ax=ax, kind=\"line\", colormap=matplotlib.colormaps.get_cmap(\"rainbow\"), legend=False)\n",
     "\n",
     "# Add labels and show plot\n",
     "plt.legend(loc='best')\n",
@@ -226,8 +227,8 @@
     "    d_cc_points=[10],\n",
     "    structure_for_limits=dose>5,\n",
     "    expansion_for_limits=40,\n",
-    "    contour_cmap=plt.cm.get_cmap(\"rainbow\"),\n",
-    "    dose_cmap=plt.cm.get_cmap(\"inferno\"),\n",
+    "    contour_cmap=matplotlib.colormaps.get_cmap(\"rainbow\"),\n",
+    "    dose_cmap=matplotlib.colormaps.get_cmap(\"inferno\"),\n",
     "    title=\"TCGA_CV_5977 Dose Metrics\")"
    ]
   },

examples/visualise.ipynb

@@ -32,10 +32,11 @@
     "# Function to grab some test data\n",
     "from platipy.imaging.tests.data import get_lung_nifti\n",
     "\n",
-    "# Usual suspects\n",
+    "# Import some common libraries which we need\n",
     "import numpy as np\n",
     "import SimpleITK as sitk\n",
     "from pathlib import Path\n",
+    "import matplotlib\n",
     "import matplotlib.pyplot as plt\n",
     "\n",
     "# Some utilities we use for generating interesting data\n",
@@ -153,7 +154,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "image_visualiser = ImageVisualiser(ct_image, window=(-900, 1100), colormap=plt.cm.get_cmap(\"terrain\"), figure_size_in=6)\n",
+    "image_visualiser = ImageVisualiser(ct_image, window=(-900, 1100), colormap=matplotlib.colormaps.get_cmap(\"terrain\"), figure_size_in=6)\n",
     "\n",
     "image_visualiser.set_limits_from_label(contours[\"LUNG_L\"] + contours[\"LUNG_R\"])\n",
     "\n",
@@ -238,7 +239,7 @@
     "image_visualiser.add_vector_overlay(\n",
     "    dvf_expansion,\n",
     "    name=\"DVF magnitude [mm]\",\n",
-    "    colormap=plt.cm.get_cmap(\"gnuplot2\"),\n",
+    "    colormap=matplotlib.colormaps.get_cmap(\"gnuplot2\"),\n",
     "    alpha=0.75,\n",
     "    arrow_scale=1,\n",
     "    arrow_width=1,\n",
@@ -314,7 +315,7 @@
     "image_visualiser.add_vector_overlay(\n",
     "    dvf_expansion,\n",
     "    name=\"DVF magnitude [mm]\",\n",
-    "    colormap=plt.cm.get_cmap(\"gnuplot2\"),\n",
+    "    colormap=matplotlib.colormaps.get_cmap(\"gnuplot2\"),\n",
     "    alpha=0.5,\n",
     "    arrow_scale=1,\n",
     "    arrow_width=1,\n",

platipy/backend/__init__.py

@@ -65,6 +65,8 @@ def on_celery_setup_logging(**kwargs):
 app = FlaskApp(__name__)
 app.config["SECRET_KEY"] = uuid.uuid4().hex
 
+app.app_context().push()
+
 # Configure SQL Alchemy
 app.config["SQLALCHEMY_DATABASE_URI"] = f"sqlite:///{env_work}/{os.path.basename(os.getcwd())}.db"
 app.config["SQLALCHEMY_TRACK_MODIFICATIONS"] = False

platipy/dicom/io/nifti_to_rtstruct.py

@@ -19,14 +19,17 @@
 import SimpleITK as sitk
 import numpy as np
 import logging
+import matplotlib
 from matplotlib import cm
 
 from rt_utils import RTStructBuilder
 
 logger = logging.getLogger(__name__)
 
 
-def convert_nifti(dcm_path, mask_input, output_file, color_map=cm.get_cmap("rainbow")):
+def convert_nifti(
+    dcm_path, mask_input, output_file, color_map=matplotlib.colormaps.get_cmap("rainbow")
+):
     """Convert a set of masks to a DICOM RTStruct object.
 
     This function now utilises the rt-utils package: https://github.com/qurit/rt-utils
@@ -37,8 +40,8 @@ def convert_nifti(dcm_path, mask_input, output_file, color_map=cm.get_cmap("rain
         mask_input (dict|list): A dictionary containing the name as key and image as value. Or a
                                 list of string with comma separated name and mask paths (name,path)
         output_file (str|pathlib.Path): The path to the file to write the RTStruct
-         color_map (matplotlib.colors.Colormap, optional): Colormap to use for output. Defaults to
-                                                           cm.get_cmap("rainbow").
+        color_map (matplotlib.colors.Colormap, optional): Colormap to use for output. Defaults to
+            matplotlib.colormaps.get_cmap("rainbow").
     """
 
     logger.info("Will convert the following Nifti masks to RTStruct:")
@@ -65,7 +68,6 @@ def convert_nifti(dcm_path, mask_input, output_file, color_map=cm.get_cmap("rain
     rtstruct = RTStructBuilder.create_new(dicom_series_path=str(dcm_series_path))
 
     for mask_name in masks:
-
         # Use a hash of the name to get the color from the supplied color map
         color = color_map(hash(mask_name) % 256)
         color = color[:3]

platipy/imaging/dose/dvh.py

@@ -87,7 +87,7 @@ def calculate_dvh_for_labels(dose_grid, labels, bin_width=0.1, max_dose=None):
         mask_array = sitk.GetArrayFromImage(mask)
 
         # Compute cubic centimetre volume of structure
-        cc = mask_array.sum() * np.product([a / 10 for a in mask.GetSpacing()])
+        cc = mask_array.sum() * np.prod([a / 10 for a in mask.GetSpacing()])
 
         bins, values = calculate_dvh(
             dose_grid, labels[k], bins=np.arange(-bin_width / 2, max_dose + bin_width, bin_width)

platipy/imaging/dose/metric.py

@@ -72,7 +72,7 @@ def calculate_d_to_volume(dose_grid, label, volume, volume_in_cc=False):
     mask_array = sitk.GetArrayFromImage(label)
 
     if volume_in_cc:
-        volume = (volume * 1000 / ((mask_array > 0).sum() * np.product(label.GetSpacing()))) * 100
+        volume = (volume * 1000 / ((mask_array > 0).sum() * np.prod(label.GetSpacing()))) * 100
 
     if volume > 100:
         volume = 100
@@ -106,7 +106,7 @@ def calculate_v_receiving_dose(dose_grid, label, dose_threshold, relative=True):
     if relative:
         return relative_volume
 
-    total_volume = (mask_array > 0).sum() * np.product(label.GetSpacing()) / 1000
+    total_volume = (mask_array > 0).sum() * np.prod(label.GetSpacing()) / 1000
 
     return relative_volume * total_volume
 

platipy/imaging/label/comparison.py

@@ -29,7 +29,7 @@ def compute_volume(label):
         float: The volume (in cubic centimetres)
     """
 
-    return sitk.GetArrayFromImage(label).sum() * np.product(label.GetSpacing()) / 1000
+    return sitk.GetArrayFromImage(label).sum() * np.prod(label.GetSpacing()) / 1000
 
 
 def compute_surface_dsc(label_a, label_b, tau=3.0):
@@ -159,7 +159,7 @@ def compute_volume_metrics(label_a, label_b):
     arr_intersection = arr_a & arr_b
     arr_union = arr_a | arr_b
 
-    voxel_volume = np.product(label_a.GetSpacing()) / 1000.0  # Conversion to cm^3
+    voxel_volume = np.prod(label_a.GetSpacing()) / 1000.0  # Conversion to cm^3
 
     # 2|A & B|/(|A|+|B|)
     dsc = (2.0 * arr_intersection.sum()) / (arr_a.sum() + arr_b.sum())

platipy/imaging/label/fusion.py

@@ -109,7 +109,7 @@ def compute_weight_map(
         view_mask = view_as_windows(arr_mask, window_box_im)
 
         # Flatten to have a list of patches (that are also flattened)
-        new_shape = (np.product(view_target.shape[:3]), np.product(view_target.shape[3:]))
+        new_shape = (np.prod(view_target.shape[:3]), np.prod(view_target.shape[3:]))
         view_target_flat = np.reshape(view_target, new_shape)
         view_moving_flat = np.reshape(view_moving, new_shape)
         view_mask_flat = np.reshape(view_mask, new_shape)