TotalSegmentator method added in volume.from_nifti()

README.md

@@ -255,3 +255,38 @@ https://github.com/mattmacy/vnet.pytorch
 F. Milletari, N. Navab, S-A. Ahmadi. V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation. arXiv:160604797. 2016.
 
 We gratefully acknowledge the support of the NVIDIA Corporation with the donation of the GPUs used for this research.
+
+## Supplimentary materials
+scripts for virtual environment
+source project_env/bin/activate
+deactivate
+
+Convert the Prostate dataset into the correct format with
+nnUNet_convert_decathlon_task -i /xxx/Task05_Prostate
+Note that Task05_Prostate must be the folder that has the three 'imagesTr', 'labelsTr', 'imagesTs' subfolders!
+
+3D full resoltion U-Net:
+nnUNet_predict -i $nnUNet_raw_data_base/nnUNet_raw_data/Task003_Liver/imagesTs/ -o nnU_OUTPUT_Task03 -t 3 -m 3d_fullres
+! change needed for different cases.
+
+image processing python:
+https://github.com/fitushar/3D-Medical-Imaging-Preprocessing-All-you-need
+
+DECOM to NEFTI:
+/home/qiyuan/Downloads/MRIcroGL/Resources/dcm2niix -f "Ped01-ref" -p y -z y "output_dir" "/media/qiyuan/My Passport/Segmentation/data/Pediatric-CT-SEG/manifest-1645994167898/Pediatric-CT-SEG/Pediatric-CT-SEG-018B687C/10-11-2008-NA-CT-72580/2.000000-RTSTRUCT-58813"
+
+scripts to ssh into lab server:
+ssh [email protected]
+
+scp -r [email protected]:/srv/data1/sean/torso_mid_result/ts_mesh /media/qiyuan/My_Passport/Segmentation/data
+
+scp -r /home/qiyuan/environments/project_env [email protected]:/home/sean/anaconda3/envs
+
+scp -r /home/qiyuan/Downloads/activate [email protected]:/home/sean/anaconda3/envs/project_env/bin
+
+scp -r /home/qiyuan/Documents/run_server.py [email protected]:/home/sean/cis2/nnUNet
+
+/home/qiyuan/Documents
+
+git remote add origin https://github.com/Ding515/3D-CT-Segmentation.git
+git push -u 

ct-org/Readme.md

@@ -0,0 +1,10 @@
+# Instruction for CT-ORG based mask generation
+
+CT-ORG is a 5-classes abdominal organ segmentation model as in [CT-ORG, a new dataset for multiple organ segmentation in computed tomography](https://www.nature.com/articles/s41597-020-00715-8). And the trained network is packaged in docker in [this link](https://github.com/bbrister/ct_organ_seg_docker).
+
+## Mask generation steps
+1. Installing the pre-trained models from [previous link](https://github.com/bbrister/ct_organ_seg_docker).
+2. Running `org_mask_batch.py`.(Currently I/O part still in in file for modification, before publication this should be worked in command line format). 
+3. In some cases the original docker setting up will not work due to the following reasons:
+   - For issues with CPU/GPUs, add `--gpus gpu_index` in ./docker/run_docker_container.py line `sudo docker run --gpus all -v $HOST_SHARED:$CONTAINER_SHARED -t $IMAGE $INFILE $OUTFILE` 
+   - For issues with `IOError: CRC check failed`, this is due to `nibabel` or nii data version, change `get_data` to `get_fdata`.

ct-org/org_mask_batch.py

@@ -0,0 +1,32 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Fri Apr 29 08:28:03 2022
+
+@author: Ding
+"""
+
+import os
+import shutil
+
+data_loading_path = '/home/sean/torso_mid_result/data'
+docker_shared_path = '/home/sean/cis2/ct_organ_seg_docker/shared'
+org_mask_path = '/home/sean/torso_mid_result/org_mask'
+file_list = os.listdir(data_loading_path)
+#os.system('cd ~/cis2/ct_organ_seg_docker')
+existing_file_list = os.listdir(org_mask_path)
+for case_name in file_list:
+    if case_name not in existing_file_list:
+        current_file_path = os.path.join(data_loading_path,case_name)
+        shared_file_path = os.path.join(docker_shared_path,case_name)
+        shutil.copyfile(current_file_path,shared_file_path)
+
+        mask_name = 'seg_'+ case_name
+        command_line = 'sh run_docker_container.sh '+ case_name+' '+mask_name
+        current_process = os.system(command_line)
+        print(current_process)
+        shutil.move(os.path.join(docker_shared_path,mask_name),os.path.join(org_mask_path,case_name))
+        os.remove(shared_file_path)
+
+
+
+

deepdrr/geo/__init__.py

@@ -18,44 +18,71 @@
 
 from .core import (
     HomogeneousObject,
-    HomogeneousPointOrVector,
+    PointOrVector,
     get_data,
     Point,
     Vector,
     Point2D,
     Point3D,
     Vector2D,
     Vector3D,
+    Line,
+    HyperPlane,
+    Line2D,
+    Plane,
+    Line3D,
     point,
     vector,
+    line,
+    plane,
+    p,
+    v,
+    l,
+    pl,
     Transform,
     FrameTransform,
     frame_transform,
     RAS_from_LPS,
     LPS_from_RAS,
 )
+from .exceptions import JoinError, MeetError
 from .camera_intrinsic_transform import CameraIntrinsicTransform
 from .camera_projection import CameraProjection
 from scipy.spatial.transform import Rotation
+from .random import spherical_uniform
 
 __all__ = [
     "HomogeneousObject",
-    "HomogeneousPointOrVector",
+    "PointOrVector",
     "get_data",
     "Point",
     "Point2D",
     "Point3D",
     "Vector",
     "Vector2D",
     "Vector3D",
+    "Line",
+    "HyperPlane",
+    "Line2D",
+    "Plane",
+    "Line3D",
     "point",
     "vector",
+    "line",
+    "plane",
+    "p",
+    "v",
+    "l",
+    "pl",
     "Transform",
     "FrameTransform",
     "frame_transform",
     "RAS_from_LPS",
     "LPS_from_RAS",
+    "JoinError",
+    "MeetError",
     "CameraIntrinsicTransform",
     "CameraProjection",
     "Rotation",
+    "spherical_uniform",
 ]

deepdrr/geo/camera_projection.py

@@ -1,17 +1,12 @@
 from typing import Union, Optional, Any, TYPE_CHECKING
 import numpy as np
 
-from .core import Transform, FrameTransform, point, Point3D, get_data
+from .core import Transform, FrameTransform, point, Point3D, get_data, Plane
 from .camera_intrinsic_transform import CameraIntrinsicTransform
 from ..vol import AnyVolume
 
-# if TYPE_CHECKING:
-#     from ..vol import AnyVolume
-# else:
-#     AnyVolume = Any
 
-
-# TODO(killeen): CameraProjection never calls super().__init__() and thus has no self.data attribute.
+# TODO: reorganize geo so you have primitives.py and transforms.py. Have separate classes for each type of transform?
 
 
 class CameraProjection(Transform):
@@ -24,7 +19,9 @@ def __init__(
         intrinsic: Union[CameraIntrinsicTransform, np.ndarray],
         extrinsic: Union[FrameTransform, np.ndarray],
     ) -> None:
-        """A generic camera projection.
+        """A class for instantiating camera projections.
+
+        The object itself contains the "index_from_world" transform, or P = K[R|t].
 
         A helpful resource for this is:
         - http://wwwmayr.in.tum.de/konferenzen/MB-Jass2006/courses/1/slides/h-1-5.pdf
@@ -47,6 +44,30 @@ def __init__(
             if isinstance(extrinsic, FrameTransform)
             else FrameTransform(extrinsic)
         )
+        index_from_world = self.index_from_camera3d @ self.camera3d_from_world
+        super().__init__(
+            get_data(index_from_world), _inv=get_data(index_from_world.inv)
+        )
+
+    @property
+    def index_from_world(self) -> Transform:
+        return self
+
+    @classmethod
+    def from_krt(
+        cls, K: np.ndarray, R: np.ndarray, t: np.ndarray
+    ) -> "CameraProjection":
+        """Create a CameraProjection from a camera intrinsic matrix and extrinsic matrix.
+
+        Args:
+            K (np.ndarray): the camera intrinsic matrix.
+            R (np.ndarray): the camera extrinsic matrix.
+            t (np.ndarray): the camera extrinsic translation vector.
+
+        Returns:
+            CameraProjection: the camera projection.
+        """
+        return cls(intrinsic=K, extrinsic=FrameTransform.from_rt(K, R, t))
 
     @classmethod
     def from_rtk(
@@ -87,10 +108,6 @@ def index_from_camera3d(self) -> Transform:
     def camera3d_from_index(self) -> Transform:
         return self.index_from_camera3d.inv
 
-    @property
-    def index_from_world(self) -> Transform:
-        return self.index_from_camera3d @ self.camera3d_from_world
-
     @property
     def world_from_index(self) -> Transform:
         """Gets the world-space vector between the source in world and the given point in index space."""
@@ -127,6 +144,9 @@ def get_center_in_world(self) -> Point3D:
             Point3D: the center of the camera in center.
         """
 
+        # TODO: can also get the center from the intersection of three planes formed
+        # by self.data.
+
         world_from_camera3d = self.camera3d_from_world.inv
         return world_from_camera3d(point(0, 0, 0))