Merge branch 'master' into master

.gitattributes

@@ -10,3 +10,4 @@ test/assets/* filter=lfs diff=lfs merge=lfs -text
 *.mhd filter=lfs diff=lfs merge=lfs -text
 *.raw filter=lfs diff=lfs merge=lfs -text
 *.ckpt filter=lfs diff=lfs merge=lfs -text
+

interface/backend/api/serializers.py

@@ -1,3 +1,7 @@
+import dicom
+import base64
+from PIL import Image
+from io import BytesIO
 from backend.cases.models import (
     Case,
     Candidate,
@@ -66,3 +70,71 @@ def create(self, validated_data):
             candidate=validated_data['candidate'],
             centroid=ImageLocation.objects.create(**validated_data['centroid']),
         )
+
+class DicomMetadataSerializer(serializers.BaseSerializer):
+    '''
+    Serialize a Dicom image metadata including a base64 version of the
+    image in following format:
+    {
+        metadata: {
+            "Specific Character Set": "ISO_IR 100",
+            "Image Type": [
+                "ORIGINAL",
+                "SECONDARY",
+                "AXIAL"
+            ],
+            "SOP Class UID": "1.2.840.10008.5.1.4.1.1.2",
+            ......
+        },
+        image: "data:image/jpg;base64,/9j/4AAQSkZJRgABAQ.....fnkw3n"
+    }
+    '''
+
+    def to_representation(self, obj):
+        '''
+        Put dicom metadata into a separate dictionary
+        '''
+        dicom_dict = {}
+        repr(obj)   # Bit hacky! But does the work to populate the elements
+        for dicom_value in obj.values():
+            if dicom_value.tag == (0x7fe0, 0x0010):
+                # discard pixel data
+                continue
+            if isinstance(dicom_value.value, dicom.dataset.Dataset):
+                dicom_dict[dicom_value.name] = self.dicom_dataset_to_dict(dicom_value.value)
+            else:
+                dicom_dict[dicom_value.name] = self._convert_value(dicom_value.value)
+        return {
+            'metadata': dicom_dict,
+            'image': self.dicom_to_base64(obj),
+        }
+
+    def dicom_to_base64(self, ds):
+        '''
+        Returning base64 encoded string for a dicom image
+        '''
+        buff_output = BytesIO()
+        img = Image.fromarray((ds.pixel_array)).convert('RGB')
+        img.save(buff_output, format='jpeg')
+        return 'data:image/jpg;base64,' + \
+                base64.b64encode(buff_output.getvalue()).decode()
+
+    def _sanitise_unicode(self, s):
+        return s.replace(u"\u0000", "").strip()
+
+    def _convert_value(self, v):
+        if isinstance(v, (list, int, float)):
+            converted_val = v
+        elif isinstance(v, str):
+            converted_val = self._sanitise_unicode(v)
+        elif isinstance(v, bytes):
+            converted_val = self._sanitise_unicode(v.decode('ascii', 'replace'))
+        elif isinstance(v, dicom.valuerep.DSfloat):
+            converted_val = float(v)
+        elif isinstance(v, dicom.valuerep.IS):
+            converted_val = int(v)
+        elif isinstance(v, dicom.valuerep.PersonName3):
+            converted_val = str(v)
+        else:
+            converted_val = repr(v)
+        return converted_val

interface/backend/api/tests.py

@@ -41,6 +41,15 @@ def test_images_available_view(self):
         url = reverse('images-available')
         response = self.client.get(url)
         self.assertEqual(response.status_code, status.HTTP_200_OK)
+
+    def test_images_metadata_view(self):
+        url = reverse('images-metadata')
+        response = self.client.get(url, {
+            'dicom_location': '/images/LIDC-IDRI-0002/' \
+            '1.3.6.1.4.1.14519.5.2.1.6279.6001.490157381160200744295382098329/' \
+            '1.3.6.1.4.1.14519.5.2.1.6279.6001.619372068417051974713149104919/-80.750000.dcm'
+        })
+        self.assertEqual(response.status_code, status.HTTP_200_OK)
 
     def test_candidates_mark(self):
         candidate = CandidateFactory()

interface/backend/api/urls.py

@@ -4,6 +4,7 @@
     NoduleViewSet,
     ImageSeriesViewSet,
     ImageAvailableApiView,
+    ImageMetadataApiView,
     candidate_mark,
     candidate_dismiss,
     case_report,
@@ -26,6 +27,7 @@
     url(r'^', include(router.urls)),
     url(r'^api-auth/', include('rest_framework.urls', namespace='rest_framework')),
     url(r'^images/available$', ImageAvailableApiView.as_view(), name='images-available'),
+    url(r'^images/metadata$', ImageMetadataApiView.as_view(), name='images-metadata'),
     url(r'^candidates/(?P<candidate_id>\d+)/dismiss$', candidate_dismiss, name='candidate-dismiss'),
     url(r'^candidates/(?P<candidate_id>\d+)/mark$', candidate_mark, name='candidate-mark'),
     url(r'^nodules/(?P<nodule_id>\d+)/update$', nodule_update, name='nodule-update'),

interface/backend/api/views.py

@@ -1,14 +1,16 @@
 import json
 import mimetypes
 import os
-
+import dicom
 from backend.api import serializers
 from backend.cases.models import (
     Case,
     Candidate,
     Nodule,
     CaseSerializer
 )
+from rest_framework.views import APIView
+from rest_framework.response import Response
 from backend.images.models import ImageSeries
 from django.conf import settings
 from django.core.files.storage import FileSystemStorage
@@ -20,7 +22,6 @@
 from rest_framework.decorators import renderer_classes
 from rest_framework.renderers import JSONRenderer
 from rest_framework.response import Response
-from rest_framework.views import APIView
 
 
 class CaseViewSet(viewsets.ModelViewSet):
@@ -43,6 +44,24 @@ class ImageSeriesViewSet(viewsets.ModelViewSet):
     serializer_class = serializers.ImageSeriesSerializer
 
 
+class ImageMetadataApiView(APIView):   
+
+    def get(self, request):
+        '''
+        Get metadata of a DICOM image including the image in base64 format.
+        Example: .../api/images/metadata?dicom_location=FULL_PATH_TO_IMAGE
+        ---
+        parameters:
+            - name: dicom_location
+            description: full location of the image
+            required: true
+            type: string
+        '''
+        path = request.GET['dicom_location']
+        ds = dicom.read_file(path, force=True)
+        return Response(serializers.DicomMetadataSerializer(ds).data)
+
+
 class ImageAvailableApiView(APIView):
     """
     View list of images from dataset directory

interface/requirements/base.txt

@@ -9,4 +9,8 @@ django-dotenv==1.4.1
 django-environ==0.4.3
 
 # Images
+Pillow==4.2.1
 pydicom==0.9.9
+
+# Data
+numpy==1.13.1

prediction/requirements/base.txt

@@ -10,4 +10,4 @@ scikit-image==0.13.0
 SimpleITK==1.0.1
 http://download.pytorch.org/whl/cu80/torch-0.2.0.post3-cp36-cp36m-manylinux1_x86_64.whl
 opencv-python==3.3.0.10
-pandas==0.20.3
+pandas==0.20.3

prediction/src/algorithms/classify/trained_model.py

@@ -6,7 +6,6 @@
     An API for a trained classification model to make predictions
     for if nodules are concerning or not.
 """
-
 from src.algorithms.classify.src import gtr123_model
 from src.preprocess.load_ct import load_ct, MetaData
 
@@ -64,6 +63,7 @@ def predict(dicom_path, centroids, model_path=None,
     else:
         preprocessed = voxel_data
 
+
     model_path = model_path or "src/algorithms/classify/assets/gtr123_model.ckpt"
 
     return gtr123_model.predict(preprocessed, centroids, model_path)

prediction/src/algorithms/identify/trained_model.py

@@ -11,8 +11,6 @@
 
 from src.algorithms.identify.src import gtr123_model
 
-
-
 def predict(dicom_path):
     """ Predicts centroids of nodules in a DICOM image.
 
@@ -37,7 +35,6 @@ def predict(dicom_path):
              'z': int,
              'p_nodule': float}
     """
-
     reader = sitk.ImageSeriesReader()
     filenames = reader.GetGDCMSeriesFileNames(dicom_path)
     if not filenames:
@@ -47,4 +44,4 @@ def predict(dicom_path):
     image = reader.Execute()
     result = gtr123_model.predict(image)
     return result
-   
+

prediction/src/tests/test_classify_trained_model_predict.py

@@ -21,14 +21,14 @@ def test_classify_predict_model_load(dicom_path, model_path):
                                       preprocess_ct=None,
                                       preprocess_model_input=None)
 
+
     assert len(predicted) == 0
 
 
 def test_classify_predict_inference(dicom_path, model_path):
     predicted = trained_model.predict(dicom_path,
                                       [{'x': 50, 'y': 50, 'z': 21}],
                                       model_path)
-
     assert len(predicted) == 1
     assert isinstance(predicted[0]['p_concerning'], float)
     assert predicted[0]['p_concerning'] >= 0.

prediction/src/tests/test_endpoints.py

@@ -5,12 +5,11 @@
     Provides unit tests for the API endpoints.
 """
 
+import json
 import os
 from functools import partial
-import json
 
 import pytest
-
 from flask import url_for
 from src.algorithms import classify, identify, segment
 from src.factory import create_app
@@ -172,3 +171,4 @@ def test_other_error(client):
     data = get_data(r)
     assert r.status_code == 500
     assert "The path doesn't contain neither .mhd nor .dcm files" in data['error']
+