updated to better trained model with correct handling of 8bit images, updating documentation, fixing figures

Author: kmader

CTBoneSegmenter.json

@@ -11,7 +11,7 @@
   "docker":
     {
     "dockerhub_repository": "4quant/bone-segmenter-cpu",
-    "digest": "sha256:41020bcd68cc5854aa1757b452defc52e9035d39ed19edc1b33439d327dd99b5",
+    "digest": "sha256:f1c96761b6a068db89411abbb720b44e5df035c59370ceda0e7c15b3cb3307ac",
     "size": "3.6 GB"
     }
   ,

README.md

@@ -1,6 +1,12 @@
 # Bone Segmenter
 A modification of the Prostate Segmenter library using our bone segmentation routines with CT images
 
+## Slicer Integration
+
+Using the DeepInfer plugin the image can be called interactively on a image loaded in Slicer. Here we show the results using the Chest CT image from the Sample Data
+
+![Slicer](figures/BoneSegGui.png)
+
 # Docker Image
 
 ## CPU
@@ -12,7 +18,7 @@ docker pull 4Quant/bone-segmenter-cpu
 
 ##### Building the docker
 ```
-docker build -t bone-segmenter-cpu -f Dockerfile.cpu .
+docker build -t 4quant/bone-segmenter-cpu -f Dockerfile.cpu .
 ```
 
 ##### Running the docker

boneseg/model.py

@@ -17,7 +17,7 @@ def predict_test(self, input_test_slices_arr):
         print("Preprocessing data.")
         test_slices_arr = self.preprocess(input_test_slices_arr, self.mean_val, self.max_val)
         print("Predicting the labelmap.")
-        test_slices_prediction = self.model.predict(test_slices_arr, verbose=1)
+        test_slices_prediction = self.model.predict(test_slices_arr, verbose=1, batch_size = 4) # reduce memory
         return test_slices_prediction
 
     def train(self, input_train_slices, input_train_segmentations, **kw_args):

figures/BoneSegGui.png

@@ -41,7 +41,7 @@ def main(argv):
     if os.path.isfile(InputVolume) and os.path.isdir(os.path.dirname(OutputLabel)):
         print("Building the model.")
         model = unet1.model(weights=True, summary=False)
-        cnn = CNNModel(model=model, mean_val = 0, max_val = 300)
+        cnn = CNNModel(model=model, mean_val = 127, max_val = 127)
         rows = 128
         cols = 128
         #