Valdo2021 - vascular space segmentation in vascular disease detection challenge (3)

Today, I will share the complete implementation process of the third step of binary segmentation of vascular space , In order to facilitate everyone to learn and understand the whole process , The whole process steps are sorted out , And give the detailed results of the steps . If you are interested, please give it a try .

One 、 Data analysis and preprocessing

First, the effective data of intracranial vascular space area in training is extracted , There are some data without vascular space areas , Not as training data , Only tag values are analyzed here 1, Other labels are 0. In total 40 Example data , There are data of vascular space 22 example .

Analyze this 22 Basic information of example data ： Average image size [243.27777778, 298.5, 168.16666667], Images Spacing Average size [0.63042518,0.63042518,0.85555538], Average size of vascular space [5.05882353,5.62745098,3.70588235]. You can see Vascular space The area of is very small , So using spacing Zoom the original image and Mask The image is unified to （0.3,0.3,0.3）.

Yes Mask By analyzing the connected domain, we can get the boundingbox, With boundingbox Cut out the center of （64,64,64）, Cut out the three modal images （5,95） The mean of 0, The variance of 1 Normalization of .

Data to enhance ： To enhance diversity , Cut the data 5 Times data expansion , Random rotation 30 degree ,x,y,z Random translation 0.1 size , level , Vertical random flip, etc .

Finally, divide the data into training sets , Validation set and test set , The ratio is 80%,10%,10% size .

Two 、 Two segment network

The main body of the network adopts VNet Network of , The loss function is binary dice, The learning rate is 0.001,droupout yes 0.5, The number of iterations is 20epochs,batchsize yes 6. The training data are 1660 example , The optimizer is AdamOptimizer.

3、 ... and 、 Network training and testing

Training loss results and accuracy results

On the test data, a central point coordinate is manually specified , Then the center point （64,64,64） Three modal image regions are segmented , The segmentation results are as follows , The left figure is the result of the gold standard , The picture on the right shows the prediction results .