Skip to main content
. 2020 Oct 16;7(5):055501. doi: 10.1117/1.JMI.7.5.055501

Fig. 1.

Fig. 1

Our segmentation review software running on AWS AppStream 2.0. AppStream allows the developer to run Windows in a virtual machine on AWS and display the output to a remote instance of Google Chrome. Any application that can be installed in Windows can be installed in the virtual machine. We developed our own application in Python 3.6 and QT 5. The program launched two instances of insight segmentation and registration toolkit (ITK)-SNAP (windows top right and top left) to display an MRI exam from the test set along with the manual technician and automatic DL tumor segmentations (red overlays). The order of the display is randomized, and the viewer is blinded to the source of the segmentation. Lesion A is always displayed in the top-left window and lesion B in the top right. The viewer can zoom in and out and move the cursor (crosshairs) to any location in the MRI volume. The two ITK-SNAP instances are synchronized so that they show the same location at all times. The bottom window provides widgets (sliders) that allow the viewer to quickly and easily score the quality of each segmentation. The bottom window also provides widgets that allow the viewer to move forward (or backward) through the MR exams in their assigned group of exams.