(Thursday, 7/18/2019) 1:00 PM - 3:00 PM
Room: 304
Purpose: To develop an automated brain metastases (BMs) segmentation platform with atlas labelling for efficient stereotactic radiosurgery treatment planning and treatment follow-up.
Methods: A web-based BMs segmentation platform was implemented on Django web framework, including both client side and back-end server. On client side, JavaScript 3D medical image viewer was adopted for visualizing DICOM and NIFIT formats images. The back-end server runs multiple image processing and segmentation algorithms to accomplish BMs segmentation and labelling, including: 1) skull removing using a learning based ROBEX algorithm; 2) a deep-learning based BMs segmentation algorithm; 3) atlas registration-based BMs labelling. Automatically segmented BMs contours are sent back to web client for reviewing/modification. The finalized contour sets are saved in a DICOM RTStruct format.
Results: We evaluated our developed platform performance on patients having BMs varied from 1-60. The platform takes 4-5 minutes in average to finish segmentation and labelling. Compared to manual segmentation/labelling, this offers substantial clinical time savings and workflow improvement.
Conclusion: A web-based BMs segmentation was developed and evaluated. The developed tool can be a useful tool for assisting radiosurgery treatment planning and treatment follow-up.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
