Graphical User Interface (GUI) Developmentfor Deep Learning Assisted Algorithm for Catheter Reconstruction During MR-OnlyGynecological Interstitial Brachytherapy
K Guo1*, A Shaaer2, M Paudel3, (1) Sunnybrook Odette Cancer Center, University of Toronto, Toronto, ON, CA, (2) Grand River Hospital, Kitchener, ON, CA, (3) Sunnybrook Odette Cancer Center, University of Toronto, Toronto, ON, CA
Presentations
PO-GePV-T-56 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: During interstitial high-dose-rate (HDR) gynecological brachytherapy, magnetic-resonance-imaging (MRI) offers excellent soft-tissue contrast for contouring. Several catheters are inserted through a standardized template, which is surgically sutured to the patient’s perineum. Current approach, manual reconstruction of the implanted plastic catheters, is time-consuming. In order to speed up this process, we have developed a novel deep-learning-assisted-semi-automatic (DLASA) algorithm to reconstruct interstitial catheters during MR-only interstitial gynecological brachytherapy. The primary aim of this study is to develop a Graphical-User-Interface (GUI) for digitizing catheters and communicate with Oncentra® Brachy treatment- planning-system (TPS).
Methods: A key component for this DLASA algorithm is to localize all catheters at the reference slice which is the slice just before the catheters enter into the template. Those localization information will be written into the input file, as well as free length of each catheter which is used to calculate the physical length of catheters inside the patient. Information in input files are passed to U-Net model to identify all possible catheter markers in a given image slice. Then, the true location of each catheter was tracked by finding the extrema in T1- and T2-weighted MR images. Once digitization complete, the marker locations will be save to xls file,or directly write into Oncentra-TPS-DICOM plan file. The 3D viewer of reconstructed catheter slices is available in GUI to QA catheter locations slice by slice.
Results: The GUI is able to digitize the catheters in an average time of 0.55s/catheter/slice for 3 tested patients which is more than 50% time saving compared to conventional manual process.
Conclusion: The adoption of this GUI has potential to improve treatment efficiency by reducing planning time, clinic resources, and manual selection errors in current workflow. In future, we need to examine GUI with more patients, and test GUI in association with TPS for further evaluation.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email