Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Gifford_Ryan_Thesis.pdf (2.49 MB)

ETD Abstract Container

Abstract Header

Deep Learning Architecture to Improve Edge Accuracy of Auto-Contouring for Head and Neck Radiotherapy

Gifford, Ryan C
http://orcid.org/0000-0002-1554-9827
http://rave.ohiolink.edu/etdc/view?acc_num=osu1658317931555616

Abstract Details

2022, Master of Science, Ohio State University, Industrial and Systems Engineering.
The manual delineation of the gross tumor volume (GTV) for Head and Neck Cancer (HNC) patients is an essential step in the radiotherapy treatment process. Methods to automate this process have the potential to decrease the amount of time it takes for a clinician to complete a plan, while also decreasing the inter-observer variability between clinicians. Deep learning (DL) methods have shown great promise in auto-segmentation problems. For HNC, we show that DL methods systematically fail at the axial edges of GTV where the segmentation is dependent on both information from the center of the tumor and nearby slices. These failures may decrease trust and usage of proposed Auto-Contouring Systems if not accounted for. In this paper we propose a modified version of the U-Net, a fully convolutional network for image segmentation, which can more accurately process dependence between slices to create a more robust GTV contour. We also show that it can outperform the current proposed methods that capture slice dependencies by leveraging 3D convolutions. Our method uses Convolutional Recurrent Neural Networks throughout the decoder section of the U-Net to capture both spatial and adjacent-slice information when considering a contour. To account for shifts in anatomical structures through adjacent CT slices, we allow an affine transformation to the adjacent feature space using Spatial Transformer Networks. Our proposed model increases accuracy at the edges by 12% inferiorly and 26% superiorly over a baseline 2D U-Net, which has no inherent way to capture information between adjacent slices.
Samantha Krening (Advisor)
Michael Rayo (Committee Co-Chair)
47 p.
Artificial Intelligence; Oncology; Systems Design
Deep Learning, Cognitive Systems, Biomedical Segmentation, Head and Neck Cancer

Recommended Citations

Citations

  • Gifford, R. C. (2022). Deep Learning Architecture to Improve Edge Accuracy of Auto-Contouring for Head and Neck Radiotherapy [Master's thesis, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1658317931555616

    APA Style (7th edition)

  • Gifford, Ryan. Deep Learning Architecture to Improve Edge Accuracy of Auto-Contouring for Head and Neck Radiotherapy. 2022. Ohio State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1658317931555616.

    MLA Style (8th edition)

  • Gifford, Ryan. "Deep Learning Architecture to Improve Edge Accuracy of Auto-Contouring for Head and Neck Radiotherapy." Master's thesis, Ohio State University, 2022. http://rave.ohiolink.edu/etdc/view?acc_num=osu1658317931555616

    Chicago Manual of Style (17th edition)

osu1658317931555616
239
© 2022, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.