The Effects of Reconstruction Filters on CBCT 2D U-Net Liver Segmentation

Michael Garcia      Tyler Kay      Lindsey Bloom      Kian Bagherlee

michael.garciaalcoser@duke.edu     tyler.kay@duke.edu     lindsey.bloom@duke.edu     kb545@duke.edu

Paper PDF

Cone-beam CT (CBCT) is a powerful tool that can help improve the effectiveness of radiation therapy treatments. By using CBCT images, a treatment plan can be adjusted to account for patient anatomy changes, and reduce radiation exposure to organs-at-risk. This work sought to determine the best reconstruction filter to accurately segment the liver in CBCT images using a 2D U-Net architecture, in attempt to reduce overexpoure and ensure healthy treatment. In the end, the Hann filter was identified as producing the most accurate liver segmentation. Link to TIGRE Toolbox used for CBCT reconstruction TIGRE.

Paper: Paper PDF

Code and Data: The following is a link to the project GitHub: CBCT_Segmentation The following is a link to the data used to train/test the model: CT Data