MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Intelligent Synthetic CT Generation From CBCT Images Via Deep Learning

L Chen*, X Liang , C Shen , S Jiang , J Wang , UT Southwestern Medical Center, Dallas, TX

Presentations

(Monday, 7/30/2018) 9:30 AM - 10:00 AM

Room: Exhibit Hall | Forum 6

Purpose: Cone-beam CT (CBCT) is scanned daily or weekly (i.e., on-treatment CBCT) for the purpose of accurate patient setup in image-guided radiotherapy (IGRT). However, inaccuracy of CT numbers prevents CBCT from advanced applications such as dose calculation and treatment planning. Motivated by the promising performance of deep learning in medical imaging, we proposed a deep U-net based approach to synthesize CT like images with accurate CT numbers from CBCT while keep the same anatomical structure as on-treatment CBCT.

Methods: We formulate CT synthesis problem under a deep learning framework, where a deep U-net architecture is employed to take advantages of anatomical structure of on-treatment CBCT and image intensity information of planning CT. U-net is chosen as it is capable of making use of both global and local features in image spatial domain, which matches our task of suppressing global scattering artifacts and local artifacts such as noise in the CBCT. In order to train the synthetic CT generation U-net (sCTU-net), we include CBCT and planning CT of 13 patients (12 for training, one for validation) as the input. CT images retaining CBCT anatomical structure are utilized as the corresponding output. To demonstrate the effectiveness of the proposed sCTU-net, we use another four independent patient cases (320 slices) for testing.

Results: We quantitatively compared the resulting synthetic CT with the original CBCT using ground truth CT images as reference. On average, the proposed sCTU-net improved SSIM by ~13% and PSNR by ~5dB. In addition, error in CT number was successfully decreased by 46.14%.

Conclusion: The proposed sCTU-net is capable of synthesizing CT-quality images with accurate CT numbers from on-treatment CBCT and planning CT. This potentially enables advanced applications of CBCT for adaptive treatment planning.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by the Cancer Prevention and Research Institute of Texas (RP160661) and US National Institutes of Health (R01 EB020366).

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email