MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All Sessions
  4. All Imaging General ePoster Discussion
  5. Pseudo-CBCT Image Prediction of Head and Neck Cancer Patient Using Principal Component Vector Fields of Early Treatment Fractions

Pseudo-CBCT Image Prediction of Head and Neck Cancer Patient Using Principal Component Vector Fields of Early Treatment Fractions

M Nakano1*, T Imae2 , T Nakamoto2 , A Haga3 , K Nawa2 , Y Nomura2 , R Chhatkuli4 , K Demachi5 , W Takahashi2 , K Yamamoto6 , K Nakagawa2 , M Hashimoto7 , Y Yoshioka1 , M Oguchi1 , (1) Cancer Institute Hospital of JFCR, Tokyo, Japan, (2) Department of Radiology, The University of Tokyo Hospital, Tokyo, Japan, (3) University of Tokushima, Tokushima, Japan, (4) National Institute of Radiological Sciences, Chiba, Japan,(5) Faculty of Engineering, The University of Tokyo, Tokyo, Japan, (6) Department of Radiology, Teikyo University Hospital, Tokyo, Japan, (7) Faculty of Allied Health Sciences, Kitasato University, Sagamihara, Japan

Presentations

(Sunday, 7/14/2019) 3:30 PM - 4:00 PM

Room: Exhibit Hall | Forum 9

Purpose: Head and neck cancer patients commonly show large shrinkage of anatomical structures during a course of radiotherapy, and it may require adaptation of radiotherapy plans in the middle of a treatment course. It might be quite beneficial if CT or cone-beam CT (CBCT) image of a patient who would lose weight in the future fractions could be predicted at the timing of early fractions. The present study proposes a method to generate predicted pseudo-CBCT images of late treatment fractions using principal component vector fields (PCVFs) extracted from deformation vector fields (DVFs) of early treatment fractions.

Methods: A series of daily CBCT images of one nasopharyngeal cancer patient treated by 35-fractionated radiotherapy was used. Fourteen DVFs between CBCT images of adjacent fractions through day-1 to day-15 were acquired by B-Spline type deformable image registration using Elastix version 4.7. Thirteen PCVFs were acquired by principal component analysis of 14 DVFs. It was assumed that the predicted DVF from CBCT day-1 to CBCT day-23 (pDVF1to23) could be described as a linear sum of vector fields: (A) 13 PCVFs and an average DVF (DVFave), and (B) 13 PCVFs, DVFave and the DVF from day-1 to day-15. The coefficients in the linear sum were estimated by Adaptive Coordinate Descent algorithm. Two kinds of pseudo-CBCT day-23 were generated as a deformed image of CBCT day-1 using pDVF1to23 predicted using vector field set (A) and (B), and compared using normalized cross correlation (NCC) values.

Results: NCC values between true-CBCT day-23 and pseudo-CBCT day-23 (A) and (B) were 0.987 and 0.992, respectively. Generated pseudo-CBCT image reproduced shrinkage of parotid grands as well as whole neck.

Conclusion: The proposed method successfully predicted pseudo-CBCT images which expressed shrinkage of patient body. Unrealistic deformation of thermoplastic shell and imperfect deformation around patient pillow must be the future work.

Funding Support, Disclosures, and Conflict of Interest: The present study was supported by JSPS KAKENHI Grant Number 16H07429 and 18K15569.

Keywords

Cone-beam CT, Deformation

Taxonomy

IM- Cone Beam CT: 4DCBCT

Contact Email

Additional Links