Encrypted login | home

Program Information

A New Penalty Generalizing Structure Tensor for CBCT Reconstruction

S Tan

L Liu1 , Q Shi1 , J Wang2 , S Tan1*, (1) Huazhong University of Science and Technology, Wuhan, Hubei,China (2) UT Southwestern Medical Center, Dallas, TX


SU-I-GPD-I-1 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: To design a new tensor penalty that extends the structure tensor for the cone-beam CT (CBCT) reconstruction.

Methods: Structure tensor total variation (STV), penalizes the Schatten norm of the structure tensor, is suitable for image reconstruction. Thanks to its ability to capture the first-order information around a local neighborhood, the structure tensor can provide more robust features of image than TV and lead to better results in CBCT reconstruction. To achieve richer image feature detection results, we proposed a novel tensor to combine the structure tensor and the second-order information of an image. The proposed tensor generalized existing tensors for image feature detection, including the structure tensor, energy tensor and boundary tensor. Furthermore, we proposed to penalize the Schatten norm of the new tensor. The objective functional was minimized using the gradient descent (GD) method. We evaluated the proposed method on a Compressed Sensing (CS) phantom and a CatPhan 600 phantom. The noise ratio (PSNR), the improvement signal to noise ratio (ISNR), the structural similarity (SSIM) and contrast-to-noise (CNR) were calculated.

Results: For the CS phantom, the reconstructed images using TV had an obvious staircase effect, while those using the proposed method and STV suppressed the staircase effect well. PSNRs, ISNRs, SSIMs of the proposed method were better than the TV and STV penalty. For CatPhan 600, CNR values of the proposed method were similar to those of TV.

Conclusion: The proposed tensor penalty retained favorable properties of STV like suppressing the staircase effect, and had a better reconstruction quality than STV.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by National Natural Science Foundation of China (NNSFC), under Grant Nos. 61375018 and 61672253. J. Wang was supported in part by grants from the Cancer Prevention and Research Institute of Texas (RP130109 and RP110562-P2), the National Institute of Biomedical Imaging and Bioengineering (R01 EB020366).

Contact Email: