Program Information
A Novel Multi-Source Multi-Objective (MSMO) Image Fusion Technique for MR-Based Treatment Planning
L Zhang*, F Yin , S Han , B Moore , J Cai , Duke University Medical Center, Durham, NC
Presentations
MO-F-205-2 (Monday, July 31, 2017) 4:30 PM - 6:00 PM Room: 205
Purpose: To develop a multi-objective driven image fusion technique using multiple single-contrast MR scans for MR-based treatment planning.
Methods: T1 weighted (T1-w), T2 weighted (T2-w), diffusion weighted images (DWI), and cine (T2/T1-w) were acquired from liver cancer patients with breath-holding. Digital human phantoms (XCAT) assigned with average of 8-10 patient MR datasets were used for simulation. The MR datasets were fed into the proposed technique. After pre-processing (deformable image registration, de-noising and normalization), the four image-sets were fused with linear weightings from -1 to 1 with 0.2 interval. The fused images, weighting factors and objective metrics (organ CNR, SNR, similarity index to T1-w, etc.) are saved into a fusion space. For different objectives, matching fusion parameters are identified and the average used as starting parameters for future patients. After fine tuning on new patients, fusion space is updated. The proposed technique was tested on XCAT phantoms and liver cancer patients.
Results: For XCAT, with enhancing tumor CNR, liver and vertebral body SNR as objective, the average weighting factors were -0.33, 0.79, 0.3, and 0.53, respectively. With negative tumor contrast as objective, the factors were 0.85, 0.53, -0.39 and -0.8, respectively. For liver cancer patients, multiple objectives were achievable. For one patient, tumor was easily observed after fusion when it was not identifiable in source images. Tumor CNR increased from 3.2, 7.7, 0.1, and -4.4 in source images to 15.5 in fused image, with high liver and vertebral body SNR (572.2 and 452.3, respectively). Objectives like organ edge-sharping, bony structure enhancement and semi-synthetic CT imaging were also achieved with different parameters.
Conclusion: A novel multi-source multi-objective MR image fusion technique was developed and examined with both XCAT phantoms and patients. Multiple objectives were achieved by fusing multi-source single-contrast MR images. This proposed technique has high potential to facilitate MR-based treatment planning.
Funding Support, Disclosures, and Conflict of Interest: This project was funded by Varian Medical Systems.
Contact Email: