WE-G-18C-04: Comparison of Image Registration Strategies to Improve DCE-MRI Uptake Curves in Esophageal Cancer

Authors


Abstract

Purpose:

Neoadjuvant chemoradiotherapy (nCRT) is often used to treat resectable esophageal cancer. In 29% of the patients, nCRT results in a pathologic complete reponse (pCR). pCR is an important prognostic factor and surgery can potentially be omitted in this patient group. We will study the potential of DCE-MRI for accurate response assessment in those patients. In preparation to this study, motion compensation by retrospective image registration strategies of DCE-MRI time series was investigated to enable correct pharmacokinetic analysis required for response assessment.

Methods:

Five patients with biopsy-proven esophageal cancer received a DCE-MRI scan (62 time-frames, 3 s/frame) prior to neoadjuvant therapy; contrast was injected after the 10th time-frame. Three different registration strategies were reviewed; registration to: (1) 1st time-frame, (2) 41th timeframe, and (3) previous time-frame; rigid or affine transformations were allowed (using Elastix toolbox). We chose these time-points as they occur well before (1) or after (2) contrast uptake. In strategy (3), the registration process is thought to be less susceptible to image intensity variations caused by contrast uptake. The effectiveness of the registration strategies was measured with SSIM (structure component of Structural SIMilarity, 1 = perfect match) metric within the volume encompassing the tumor. The 41th frame was chosen as reference image.

Results:

The average SSIM over all time-frames and patients was: pre-registration: 0.67, post-registration rigid to 1st/41th/previous frame: 0.68/0.70/0.66, affine: 0.68/0.69/0.66. For rigid registration to frame 41, the minimum and maximum SSIM change per time-frame ranged from −0.09 to 0.14.

Conclusion:

In this analysis, rigid registration to a time-frame after contrast-enhancement led to the best retrospective motion compensation of DCE-MRI in esophageal cancer measured by SSIM. After verification in a larger patient group, this correction will be used prior to semi-quantitative analysis and pharmacokinetic modeling to assess response to neoadjuvant therapy.

Ancillary