Main Session
Sep 29
PQA 05 - Physics

3199 - Reduced-Distortion Diffusion Weighted MRI for Cervical Cancer Radiotherapy Treatment Planning

12:30pm - 01:45pm ET
Poster Hall - Exhibit Hall A
Screen: 17
POSTER

Presenter(s)

Victoria Yu, PhD - Memorial Sloan Kettering Cancer Center, New York, New York

V. Y. Yu1, E. Aliotta1, C. Wu1, D. Aramburu Nunez1, A. Moore2, M. A. Kollmeier2, and V. M. Williams2; 1Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, 2Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY

Purpose/Objective(s): Diffusion Weighted Imaging (DWI) has potential to guide radiotherapy treatment planning given its demonstrated ability to predict treatment response. However, clinical translation is limited due to severe geometric distortions in conventional single-shot Echo-Planar-Imaging (ssEPI) techniques. In this study, we evaluated the ability of reduced-distortion DWI techniques to improve cervical cancer tumor delineation.

Materials/Methods: Pre-treatment MRI were obtained in treatment position for 12 women undergoing chemoradiation for cervical cancer enrolled on a prospective clinical trial. Two reduced-distortion DWI sequences: multi-shot EPI (msEPI) and multi-shot Turbo-spin-echo (msTSE), were compared with ssEPI (protocol details in Table 1) as well as a T2-weighted (T2w) TSE scan. For each subject, a tumor contour based on the T2w scan (TumorT2), and tumor subvolumes for each sequence (SubVolssEPI, SubVolmsEPI, SubVolmsTSE) were delineated by a radiation oncologist. Subvolumes were defined as hypointense regions on the ADC maps (from all b values) and hyperintense regions on the high b value images overlapping TumorT2. Subvolumes were compared between sequences using Hausdorff Distance (HD), Mean-Distance-to-Agreement (MDA), and Dice Similarity Coefficient (DSC) metrics.

Results: msEPI qualitatively showed the clearest tumor contrast on the corresponding ADC maps. However, in all but one case, gas adjacent to the tumor resulted in signal pile-up and partial image distortion in both ssEPI and msEPI images. msTSE, although noisier, did not suffer from distortion or artifacts from gas. A one-sided paired t-test showed that SubVolssEPI was significantly smaller than SubVolmsEPI and SubVolmsTSE. Statistics of all volume comparisons are shown in Table 1. msEPI showed the largest difference from ssEPI in all pairwise comparison metrics. Smaller differences were seen in the average volume for msEPI and msTSE. HD and MDA results indicate on average 8-10 mm of maximum difference and mean distance difference of 1.5 mm between subvolumes.

Conclusion: The msTSE sequence was the most robust for tumor delineation as it maintained superior spatial integrity in the unavoidable presence of gas neighboring the cervix which distorted EPI-based sequences. It also identified a larger subvolume than ssEPI, which may impact treatment planning. Clinical significance of observed differences will be examined in future work.

Table 1: sequence properties and summary of volume comparisons
µ ± s

1-sided paired t-test p-value of Volume

HD (mm)

MDA (mm)

DSC

msEPI vs. ssEPI

10.3 ± 4.0

1.5 ± 0.7

0.81 ± 0.1

0.004

msEPI vs. msTSE

8.4 ± 4.0

1.3 ± 0.6

0.84 ± 0.1

0.053

ssEPI vs. msTSE

8.8 ± 2.8

1.2 ± 0.5

0.85 ± 0.1

0.046

Volume (cc): µ ± s

Scan Time (min)

b-values (NSA)

msEPI (2-shot)

50.5 ± 40.1

4:36

0(2), 50(1), 800(4)

ssEPI

44.6 ± 35.5

5:21

0(1), 10(1), 30(1), 50(1), 100(2), 500(3), 800(5), 1000(7)

msTSE (2-shot PROPELLER)

48.2 ± 39.7

6:07

0(1), 800(2)

TumorT2

69.1 ± 50.7