ESTRO 2024 - Abstract Book

S3860

Physics - Image acquisition and processing

ESTRO 2024

resulting in signal blurring and misalignment of images acquired with different b-values. Several approaches have been proposed to manage respiratory motion during MRI acquisitions, e.g. navigator triggering [2] and motion reduction using an abdominal compression belt [3]. Although several studies showed that average ADC values are comparable between different motion compensation methods [4-5], it is unknown whether these methods improve the repeatability. As sufficient repeatability of ADC measurements is crucial for treatment response monitoring, the aim of this study was to investigate the effect of different motion compensation techniques on the repeatability of ADC measurements on a 1.5T MR-linac system.

Material/Methods:

For this study, a free breathing (FB) and a navigator-triggered (NT) acquisition with and without the use of an abdominal compression belt (Aspen Medical Products, Irvine, USA; FB+belt and NT+belt) were compared on a 1.5T MR-linac (Unity, Elekta AB). Most acquisition parameters were equal between FB and NT: axial single-shot EPI readout with descending slice order, field-of-view = 510x319mm 2 , acquired voxel size=3.5x3.5x5mm 3 , slice gap = 0.5mm, TE = 83ms, and b-values = 0, 200, 500s/mm 2 ). Total scan time for the FB acquisition was 2m51s. In order to keep the scan time for the NT acquisition around 5min, different settings were used for TR (FB: 3264ms; NT: 2176ms), number of slices (FB: 27; NT: 18), and number of averages per b-value (FB: 4, 8, 16; NT: 2, 6, 12). ADC was derived from the b = 200 and 500s/mm 2 .

To determine the accuracy of the FB and NT acquisitions with respect to a calibration sequence, the NIST/QIBA diffusion phantom was scanned. The ADC bias at isocenter was estimated according to QIBA guidelines [6].

To assess the repeatability, a test-retest study was performed with 26 volunteers. The study was approved by the institutional review board and all volunteers signed informed consent. DWI data of all four acquisitions were acquired on two separate days (max 2 weeks apart). For each acquisition, circular ROIs of 1 or 2cm in diameter were delineated on the b = 0s/mm 2 image at three consecutive slices in homogeneous areas of the liver (2 ROIs), left kidney (1 ROI) and spleen (1 ROI), avoiding large vessels and organ borders. For each ROI, the median ADC was determined and compared between the four acquisitions of the first scan session using the Friedman test and Bonferroni corrected post-hoc testing. The repeatability coefficient (RC = 2.77 ✕ wSD) and percentage RC (%RC = 2.77 ✕ wCV) were determined from the two repeated measurements for each acquisition [7].

Results:

Compared to the reference value of the vial in the isocenter, the FB and NT acquisitions underestimated the ADC of - 0.057 ✕ 10 -3 and -0.060 ✕ 10 -3 mm 2 /s respectively, which was smaller using the calibration sequence (-0.013 ✕ 10 3 mm 2 /s). A paired t-test showed that ADC values from the FB and NT acquisition were similar (p=0.79). Generally, image quality improved using the NT acquisition and the compression belt (Figure 1). Mean scan time of the NT duration was 5m39s ± 46s. So far, the data from 11 volunteers were analyzed. ADC values were similar across acquisitions except for one of the liver ROIs (Table 1). However, after multiple comparison correction, none of the differences remained statistically significant during post-hoc testing (all adjusted p>0.16). RC values ranged from 0.17 ✕ 10 -3 mm 2 /s to 0.47 ✕ 10 -3 mm 2 /s across ROIs and acquisitions (Table 1).