ESTRO 2024 - Abstract Book

S4982

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

Michael J van Rijssel, Cornelis A T van den Berg, Astrid L H M W van Lier

UMC Utrecht, Radiotherapy, Utrecht, Netherlands

Purpose/Objective:

Relaxometry may aid in response monitoring and both OAR and target definition [1], [2]. Therefore, there is increased interest in using MRL systems for functional imaging and relaxometry [3], [4]. However, in the online setting scan time is limited, so additional sequences for relaxometry should be fast and efficient and ideally intertwined with anatomical imaging. Current methods require approximately 15 minutes of additional scan time to quantify T 1 and T 2 . We propose a method that measures both and produces high-quality anatomical images, all in a scan time of approximately four minutes. We investigate the accuracy and precision of this method in a calibrated phantom and demonstrate our method in a healthy volunteer.

Material/Methods:

Relaxometry Phase-cycled balanced steady state free precession (PC-bSSFP) acquisitions allow for simultaneous T1 & T2 mapping if multiple phase cycles (PCs) are acquired [5], [6]. Additionally, an image free from banding artefacts that hamper balanced acquisitions can be calculated, which could be used for position verification and online replanning [7]. We use the MIRCALE method to estimate T 1 & T 2 [5]. In MIRACLE, image intensity versus PC information is decomposed into modes using a Fourier transform. Ratios of these modes are in turn used to derive T 1 and T 2 . Only the central three modes are used in this procedure, therefore a minimum of three PCs should be measured. However, measuring fewer PCs will lead to aliasing and inaccurate estimates of both the ratios and the derived T 1 and T 2 estimates. In an iterative process we estimate this error and correct for it [8]. This allows for a four-PC scan of approximately four minutes, where existing fitting methods would require approximately ten minutes. experiment We imaged the Eurospin II TO5 phantom (Diagnostic Sonar LTD, Livingston, Scotland) with 12 tubes with known T 1 and T 2 on a Unity MRL (Elekta AB, Stockholm, Sweden) with a 3D PC-bSSFP with 11 equidistant PCs over 360° (9.88 Hz offset between phases), TR 9.2 ms, TE 4.6 ms, FA 35°, readout bandwidth 430 Hz/pix, voxel size 0.7x0.7x2 mm 3 , in 53 s/PC. To assess the influence of the iterative reconstruction, the MIRACLE method was applied both on the full dataset and on a subset of only four PCs, both with and without iterative reconstruction [5], [8]. The means and standard deviations of ROI’s placed in the tubes were calculated. Accuracy was assessed on the means using Bland & Altman’s limits of agreement with the reference; precision was assessed using the coefficient of variation [9]. In-vivo We scanned a healthy volunteer with a similar scan protocol, slightly adjusted to allow for fat suppression. We used a 3D PC-bSSFP with 11 equidistant PCs over 360° (9.09 Hz offset between phases), TR 10 ms, TE 5.0 ms, FA 35°, readout bandwidth 430 Hz/pix, voxel size 0.7x0.7x2 mm 3 , in 67 s/PC. A banding-free anatomical image, T 1 and T 2 maps were calculated using MIRACLE in combination with iterative reconstruction and ROI analyses were performed [5], [8]. Phantom

Results: