ESTRO 2020 Abstract book

S1045 ESTRO 2020

The End-to-End test for small animal radiation research has been developed and implemented. Absolute dosimetry with alanine in relevant preclinical irradiation conditions showed reasonable levels of accuracy when compared to TPS calculations. It was possible to detect problems related to the commissioning of Muriplan. This work is a step forward in providing an independent and traceable dosimetric validation in pre-clinical radiotherapy. PO-1782 Compressed sense accelerated MRI simulation: image quality assessment using an ACR MRI phantom O.L. Wong 1 , J. Yuan 1 , Y. Zhou 1 , K.Y. Cheung 1 , S.K. Yu 1 1 Hong Kong Sanatorium & Hospital, Medical Physics and Research Department, Hong Kong, Hong Kong SAR China Purpose or Objective MRI is increasingly used in RT but still suffers from long scan, even with fast sequences and parallel imaging. Compressed sense (CS) MRI, based on incoherent K-space undersampling and dedicated reconstruction, has potentials to further shorten RT MRI simulation, but is rarely explored. We aim to quantitatively evaluate CS-MRI image quality for MRI simulation use. Material and Methods ACR MRI large phantom was scanned on a 1.5T MR-sim (a 16-ch array and 2 surface loops for head MR-sim setting) using 2D fast spin echo (FSE) (TR/TE = 500/20ms (T1w); TR/TE=2000/80ms (T2w), voxel = 0.93x0.93x5mm 3 , phase encoding PE=LR) and a customized 3D T1w FSE sequence (TR/TE=420/29ms, voxel = 1x1x1mm 3 , PE=LR) for head MR- simulation, with various CS acceleration factors (2-20). Denoising level was set as high. First, image artifacts were visually assessed. Image quality was then quantitatively evaluated in terms of signal-to-noise ratio (SNR), percent intensity uniformity (PIU), low contrast detectability measured by contrast to noise ratio (CNR), resolvable spatial resolution (SR), edge sharpness measured by the 10-90% image intensity rise width (RW) of the cylindrical insert, geometric distortion (GD) of the inner diameter (Dtrue=190mm). Results Traditional SENSE undersampling artifact along PE was only observable with CS >8 for 2D and CS>13 for 3D scan. Visual assessment observed noticeable over-smoothed appearance for all images. This over-smoothing was more severe at high CS factors, and showed patch-like patterns, indicating over regularization of sparsity constraining in the wavelet domain. Decreasing SNR was found with larger CS factor. SNR reduction from CS2 to CS20 were 57% for 2D T1w, 78% for 2D T2w and 72% for 3D T1w. This SNR reduction rate seemed relatively low, probably due to high denoising level setting. PIU reduced for both T1w (2D: 94.7 to 81.9%; 3D: 94.3 to 74.8%) and T2w (2D: 97.1 to 81.6%) from CS2 to CS20. High CS factor reduced low contrast detectability. where reduction between CS2 and CS20 were 47% for T1w2D, 66% for T2w2D and 92% for 3D scans. All low contrast inserts became invisible above CS8 for 2D and CS17 for 3D. An overall decreasing CNR was exhibited. CNR reduction from CS2 to CS20 were 47% for 2D T1w, 66% for 2 T2w and 92% for 3D. For 3D scan, 1mm and 1.1mm SR was unresolvable at CS>2 and CS>8, respectively. For 2D scan, 0.9mm, 1mm and 1.1mm SR became unresolvable when CS>5, CS>6 and CS>9 for T1w, and CS>5, CS>7 and CS>9 for T2w. No apparent reduction in edge sharpness was noted when CS<14. On all images, measured inner diameters were all close to true value with sub-mm deviations. This suggested that image distortion was barely affected, even with compromised edge sharpness at high CS factors.

The End-to-End test was developed and implemented based on alanine from the NPL alanine measurements service [1] as the reference detector, placed in an intracranial cavity of a purposely designed, anatomically correct mouse phantom (with tissue equivalent material) [2] (Figure 1). A set of standard conditions for planning and irradiations for the End-to-End test at five UK Xstrahl SARRP devices were defined: single field, at zero degrees gantry angle with the 10 mm x 10 mm collimator (or for institutions with bespoke build collimators, the closets dimension to it) and 15 Gy prescribed to a point centred in the alanine pellet. A workflow with specific procedures for each step: phantom preparation, positioning, imagining, segmentation, pellet contouring, time and dose distribution calculations, was defined and followed. A minimum of three irradiations were performed at each institution. Correction to alanine energy dependence was applied according to each institution’s half value layer (HVL) quality index. Mean dose to the pellets, calculated by Muriplan treatment planning system (TPS) were compared to alanine measured dose as % difference = 1- Dose TPS /Dose alanine .

Results Figure 2 presents a summary of the End-to-End test results for all the participating institutions. The five institutions are labelled I1 – I5. Initial results in I1 showed on average a 12.9% difference between the TPS calculated and the alanine measured dose. Considering that prior to this work, measurements of I1’s SARRP reference output carried out with ionization chamber and alanine, compared within 2%, Xstrahl was contacted about the commissioning of the TPS. As a consequence, the TPS was recommissioned and a new set of measurements for the End-to-End test were performed which resulted in an average difference of 4.7%. Only in the case of Institution 4 (I4), the averaged dose difference was larger than 5%. It was not possible to verify the HVL at I4, which could have had an effect in the magnitude of the difference.

Conclusion