An analysis of inter-fraction prostate deformation relative to implanted fiducial markers using finite element modelling

Abstract

To determine the probability and magnitude of deformation of the prostate relative to implanted fiducial markers and to quantify Internal Margins for image-guided radiotherapy. Seventeen patients were imaged on two occasions and the change in shape of the prostate was studied using finite element modelling. Patients were eligible if they had a pathological diagnosis of adenocarcinoma of the prostate with clinical stage T1c-T2c and consented to radiotherapy with online setup correction, but were ineligible if they had concurrent androgen deprivation. Three gold fiducial markers were implanted under ultrasound guidance to the apex, the posterior mid-point and the anterior base of the prostate in a para-mid sagittal plane. Patients had axial MRIs in a 1.5 T Signa Infinity™ (GE, Milwaukee, WI) scanner using a gradient recalled echo sequence optimized for viewing gold markers (GRE TR/TE: 650/15 ms, FA: 25) and T2 fast spin echo (T2 FSE TR/TE: 4400/102 ms) with 2 mm slice thickness and no spacing at the time of their computed tomography simulation and once during radiotherapy. A single observer contoured the prostate and gold seed markers on the GRE images, with reference to the fused T2 images, using Pinnacle™ (ADAC Milpitas, CA) treatment planning software. The contours were exported to a finite element modelling package Hypermesh™, (Altair, Troy, MI) where a three node, tria-element surface mesh was created to represent the prostate surface. The average surface area per element was 0.0025 cm2. Different methods of aligning the prostates between the two MRIs were evaluated: 1) a translation of the center of mass of the three fiducial markers without rotation 2) a translation and rotation of the center of mass of the three fiducial markers. Following the seed alignment procedure, the residual error, representing deformation of the prostate surfaces between image studies was calculated by performing 3D deformable alignment. Each node from the mesh representing the prostate surface on the first MRI was projected to the surface constructed from the mesh representing the prostate on the second MRI. On the initial MRI scan, the inter-seed distances were from base-mid: 16.3 mm (range: 6.6–25.7), mid-apex: 17.9 mm (range: 10.6–24.9) and base-apex: 27.6 mm (range: 16.5–38.6). The mean prostate volume was not significantly different between the first and second scans: 38.9 vs 37.6 cc (p = 0.9). The differences in inter-seed distances between the two scans were respectively: 1.0 mm (CI: −4.6–6.6), 0.0 mm, (CI: −2.6–2.6) and 0.1 mm (CI: −3.0–5.0). With translational alignment of the center of mass of the fiducial markers, the maximal deformations of the surface of the prostate were respectively, in transverse, anterior-posterior and superior-inferior directions: 5 mm, 7 mm and 7 mm. The proportion of prostates that had less than 3 mm deformation over 90% of the surface was 35%. On average, 12% (range: 0.2–29.0) of the prostate surface deformed by greater than 3 mm. With translational and rotational alignment of the center of mass of the fiducial markers, the maximal deformations of the surface of the prostate were respectively, in transverse, anterior-posterior and superior-inferior directions: 5 mm, 9 mm and 7 mm. The proportion of prostates that had less than 3 mm deformation over 90% of the surface was 47%. On average, 14% (range: 0.7–43.0) of the prostate surface deformed by greater than 3 mm. No fiducial marker migration was identified. With both methods of alignment to fiducial markers, most prostates exhibit inter-fraction deformations greater than 3 mm, some as large as 9 mm. Accurate assessment of marker surrogacy is critical for the determination of appropriate margins in on-line marker-based image-guided radiation therapy of the prostate. Further margin reductions may be possible with imaging methods that can discern prostate deformation.