Original article Strahlenther Onkol 2014 DOI 10.1007/s00066-014-0676-3 Received: 5 December 2013 Accepted: 10 April 2014
Jatta Berberat1 · Jane McNamara2 · Luca Remonda1 · Stephan Bodis2,3 · Susanne Rogers2
© Springer-Verlag Berlin Heidelberg 2014
3 Department of Radiation Oncology, University Hospital, Zurich, Switzerland
1 Department of Neuro-radiology, Cantonal Hospital, Aarau, Switzerland 2 Department of Radiation Oncology, Cantonal Hospital, Aarau, Switzerland
Diffusion tensor imaging for target volume definition in glioblastoma multiforme Glioblastoma multiforme (GBM) is a highly aggressive tumour associated with a poor prognosis. A combination of maximum safe tumour resection followed by postsurgical chemoradiation and adjuvant chemotherapy—both with temozolomide—has demonstrated significant survival improvements in a subset of patients [1]. Relapse remains inevitable, however, and peritumoural microscopic infiltration is a major factor in the failure of current treatment modalities for GBM [2]. High-grade gliomas are heterogeneous and it is difficult to predict whether they will relapse within the radiation field, at the margins or more distantly [3]. Pathological studies have revealed that individual glioblastoma cells tend to disperse preferentially along white matter tracts [4–5]. Conventional magnetic resonance (MR) imaging has known limitations in the detection of peritumoral microscopic infiltration. Therefore, current clinical practice involves the addition of a wide isotropic margin to encompass probable sites of tumour dissemination [6]. Alternative imaging modalities such as fluoroethyl tyrosine positron-emission tomography (FET-PET) have been explored in an attempt to optimize tumour coverage, as these techniques may identify areas of viable glioblastoma [7]. Diffusion tensor imaging (DTI) is an MR-based imaging technique that is sensitive to subtle disruptions of white matter (tensor) tracts. Newer DTI analytical techniques permit differentiation between tumour and adjacent brain that has been infiltrated by glioblastoma cells [8–13].
Longitudinal studies have shown that DTI abnormalities can identify the presence of tumour [14–15] and may even predict the pattern of recurrence [2]. Furthermore, diffusion-weighted MRI has shown promise in assessing the response to chemoradiation [16]. Whilst previous studies of apparent diffusion coefficients (ADC) and fractional anisotropy (FA) have failed to differentiate between tumour cell-infiltrated oedema and uninvolved oedema [17–18], postprocessing of DTI data has enabled the creation of maps representing isotropic and anisotropic water diffusion in the brain, which may assist in the identification of tumour and the peritumoural region. The purpose of this study was to identify tumour cell infiltration along white matter tracts in high-grade gliomas and to exploit this for radiotherapy target volume delineation.
Materials and methods Radiotherapy plans from 13 patients (mean age 56 years ± standard deviation, SD 13 years) were included in the study. Inclusion criteria were: histologically confirmed diagnosis of glioblastoma, a DTIMRI scan prior to neurosurgery and availability of three monthly follow-up MRI evaluations. Postoperative chemoradiation comprising a total dose of 60 Gy combined with temozolomide (75 mg/m2 daily) was received by 10 patients; 3 patients were treated with preoperative radiotherapy (25 Gy in 10 fractions over 5 days) followed by maximal safe tumour resection
and adjuvant temozolomide (200 mg/m2 days 1–5) in the context of a phase I clinical study. The study was approved by the local ethics committee. A high-speed multislice computed tomography (CT) scanner (Siemens Healthcare, Erlangen, Germany) was used for CT imaging (pitch 1, collimation 24 × 1.2, slice thickness 2 mm, 300 mAs, 120 kVp, image matrix 512 × 512). MRI data acquisition was performed on a Siemens 1.5T Avanto scanner (Siemens Healthcare) with a 12-channel head array coil. Each imaging session consisted of a spin-echo single-shot echo planar imaging (EPI) DTI scan with the following parameters: scan duration 4 min; voxel size 2 × 2 × 2; matrix size 128 × 128; slices 63; echo time (TE) 99 ms; repetition time (TR) 7700 ms; b-factor 1000 s/mm2; 21 diffusion encoding directions and one signal average were performed. A proton density (PD) and T2-weighted fast spin-echo (fse) data set (acquisition time 3.51 min; fat saturation, fs; slice thickness 6 mm; TR 2280 ms; TE 12 ms; echo train length, ETL = 5; matrix size 224 × 512 and three averages) was acquired, followed by a T1-weighted anatomical 3D volumetric interpolated brain examination (VIBE) sequence with contrast agent (Dotarem 0.5 mmol/ml; Guerbet AG, Roissy CdG Cedex, France; T1Gd data set) [19]. For clinical follow-up, T1Gd and T2-weighted data sets were acquired (sequence settings as above). Tractography was analysed by placing seed points in the brainstem (seeding threshold FA 0.10; step size 0.5 mm; stopping criteria for tracking FA 0.10 and de Strahlentherapie und Onkologie X · 2014
| 1
Original article
Fig. 1 8 a Axial MR slice displaying the reconstructed white matter tracts overlaid onto the T1 plus contrast agent (T1Gd) anatomical data set and showing the aberrant right lateral corticospinal tract. b Additional coregistation of the isotropic/anisotropic component of diffusion (pq) map yields the final “infiltration map” and reveals infiltration lateral to the primary tumour. c Coronal views of the radiotherapy volumes: primary tumour plus 1 cm isotropic expansion (yellow), diffusion tensor imaging planning target volume (DTI-PTV; purple) and conventional PTV (green). The volume of DTI-PTV was 47 % smaller than the clinical PTV. d The pattern of recurrence on T1Gd MRI 11 months after diagnosis suggests a route of spread via the abnormal right corticospinal tract. The primary tumour is delineated in yellow in c; both the DTI-PTV (purple in c) and the PTV (green in c) included the site of recurrence (red arrow)
flection angle 60°) using the Syngo workstation (Siemens Healthcare). Then the data was overlaid on the T1Gd VIBE images. Further image processing of the diffusion tensor data set was performed using an in-house program implemented in Matlab (Mathworks Inc., Natick, MA, US) and BrainVoyager QX v2.4 (Brain Innovation BV, Maastricht, Netherlands), based on the method by Basser et al. [20]. For each voxel, the eigenvalues were computed and then used to calculate FA, as well as the isotropic (p) and the anisotropic component of diffusion (q) [11]. For each patient, a mirror-image region of contralateral normal brain was used as an internal control. Regions rep-
2 | Strahlentherapie und Onkologie X · 2014
resenting microscopic tumour infiltration were generated according to biopsy-validated p and q values (% of control) determined by Price et al [11]. These “pq” infiltration maps were coregistered with the diffusion tensor tracts (. Fig. 1a) and overlaid on T1Gd VIBE images to create a final pq map (. Fig. 1b). The analysis was performed blinded to the subsequent pattern of relapse.
Radiation treatment planning Radiotherapy plans had previously been created for clinical use by coregistering the planning CT with preoperative T1Gd and T2-weighted MR images. For each patient,
a pq map was additionally fused with the planning CT scan. The gross target volume (GTV) was first delineated conventionally as the enhancing tumour visible on T1Gd images and then modified to include any further regions defined as tumour by the pq map. Peritumoral infiltration was contoured separately according to the pq map and combined with the GTV to create the “DTI-CTV”. The abnormal tensor tracts were defined visually by loss of symmetry with the contralateral side (. Fig. 1a). The DTI-CTV was isotropically expanded by 1 cm and then extended a further 1 cm in length and width along the visible, apparently normal, white matter tracts adjacent to the tumour to create the DTI-PTV (. Fig. 1b). The aim was to generate an anisotropic margin with 1-cm coverage of the region at lowest risk of peritumoural infiltration and a 2-cm margin along the likely direction of tumour cell dissemination. The clinical PTV was defined as the GTV (T1Gd) plus a 25-mm isotropic margin. Additional intensity-modulated radiation therapy (IMRT) plans were produced for the DTI-PTVs, but these were not used for clinical treatment. The conventional (T2-weighted CTV and T1Gd PTV) and research (DTI-CTV and DTIPTV) volumes were compared.
Statistics A paired t-test was performed to compare the planning volumes (GraphPad Software, La Jolla; CA, USA). Statistical significance was defined as p
Jatta Berberat1 · Jane McNamara2 · Luca Remonda1 · Stephan Bodis2,3 · Susanne Rogers2
© Springer-Verlag Berlin Heidelberg 2014
3 Department of Radiation Oncology, University Hospital, Zurich, Switzerland
1 Department of Neuro-radiology, Cantonal Hospital, Aarau, Switzerland 2 Department of Radiation Oncology, Cantonal Hospital, Aarau, Switzerland
Diffusion tensor imaging for target volume definition in glioblastoma multiforme Glioblastoma multiforme (GBM) is a highly aggressive tumour associated with a poor prognosis. A combination of maximum safe tumour resection followed by postsurgical chemoradiation and adjuvant chemotherapy—both with temozolomide—has demonstrated significant survival improvements in a subset of patients [1]. Relapse remains inevitable, however, and peritumoural microscopic infiltration is a major factor in the failure of current treatment modalities for GBM [2]. High-grade gliomas are heterogeneous and it is difficult to predict whether they will relapse within the radiation field, at the margins or more distantly [3]. Pathological studies have revealed that individual glioblastoma cells tend to disperse preferentially along white matter tracts [4–5]. Conventional magnetic resonance (MR) imaging has known limitations in the detection of peritumoral microscopic infiltration. Therefore, current clinical practice involves the addition of a wide isotropic margin to encompass probable sites of tumour dissemination [6]. Alternative imaging modalities such as fluoroethyl tyrosine positron-emission tomography (FET-PET) have been explored in an attempt to optimize tumour coverage, as these techniques may identify areas of viable glioblastoma [7]. Diffusion tensor imaging (DTI) is an MR-based imaging technique that is sensitive to subtle disruptions of white matter (tensor) tracts. Newer DTI analytical techniques permit differentiation between tumour and adjacent brain that has been infiltrated by glioblastoma cells [8–13].
Longitudinal studies have shown that DTI abnormalities can identify the presence of tumour [14–15] and may even predict the pattern of recurrence [2]. Furthermore, diffusion-weighted MRI has shown promise in assessing the response to chemoradiation [16]. Whilst previous studies of apparent diffusion coefficients (ADC) and fractional anisotropy (FA) have failed to differentiate between tumour cell-infiltrated oedema and uninvolved oedema [17–18], postprocessing of DTI data has enabled the creation of maps representing isotropic and anisotropic water diffusion in the brain, which may assist in the identification of tumour and the peritumoural region. The purpose of this study was to identify tumour cell infiltration along white matter tracts in high-grade gliomas and to exploit this for radiotherapy target volume delineation.
Materials and methods Radiotherapy plans from 13 patients (mean age 56 years ± standard deviation, SD 13 years) were included in the study. Inclusion criteria were: histologically confirmed diagnosis of glioblastoma, a DTIMRI scan prior to neurosurgery and availability of three monthly follow-up MRI evaluations. Postoperative chemoradiation comprising a total dose of 60 Gy combined with temozolomide (75 mg/m2 daily) was received by 10 patients; 3 patients were treated with preoperative radiotherapy (25 Gy in 10 fractions over 5 days) followed by maximal safe tumour resection
and adjuvant temozolomide (200 mg/m2 days 1–5) in the context of a phase I clinical study. The study was approved by the local ethics committee. A high-speed multislice computed tomography (CT) scanner (Siemens Healthcare, Erlangen, Germany) was used for CT imaging (pitch 1, collimation 24 × 1.2, slice thickness 2 mm, 300 mAs, 120 kVp, image matrix 512 × 512). MRI data acquisition was performed on a Siemens 1.5T Avanto scanner (Siemens Healthcare) with a 12-channel head array coil. Each imaging session consisted of a spin-echo single-shot echo planar imaging (EPI) DTI scan with the following parameters: scan duration 4 min; voxel size 2 × 2 × 2; matrix size 128 × 128; slices 63; echo time (TE) 99 ms; repetition time (TR) 7700 ms; b-factor 1000 s/mm2; 21 diffusion encoding directions and one signal average were performed. A proton density (PD) and T2-weighted fast spin-echo (fse) data set (acquisition time 3.51 min; fat saturation, fs; slice thickness 6 mm; TR 2280 ms; TE 12 ms; echo train length, ETL = 5; matrix size 224 × 512 and three averages) was acquired, followed by a T1-weighted anatomical 3D volumetric interpolated brain examination (VIBE) sequence with contrast agent (Dotarem 0.5 mmol/ml; Guerbet AG, Roissy CdG Cedex, France; T1Gd data set) [19]. For clinical follow-up, T1Gd and T2-weighted data sets were acquired (sequence settings as above). Tractography was analysed by placing seed points in the brainstem (seeding threshold FA 0.10; step size 0.5 mm; stopping criteria for tracking FA 0.10 and de Strahlentherapie und Onkologie X · 2014
| 1
Original article
Fig. 1 8 a Axial MR slice displaying the reconstructed white matter tracts overlaid onto the T1 plus contrast agent (T1Gd) anatomical data set and showing the aberrant right lateral corticospinal tract. b Additional coregistation of the isotropic/anisotropic component of diffusion (pq) map yields the final “infiltration map” and reveals infiltration lateral to the primary tumour. c Coronal views of the radiotherapy volumes: primary tumour plus 1 cm isotropic expansion (yellow), diffusion tensor imaging planning target volume (DTI-PTV; purple) and conventional PTV (green). The volume of DTI-PTV was 47 % smaller than the clinical PTV. d The pattern of recurrence on T1Gd MRI 11 months after diagnosis suggests a route of spread via the abnormal right corticospinal tract. The primary tumour is delineated in yellow in c; both the DTI-PTV (purple in c) and the PTV (green in c) included the site of recurrence (red arrow)
flection angle 60°) using the Syngo workstation (Siemens Healthcare). Then the data was overlaid on the T1Gd VIBE images. Further image processing of the diffusion tensor data set was performed using an in-house program implemented in Matlab (Mathworks Inc., Natick, MA, US) and BrainVoyager QX v2.4 (Brain Innovation BV, Maastricht, Netherlands), based on the method by Basser et al. [20]. For each voxel, the eigenvalues were computed and then used to calculate FA, as well as the isotropic (p) and the anisotropic component of diffusion (q) [11]. For each patient, a mirror-image region of contralateral normal brain was used as an internal control. Regions rep-
2 | Strahlentherapie und Onkologie X · 2014
resenting microscopic tumour infiltration were generated according to biopsy-validated p and q values (% of control) determined by Price et al [11]. These “pq” infiltration maps were coregistered with the diffusion tensor tracts (. Fig. 1a) and overlaid on T1Gd VIBE images to create a final pq map (. Fig. 1b). The analysis was performed blinded to the subsequent pattern of relapse.
Radiation treatment planning Radiotherapy plans had previously been created for clinical use by coregistering the planning CT with preoperative T1Gd and T2-weighted MR images. For each patient,
a pq map was additionally fused with the planning CT scan. The gross target volume (GTV) was first delineated conventionally as the enhancing tumour visible on T1Gd images and then modified to include any further regions defined as tumour by the pq map. Peritumoral infiltration was contoured separately according to the pq map and combined with the GTV to create the “DTI-CTV”. The abnormal tensor tracts were defined visually by loss of symmetry with the contralateral side (. Fig. 1a). The DTI-CTV was isotropically expanded by 1 cm and then extended a further 1 cm in length and width along the visible, apparently normal, white matter tracts adjacent to the tumour to create the DTI-PTV (. Fig. 1b). The aim was to generate an anisotropic margin with 1-cm coverage of the region at lowest risk of peritumoural infiltration and a 2-cm margin along the likely direction of tumour cell dissemination. The clinical PTV was defined as the GTV (T1Gd) plus a 25-mm isotropic margin. Additional intensity-modulated radiation therapy (IMRT) plans were produced for the DTI-PTVs, but these were not used for clinical treatment. The conventional (T2-weighted CTV and T1Gd PTV) and research (DTI-CTV and DTIPTV) volumes were compared.
Statistics A paired t-test was performed to compare the planning volumes (GraphPad Software, La Jolla; CA, USA). Statistical significance was defined as p
