As a powerful molecular imaging tool, positron emission tomography (PET) is increasingly being used for early assessment of tumour response to therapy [
Response assessment using SUVs requires the selection of either a representative tumor voxel or a region of interest (ROI) for quantification. One of the simplest and most common methods of quantifying tumor uptake is to use the single voxel containing the maximum SUV (
Table
A summary of previous Intra-Tx tumour response assessment studies that used the Fixed size ROI method.
Study |
|
Site | Pre-Tx ROI | Intra-Tx ROI | Res. Thr |
---|---|---|---|---|---|
Schelling et al. (2000) [ |
22 | Breast | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at max | 55% |
Weber et al. (2003) [ |
57 | Lung | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at max | 20% |
Avril et al. (2005) [ |
33 | Ovarian | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at max | 20% |
Brun et al. (2002) [ |
47 | Head and neck | Fixed size, square (4 or 9 pixels) at max | Fixed size, square (4 or 9 pixels) at max | Median |
Rousseau et al. (2006) [ |
64 | Breast | Fixed size, 5 to 10 mm at max | Fixed size, 5 to 10 mm at max | 40% |
Maisonobe et al. (2013) [ |
40 | Colorectal | Fixed size, |
Fixed size, |
N/Sp |
Ott et al. (2006) [ |
65 | EG junction | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at the same position using landmark | 35% |
Weber et al. (2001) [ |
40 | EG junction | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at the same position using landmark | 35% |
Wieder et al. (2007) [ |
24 | EG junction | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at the same position using landmark | 35% |
Ott et al. (2003) [ |
44 | Gastric | Fixed size, circular 15 mm at max | Fixed size, circular 15 mm at the same position using landmark | 35% |
Wieder et al. (2004) [ |
38 | Esophagus | Fixed size, circular 15 mm at max | N/Sp | 30% |
Schwarz et al. (2005) [ |
11 | Breast | Fixed size, circular (size N/Sp) manually placed | Fixed size, circular (size N/Sp) manually placed | 20% |
The distribution of uptake within the tumour may change in response to therapy such that the maximum uptake point in the Intra-Tx study is found at an anatomically different location than it was prior to treatment. This is illustrated in Figures
Change in the distribution of FDG uptake during treatment. The PET/CT images of Pre-Tx (a) and Intra-Tx (b) are of a patient with a base of tongue primary tumour. Two circular ROIs of 15 mm diameter are centered at the maximum uptake points on both Pre-Tx (green) and Intra-Tx (red) images denoted by “M.” An additional 15 mm diameter circular ROI is placed on the Intra-TX image (blue) in a position judged to correspond to the same anatomical location as the ROI as in the Pre-Tx. The FDG uptake profiles along the black lines connecting the two Intra-Tx ROIs are shown in Figure
The uptake profile from Figure
Using the ROIsame method is reasonable if the goal is to evaluate the change in uptake in the same area of the tumour. This method has been recommended by the European Organization for Research and Treatment of Cancer (EORTC) [
Both tumors and normal tissues may shrink and shift during the treatment. The coregistered PET/CT images of Pre-Tx (a) and Intra-Tx (b) are cross-sectional images of a patient with a primary tumour of the tonsil. The patient’s left parotid gland in Intra-Tx (yellow contour) shows volume loss and shift relative to that in Pre-Tx (red contour). Similarly, the gross tumour volume for one nodal disease site in Intra-Tx (dotted green contour) shows volume loss and shift relative to that in Pre-Tx tumour (dotted blue contour).
Uncertainty in the placement of the Intra-Tx ROI could significantly affect the accuracy of quantitative response assessment. Uncertainties in quantitative response assessment could have significant impact on treatment decisions and clinical outcome. Consequently, we investigated the effects of fixed-size ROI placement on quantitative response assessment. The purpose of this study was twofold: (1) to evaluate quantitative response assessment when Intra-Tx PET images are measured using the ROIpeak and ROIsame methods; (2) to quantify the geometric changes of both tumour and normal tissues and their impact on quantitative response assessment using the ROIsame method.
Two independent populations (A and B) were used. Population A consisted of 15 patients with a total of 38 gross tumour volumes (GTV) identified by experienced radiation oncologists. Population A was used to compare two quantitative tumour response assessments based on using the ROIpeak and ROIsame methods. Population B consisted of 10 patients with a total of 33 GTVs identified by experienced radiation oncologists and was used to quantify geometric changes of both tumour and normal tissues during therapy. The impact of these geometric changes on quantitative tumour response assessment was evaluated in population A. Both populations A and B were part of a clinical trial at Sunnybrook Health Sciences Centre (Toronto, Canada) to assess tumour response in patients with advanced HNC. Population B consisted of patients entered in the pilot study which proceeded the main trial while population A consisted of patients entered in the clinical trial itself.
While populations A and B were very similar, there were some slight differences, primarily in the CT-voxel size used and the average time between the Pre-Tx and Intra-Tx scans. All patients in both groups had locally advanced HNC (stage III or IV) and underwent 6.6 weeks of radical radiotherapy with concurrent chemotherapy. Patients received intensity modulated radiation therapy (IMRT) of 70 Gy in 33 fractions to all GTVs for both primary (GTVp) and involved lymph nodes (GTVn). All patients also received concurrent bolus platinum chemotherapy as tolerated by intravenously injecting 100 mg/m2 Cisplatin on days 1, 22, and 43. Patients underwent two sequential FDG-PET/CT scans, one Pre-Tx and one Intra-Tx, both supine in the same position using a thermoplastic radiotherapy immobilization mask. One 18 cm axial field of view (FOV) that covered the head and neck area was used. The PET/CT scanner was the GEMINI System (Philips Medical System, Cleveland, Ohio). Prior to the PET/CT scans, patients were injected with 5 MBq of FDG per kg. Patients heavier than 75 kg were injected with a fixed dose of 370 MBq of FDG. PET images were reconstructed using a 3-Dimensional Row-Action Maximum Likelihood Algorithm (3D-RAMLA) and corrected for attenuation using CT. In order to register the Intra-Tx CT to the Pre-Tx CT images, a Chamfer matching algorithm [
Pre-Tx FDG PET/CT scans were performed 14 ± 4 days (range, 8–22) prior to the start of the treatment. Intra-Tx FDG PET/CT scans were performed 16 ± 2 days (range, 11–20) after the first treatment day. The CT-voxel size was 0.59 × 0.59 × 1.60 mm3 and the CT FOV was 300 × 300 × 210 mm3 in lateral, anterior-posterior, and superior-inferior directions, respectively. The PET voxel size was 2 × 2 × 2 mm3 and the PET FOV was 576 × 576 × 180 mm3 in lateral, anterior-posterior, and superior-inferior directions, respectively. PET images were acquired 50 minutes after injection for 2.5 minutes. The Pre-Tx and Intra-Tx PET postinjection acquisition times were matched within 5 minutes.
The SUVs were normalized to the patients’ body weight. For each GTV, ROIpeak (a circular ROI of 15 mm diameter) was placed on a single transaxial slice centered at the maximum FDG uptake point in both Pre-Tx and Intra-Tx images. For each GTV, ROIsame (a circular ROI of 15 mm diameter) was also placed on a single transaxial slice at the location of the Intra-Tx image that corresponded to the same physical location as the Pre-Tx max-point ROI. A dual-board certified, nuclear medicine/radiology physician positioned ROIsame based on anatomical landmarks. Thus, each GTV had two Intra-Tx ROIs. The distance between the centers of these two ROIs was measured in 3D geometry.
On the same transaxial slice where ROIsame was located, the Intra-Tx GTV size was measured by averaging the anterior-posterior and lateral extents of an oncologist drawn GTV. In order to reduce errors in FDG uptake from partial volume effects, only Intra-Tx GTVs larger than 15 mm were subsequently analyzed, reducing the total number of GTVs available for analysis from 38 to 26.
Tumour response assessments were obtained using two different methods, called
In order to determine how uncertainties in positioning ROIsame due to geometric changes may impact
Pre-Tx FDG PET/CT scans were performed 17 ± 5 days (range, 13–28) prior to the start of the treatment. Intra-Tx FDG PET/CT scans were performed 33 ± 4 days (range, 28–40) after the first treatment day. The CT-voxel size was 1.17 × 1.17 × 6.5 mm3 and the CT FOV was 600 × 600 × 208 mm3 in the lateral, anterior-posterior, and superior-inferior directions, respectively. The PET voxel size was 2 × 2 × 2 mm3 and the PET FOV was 576 × 576 × 180 mm3 in the lateral, anterior-posterior, and superior-inferior directions, respectively.
GTVs were contoured manually by an oncologist experienced in treatment of HNC. All the GTVs were contoured on CT images guided by coregistered PET images. Noncoregistered diagnostic MR images were available to aid contouring in all patients except one where no MRI study was performed. Radiology reports on both PET/CT and MRI studies were also used to aid in contouring.
Geometric changes of the GTVs and normal tissues during treatment were thought to be potentially important in influencing the accuracy of placement of an ROI for quantitative tumour response assessment. In addition to GTVs, geometric changes of some normal tissues were also quantified. While the geometric shifts in tumours, not normal tissues, were of primary interest, the uncertainty in estimating geometric shifts in GTVs was greater than the uncertainty in estimating shifts in other structures, simply due to the difficulty in accurately delineating the GTV after treatment. Thus, the geometric shifts in normal tissues were used as surrogate measures of possible shifts in GTVs. Ten normal tissues were contoured on both Pre-Tx and Intra-Tx CT images for each patient. These normal tissues included the C2 vertebral body, mandible, hyoid, spinal cord, right and left sternocleidomastoid muscles, right and left parotid glands, and right and left submandibular glands. All normal tissues were contoured using consistent window and level settings under the guidance of an experienced oncologist. The most inferior extent for contouring the spinal cord and the sternocleidomastoid muscles was the most superior aspect of the apex of the lung. The most superior extent of the spinal cord was chosen to correspond to the most superior extent of the C2 vertebral body. Mandible and parotid contours were excluded from one patient since the scan did not include the entire organs in the superior direction.
Using both Pre-Tx and Intra-Tx contours for normal tissues and GTVs, an IDL program was developed in house to quantify the geometric changes by calculating percentage volume changes, that is, Intra-Tx volume relative to Pre-Tx volume, shift of center Of mass (COM), that is, Intra-Tx COM relative to Pre-Tx COM. The shifts were calculated as a shift vector in a 3D geometry and the reported values are the absolute values of these vectors.
Patient characteristics for both populations are listed in Table
Patient characteristics.
Population A | Population B | |
---|---|---|
No. of patients | 15 | 10 |
Sex (F, M) | 3 F, 12 M | 1 F, 9 M |
Age | ||
Mean age ± SD | 58.3 ± 5.7 yr | 58.7 ± 11.6 yr |
Age range | 49–68 yr | 42–79 yr |
Clinical stage | ||
Stage III | 5 | 1 |
Stage IV | 10 | 9 |
Total no. of GTV | 38 | 33 |
GTVp | 15 | 10 |
GTVn | 23 | 23 |
Site | ||
Tongue | 5 | 4 |
Tonsil | 3 | 3 |
Hypopharynx | 5 | 1 |
Larynx | 2 | 1 |
Paranasal sinus | 0 | 1 |
F: female, M: male, SD: standard deviation, GTV: gross tumour volume, GTVp: primary tumour, GTVn: involved lymph node.
The mean Intra-Tx GTV size (i.e., average of the anterior-posterior and lateral extents) was
Histogram of distances between the centers of the two Intra-Tx ROIs.
Figure
Comparison between the two quantitative tumour response measurements when two different Intra = Tx ROI methods were used. Plot (a) is a scatter plot of the two methods. The solid line in this graph is the unity line where
As expected, Intra-Tx SUVpeak had a higher value than SUVsame for most GTVs, resulting in a lower value for
Figure
Figure
Plots showing how uncertainties in positioning ROIsame impact tumor response assessment measured by
A total of 97 normal tissue regions were contoured in 10 patients in both Pre-Tx and Intra-Tx. Figure
Geometric changes due to therapy for GTV and normal tissues characterized by percentage volume changes (a) and shifts (b). The bars show the median values and the error bars show the standard errors. GTVp: gross tumor volume (primary), GTVn: involved lymph node, LT: left, RT: right, sman: submandibular, sMuscle: sternocleidomastoid muscle.
For normal tissues, significant volume losses were only found for the salivary glands. Median volume losses were 28.1% (range, 7.3–45.6%) for all parotid glands and 31.0% (range, 13.3–48.7%) for all submandibular glands. Other soft tissues (i.e., sternocleidomastoid muscles and spinal cord) and bones did not show significant volume losses.
Figure
Geometric changes during therapy can be expected to influence the accuracy with which an expert can place the tumour ROI and thus could affect tumour response assessment using the ROIsame approach. Our results in Figure
The placement of the fixed size ROI could have a significant effect on PET quantification for tumour response assessment. In this study, we found that
With
Considering the typical response thresholds which have been used to separate responding patients from nonresponding patients (last column in Table
Many recent studies on early tumour response assessment have used the single-voxel based
We found that the two ROI methods gave rise to highly correlated (
The EORTC [
PET quantification for assessing treatment response using a fixed size ROI is sensitive to the placement of the ROI within the tumour. The difference between the current recommendations favoring ROIpeak (over
This work was supported in part by the Ontario Institute for Cancer Research, a Beatty Fellowship, and an Ontario Graduate Scholarship. The authors thank Dr. Daniel Mozeg and Ms. Maggie Kusano for their scientific contributions. No conflict of interests exists.