Abstract
Background
High-definition (HD) positron emission tomography (PET) image reconstruction is a new image reconstruction method based on the point spread function system, which improves the spatial resolution of the images.
Purpose
To compare the utility of HD reconstruction of PET images for staging lung cancer with that of conventional 2D ordered subset expectation maximization + Fourier rebinning (2D) reconstruction.
Material and Methods
Thirty-five lung cancer patients (24 men, 11 women; median age, 66 years) who underwent surgery after 18F-2-deoxy-fluoro-D-glucose (FDG)-PET-CT were studied. Their PET data were reconstructed with 2D and HD PET reconstruction algorithms. Two radiologists individually TNM staged both sets of images. They also evaluated the quality of the images and the diagnostic confidence that the images afforded them using 5-point scales.
Results
T, N, and M stages were correctly diagnosed on both the 2D and HD reconstructed images in 23 (66%), 25 (71%), and 30 (86%) of 35 cases, respectively. Overall TNM stage was correctly diagnosed on both types of reconstructed images in 23 cases (66%), underestimated in three (9%), and overestimated in nine (26%). No significant difference in T, N, or M stage or overall TNM stage was observed between the two reconstruction methods. However, the HD reconstructed images afforded a significantly higher level of diagnostic confidence during TNM staging than the 2D reconstructed images and were also of higher quality than the 2D reconstructed images.
Conclusion
Although HD reconstruction of FDG-PET images did not improve the diagnostic accuracy of lung cancer staging compared with 2D reconstruction, the quality of the HD reconstructed images and the diagnostic confidence level they afforded the radiologists were higher than those of the conventional 2D reconstructed images.
Keywords
It is necessary to stage lung cancer accurately to decide treatment strategy. With the advent of new treatment modalities such as stereotactic body radiotherapy and radiofrequency ablation, treatment options have increased for stage I non-small-cell lung cancer (1, 2), so confirmation of the absence of lymph node and distant metastases is now especially important. 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) is an indispensable imaging modality for the pre-treatment staging of lung cancer (3, 4). However, this modality has limitations with respect to its spatial resolution and the fact that uptake on FDG-PET reflects glucose metabolism (3–6). Although integrated PET-CT is a useful method for compensating for this weakness, providing matched morphological and functional images (3, 7, 8), improving the spatial resolution of PET itself are also necessary (9, 10).
High-definition (HD) PET image reconstruction is a new image reconstruction method based on the point spread function (PSF) system, which improves the spatial resolution of the images to approximately 2 mm. Thus, HD PET reconstruction may solve the problem of the insufficient spatial resolution of FDG-PET, but there have been very few reports about this new image reconstruction approach (11).
Therefore, in this study we compared the utility of images produced using this reconstruction approach for lung cancer staging with that of images produced using conventional 2D ordered subset expectation maximization (OSEM) reconstruction. The purpose of this study was to verify the hypothesis that compared with conventional 2D OSEM reconstruction HD PET reconstruction increases the diagnostic confidence of physicians and improves diagnostic accuracy during lung cancer staging.
Material and Methods
Subjects
Thirty-six consecutive lung cancer patients underwent FDG-PET scans for staging and surgical treatment between 2008 and 2009. Of them, one was not fully histologically evaluable because of tumor necrosis induced by chemotherapy prior to surgical treatment and was therefore excluded. Thus, 35 patients were analyzed in this study. The patients ranged in age from 35 to 82 years (median, 66 years) and included 24 men and 11 women. The histological subclassifications of the tumors were as follows: 22 adenocarcinomas, eight squamous cell carcinomas, two large cell carcinomas, one carcinoid tumor, one adenosquamous cell carcinoma, and one pleomorphic carcinoma.
PET-CT image acquisition and reconstruction
A Biograph 40 True Point PET/CT scanner (Siemens Medical Solutions USA, Knoxville, TN, USA) was used to acquire the FDG-PET images. All patients fasted for at least 6 h before the PET-CT examination. The scans were performed 90 min after the intravenous injection of 3.7 MBq 18F-FDG per kg of body weight. The PET scans were obtained using the following parameters: 3D acquisition mode; 168 × 168 matrix; trans-axial field of view (FOV), 605 mm; axial FOV, 216 mm; acquisition time, 2 min per bed position in 29 cases and 2.5 min per bed position in six cases; and slice thickness, 2 mm. At our institution, the mean trans-axial resolutions of the 2D and HD reconstructed images at full width at half maximum (FWHM) were 2.8 and 1.6 mm, respectively, and the mean axial resolutions of the 2D and HD reconstructed images at FWHM were 4.2 and 2.3 mm, respectively. Images of CT for attenuation correction (CTAC) were obtained with the following parameters: 120 KVp; 100 mA (50 mAs CARE DOSE4D); pitch factor, 1.2; rotation time, 0.5 s; trans-axial scan field, 70 cm; 512 × 512 matrix; and slice thickness, 2 mm. No contrast material was used for CTAC. The PET images were reconstructed with the 2D OSEM and Fourier rebinning (OSEM + FORE), and the HD PET reconstruction (3D-OSEM + PSF) algorithms.
Image interpretation
Two chest radiologists (MH and YO) independently interpreted the 2D OSEM + FORE and HD reconstructed PET images to TNM stage the patients' tumors. The interpreted images consisted of 2-mm-thick axial and coronal images, maximum intensity projection (MIP) images derived from PET images, 2-mm-thick axial CTAC images, and fusion images composed of the PET and CTAC images.
When diagnosing tumor (T) status, we measured the diameter of the primary tumor on CTAC images and applied these results to both the 2D and HD PET reconstructed images. Tumor invasion into the adjacent organs and the chest wall was examined on the PET and CTAC images. FDG uptake that clearly extended to the adjacent structures on PET, in addition to CTAC findings suggestive of tumor invasion, was interpreted as a finding of invasion. We also measured the maximum standardized uptake value (SUV) of the primary tumors on MIP images. When diagnosing nodal (N) status, increased 18F-FDG uptake compared with the surrounding mediastinal level was interpreted as a positive result, and bilateral symmetric hilar uptake was interpreted as physiologically normal (negative). We used the CTAC images to judge nodal stations. Metastasis (M) status was assessed using a combination of CTAC findings and FDG-PET uptake.
The radiologists also scored image quality and the level of diagnostic confidence afforded by the images regarding TNM status as described below. The level of diagnostic confidence afforded by the images regarding TNM status was assessed using the following scale: grade 1, 0–20% (least sure about the diagnosis); grade 2, 21–40%; grade 3, 41–60%; grade 4, 61–80%; and grade 5, 81–100% (most sure about the diagnosis). Image quality was assessed using the following scale: grade 1, the margins of the spine and ribs were unclear; grade 2, intermediate between grades 1 and 3; grade 3, the margins of the spine and ribs were fairly clear; grade 4, intermediate between 3 and 5; and grade 5, the margins of the spine and ribs were very clear. We used the seventh edition of the TNM classification for lung cancer developed by the Union Internationale Contre le Cancer and the American Joint Committee for Cancer Staging. Each TNM factor and the overall stage were determined by consensus.
To determine the pathological TNM stage, the patients' surgical and pathological results were examined. All patients underwent thoracotomy at 9–53 days (mean, 30 days) after the PET-CT examination. During the thoracotomy, both the N1 and N2 nodes were resected in 32 cases, the N1 node alone was resected in one case, and no node resection (ND0) was performed in two cases. Distant metastases (M stage) were suspected on PET-CT in six patients (2: M1a, 4: M1b) and we followed up the lesions for at least 1 year with various examinations including PET-CT, CT, MRI, and tumor markers. Then, we judged whether the lesion was metastatic or not. We also followed up the nodal status of the three patients who did not undergo mediastinal nodal resections for 1 year or longer and assessed whether nodal metastases had been present at the initial PET/CT scan.
Statistical analysis
The McNemar test was used to compare the accuracy of the TNM factor staging and overall TNM staging between the 2D and HD PET reconstructed images. The Wilcoxon signed rank test was also used to compare the diagnostic confidence level of the radiologists regarding TNM staging and image quality between the two types of images. P values <0.05 were considered to indicate a significant difference.
Results
T, N, and M status on 2D reconstructed images, on HD reconstructed images, and at the final diagnosis in 35 lung cancer patients
Stage determined by 2D reconstructed images, HD reconstructed images, and pathological and follow-up examinations in 35 lung cancer patients
Accuracy of TNM factor and overall staging on 2D and HD reconstructed images
The HD reconstructed images afforded the assessors a significantly higher level of diagnostic confidence than the 2D images during TNM staging. A similar result was obtained for the image quality evaluation (Figs. 1 and 2).
The level of diagnostic confidence regarding TNM status and image quality. (a, b, c) The HD reconstructed images afforded the radiologists with a significantly higher level of confidence during TNM staging than the 2D images; (a) T factor; P = 0.0001, (b) N factor; P = 0.0003, and (c) M factor; P = 0.013). (d) The HD reconstructed images were of higher quality than the 2D reconstructed images (P < 0.0001) Coronal PET images of a 36-year-old man with squamous cell carcinoma of the left lower lobe that were indicative of T2bN2M0 (IIIA) stage disease. The contours of the primary tumor (arrow head) and the N2 lymph node (arrow) were less clear on the conventional 2D images (a, c) than on the HD reconstructed images (b, d). The clinical stage was diagnosed as T3N2M0 (stage IIIA) on both types of images. The median TNM status confidence levels of the two radiologists were 4.5, 3, and 3.5, respectively, on the 2D images, and 4.5, 4, and 3.5, respectively, on the HD reconstruction images. The median image quality value was 4 on the 2D images and 5 on the HD reconstructed images
Discussion
Our study showed that although the accuracy of lung cancer staging using HD reconstructed images was similar to that using conventional 2D OSEM + FORE images, HD reconstruction provided the radiologist with significantly higher levels of diagnostic confidence and image quality than 2D reconstruction. The fact that the HD reconstructed images did not improve the diagnostic accuracy of lung cancer staging can be partly explained by the fact that T stage was mainly determined by measuring tumor size on CT, and 77% of the patients in our study displayed N0 disease. However, as our results showed that the HD reconstructed images were of higher quality and increased the diagnostic confidence of the radiologists, it might be easier to interpret HD reconstructed images than 2D reconstructed images during lung cancer staging.
We could not find any article about the accuracy or benefits of the HD PET reconstruction technique for lung cancer staging. In the present study, we used the standard TNM staging method with FDG-PET, i.e. PET/CT-based interpretation. The accuracy of PET/CT for diagnosing N stage has been reported to range from 78% to 93% (7, 8, 12, 13), which is slightly higher than that detected in our study. A recent study correctly predicted T status in 82% of cases (378/463) using PET/CT (14), which was higher than the correct prediction rate in our study (23/35, P = 0.028 according to Fisher's exact test).
HD PET reconstruction might improve the quality of images used for routine lung cancer staging and increase the diagnostic confidence of radiologists. HD PET reconstruction can provide an improved image quality due to its high spatial resolution. As a result, it could become easier to separate two or more adjoining lymph nodes, especially when lymph nodes extend over multiple stations. In our study, 77% of the patients had N0 disease, and 63% of patients had stage IA or IB disease. Further studies with a larger number of patients with advanced stage cancer are needed. The schedule of PET examinations is tight and HD reconstruction requires a slightly longer time than 2D reconstruction. However, this extra time is not too long for the examination schedule to be changed.
In conclusion, although HD PET reconstruction did not improve the diagnostic accuracy of lung cancer staging compared with conventional 2D OSEM + FORE images, the HD PET reconstructed images increased the diagnostic confidence level of the radiologists and were of higher quality than the conventional 2D OSEM images. Our method might make it easier to stage lung cancer using FDG-PET/CT.