Introduction
Breast cancer is the most frequent malignancy in women in Western countries and the second leading cause of cancer-related deaths among women. Axillary lymph node status is the most important prognostic factor for recurrence and survival. Therefore, accurate staging at the time of initial diagnosis is crucial. PET using
18F-FDG has been used for staging, restaging and therapy monitoring in a variety of cancer types, including breast cancer [
1‐
5]. In newly diagnosed breast cancer,
18F-FDG PET is not recommended for routine staging of axillary lymph nodes because its sensitivity is too low [
1]. A recent systematic review found that the sensitivity of PET/CT systems in the detection of axillary nodal metastases ranges from 44 % to 67 % [
6]. This low sensitivity is due in part to the limited spatial resolution of PET systems, leading to partial volume effects (PVE) that cause significant underestimation of the radioactivity concentration in lesions smaller than two to three times the spatial resolution of the system. Consequently, small cancer deposits and especially micrometastases (<2 mm) are very unlikely to be detected.
In recent years, major hardware and software improvements have been implemented in PET imaging. In particular, advanced reconstruction algorithms that model the point spread function (PSF) of a system have recently become commercially available [
7,
8]. PSF reconstruction improves spatial resolution throughout the entire field of view (FOV), reduces PVE and improves image contrast. As a result, newer-generation clinical PET systems equipped with such algorithms can be expected to detect small-volume metastases better.
The aim of this prospective study in single referral centre was to evaluate the impact of PSF reconstruction on quantitative values and the diagnostic accuracy of 18F-FDG PET for the axillary staging of breast cancer patients, as compared with a conventional algorithm, OSEM (ordered subsets expectation maximization). PET results were compared with pathological results with special emphasis on the size of nodal metastases.
Discussion
Evaluation of new technologies that are being implemented in PET imaging is needed. Some of them such as advanced reconstruction algorithms are likely to not only improve diagnostic performance but also change quantitative and image features, requiring new diagnostic thresholds to be defined. PSF reconstruction is a new reconstruction algorithm available from all major vendors of whole-body PET/CT systems (namely TrueX from Siemens Healthcare [
17], SharpIR for GE Healthcare [
18] and Astonish TF from Philips Healthcare) which improves spatial resolution and is therefore expected to lead to the detection of smaller metastases than can be achieved by conventional algorithms such as OSEM. So far, PET imaging has failed to solve the clinical issue of proper axillary lymph node staging in breast cancer patients at least in part because of its limited spatial resolution, and could therefore benefit from PSF reconstruction. In this prospective study in a single referral centre, the use of PSF reconstruction led to an increase in SUV
max of 66 % on average (Fig.
1) as compared with OSEM reconstruction. PSF reconstruction was able to detect more involved nodes and to improve PET sensitivity in the detection of axillary nodal involvement, especially in patients in whom the largest nodal metastasis was <7 mm.
The implementation of PSF reconstruction improves spatial resolution in a more important manner at the edges of the FOV where the PSF broadens because of the oblique penetration of 511-keV photons into scintillation crystals. Evaluation of the FWHM of our system in the geometry of a human breast and axilla PET examination showed a strong improvement thanks to PSF reconstruction (median radial FWHM: 2.50 mm) as compared to a conventional OSEM algorithm (median radial FWHM: 6.34 mm). In addition, PSF reconstruction minimized PVE and would be expected to improve activity recovery more importantly in lesions smaller than twice the spatial resolution of the PET system. Therefore, the improvement in quantitative values and the ability to detect small lesions would be expected to be higher in the axillae, as compared with more centrally located malignancies and/or larger tumours. Indeed, the 66 % improvement in SUV
max in this study can be compared with the 48 % improvement observed in a previous study dealing with thoracic lymph node staging in patients with non-small-cell lung cancer (NSCLC) that used the same PET system and acquisition/reconstruction parameters [
8].
In our study, PET with PSF reconstruction detected more involved nodes than PET with OSEM reconstruction . This is of importance, as the number of involved nodes is itself a prognostic factor [
19]. In addition, this result strengthens the findings of Vinh-Hung et al. [
5], demonstrating that PET may be a powerful tool for distinguishing patients with a low versus those with a high burden of lymph node involvement. PET with PSF reconstruction did not detect involved nodes outside Berg I and II levels that would have been overlooked by PET with OSEM reconstruction . Detecting these nodes, which are usually not addressed during an axillary clearance procedure, is valuable and can change a patient’s management [
20].
PSF reconstruction performed better than OSEM reconstruction in detecting nodal metastases <7 mm. We chose the median size of all nodal metastases in our series as a cut-off value, but it is noteworthy that 7 mm was also roughly twice the spatial resolution of our PET system, a size below which PVE is significant. To the best of our knowledge, there is only one study that has compared PET results to the exact size of the intranodal metastases [
2]. This is of importance in breast cancer, a malignancy in which a given nodal metastasis is frequently smaller than the involved node by itself. In a recent study evaluating diagnostic full-dose
18F-FDG PET/CT for axillary staging of breast cancer patients [
2], 10 out of 61 included patients were false-negative. In these patients, apart from an overlooked 24-mm nodal metastasis immediately adjacent to a primary tumour, nodal size ranged from 0.8 to 6 mm (mean 3 mm). In our study, we took into account the size of the largest nodal metastasis per axilla, because this is the lesion most likely to be detected by PET. When taking into account all nodal metastases to compare our results with those of Heusner et al. [
2], who used a similar methodology, the size of the metastases in the five patients in whom both PSF and OSEM PET were false-negative ranged from 1 to 9 mm (mean 3.3 mm). In the three patients in whom PSF reconstruction was true-positive while OSEM reconstruction was false-negative, the size of the largest metastases ranged from 1.8 to 8 mm (mean 4.9 mm). The median size of the largest nodal metastasis per axilla was lower in the PSF+/OSEM− group than in the PSF+/OSEM+ group (Fig.
6). Yet in the PSF+/OSEM+ group there were some small metastases (Fig.
6). This illustrates the fact that the ability of PET imaging to detect small cancer deposits depends not only on spatial resolution, but also on other factors such as
18F-FDG avidity and contrast between lesion and background.
There is an ongoing debate as to the potential role of
18F-FDG PET for initial staging of breast cancer [
15,
21‐
23]. Despite a low sensitivity,
18F-FDG PET is generally reported to have a good specificity. Some authors advocate the use of
18F-FDG PET to reduce the use of sentinel node biopsy (SNB) [
21,
23] (i.e. if findings are positive in the axilla, SNB is no longer required and ALND can be performed immediately), while others suggest that
18F-FDG PET should be used to extend the use of SNB [
2,
15]. Regarding the latter option, Heusner et al. [
2] suggested that in patients with a high risk of axillary lymph node metastases, an unremarkable
18F-FDG PET scan could help identify a subgroup of patients who can safely undergo SNB. It is noteworthy that the first strategy is based on a good specificity of PET, generally reported to be higher than 80 %. Similar to the study by Lasnon et al. [
8] in patients with NSCLC, we found that PET with PSF reconstruction improves sensitivity at the expense of a slightly lower specificity, probably because PSF reconstruction improves activity recovery in nodes with moderate uptake because of benign disease. In our study, PET with PSF reconstruction had a higher sensitivity in patients with a primary tumour ≤30 mm, whereas both reconstructions performed equally well in patients with a primary tumour >30 mm, a size above which the risk of macrometastases is higher (Table
2). In addition, the LR− was lower with PSF reconstruction than with OSEM reconstruction, and Biggerstaff graphical comparison showed PSF reconstruction to be globally superior to OSEM reconstruction (Fig.
5). Therefore, altogether these data suggest that the use of PET with PSF reconstruction could enable SNB to be performed more safely in patients with a primary tumour ≤30 mm and with unremarkable PET results in the axilla.
Finally, it is noteworthy that this study and others evaluating the impact of PSF reconstruction or a combination of PSF reconstruction and time-of-flight on quantitative values in oncology [
7,
8,
24,
25] have been performed at a time when many efforts are being made to harmonize SUV values in multicentre trials [
26‐
28]. The use of different generation PET systems in which FDG PET is used for therapy monitoring in breast cancer patients could lead to inaccurate response evaluation. If, for example, a patient underwent a pretreatment scan on a PET system using a conventional algorithm and a posttreatment scan on a PET system equipped with PSF reconstruction, response would be incorrectly minimized. This may occur in centres running two or more PET systems or updating their equipment during the course of a trial. The use of PSF reconstruction may also be an issue when pooling SUVs coming from PET systems of different generations to determine whether the SUV of metastatic nodes is a prognostic factor [
29]. A solution to overcoming these problems is to harmonize SUV using an additional filtering step [
30] or by generating two sets of images, one to provide optimal diagnostic quality and the second to meet quantitative harmonizing standards [
31].
Conclusion
In this prospective study in a single referral centre, the use of PSF reconstruction led to an increase in SUVmax of 66 % on average compared with OSEM reconstruction, detected more involved nodes and improved PET sensitivity in the detection of axillary nodal involvement, especially in patients in whom the largest nodal metastasis was <7 mm. Although the sensitivity of PET with PSF reconstruction appears to be insufficient for it to replace surgical approaches for axillary staging, our data suggest that the use of PET with PSF reconstruction could allow SNB to be performed more safely in patients with a primary tumour ≤30 mm and with unremarkable PET results in the axilla.