Evaluation of image quality at the detector’s edge of dedicated breast positron emission tomography: Can small breast cancer near the chest wall be detected ? CURRENT

Purpose : We assessed image quality of dedicated breast positron emission tomography (dbPET) at the detector's edge by phantom and clinical studies. Methods: A breast phantom with four spheres (16, 10, 7.5, and 5 mm in diameter) was filled with 18 F-fluorodeoxyglucose solution of sphere-to-background ratio was 8:1. It was positioned such that the spheres were five different positions from the top edge to the centre of the detector and scanned for 5 min in each position. Reconstructed images were visually evaluated, and % background variability ( %N 5mm ), % contrast ( %Q H ,5mm ), contrast-to-noise ratio ( Q H ,5mm / N 5mm ), and coefficient of variation of the background ( CV background ) were calculated. Next, tumour-to-background ratios (TBRs) between breast cancer near the chest wall (close to the detector’s edge; peripheral group) and at other locations (non-peripheral group) were compared. The TBR of each lesion was also compared between dbPET and PET/computed tomography (CT). Results: As closer to the detector’s edge, the %N 5mm and CV background increased and %Q H ,5mm and Q H ,5mm / N 5mm decreased in the phantom study. The disadvantages of this placement were visually confirmed. With regard to clinical images, TBR of dbPET was significantly higher than that of PET/CT in both the peripheral (12.1±6.2 vs. 6.5±3.4, p =0.0001) and non-peripheral (13.1±7.1 vs. 7.7±7.4, p =0.0004) groups. There was no significant difference in TBR of dbPET between the peripheral and non-peripheral groups (12.1±6.2 vs. 13.1±7.1, p= 0.6367). Conclusion : In the phantom study, the image quality decreased closer to the detector’s edge than at a depth of 1/8. In clinical studies, however, the lesion detectability of dbPET was the same even if the lesion was close to the detector’s edge or not, and it was higher than that in PET/CT. dbPET has a great potential for detecting breast lesions near the chest wall even in young women with small breasts.


Introduction
F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) has become one of the most useful tools in diagnostic imaging for cancer. Many studies have demonstrated the efficacy of whole-body FDG-PET/CT in staging or re-staging, in monitoring the 3 response to therapy, and for predicting the prognosis of patients with breast cancer [1][2][3]. It is important to detect breast cancer at an early stage when the lesions are it is small, since mortality increases with tumours exceeding 1 cm in size [4,5]. However, detection of small breast cancers by whole-body PET/CT is challenging because of its limited spatial resolution [6]. High-resolution dedicated breast PET (dbPET) scanners have been developed to detect small breast lesions. There are two types of high-resolution dbPET, i.e. positron emission mammography (PEM) and a tomographic technique using a ring-shaped scanner (ring-shaped dbPET) [7]. PEM systems depict breast tissue via soft compression of the breast with two opposing plate-like detectors and have higher sensitivity than whole-body PET/CT [8][9][10] while ring-shaped dbPET scanners can visualise breast cancer more clearly than whole-body PET/CT [11,12]. These high-resolution breast PET systems have greater photon sensitivity and can improve spatial resolution by setting the detector close to the breast, reducing respiratory movement, and using smaller detection units with reconstruction methods that are different to those used for whole-body PET/CT. Their performances have been evaluated using NEMA-NU4-2008 standards [13], and the physical parameters of dbPET and whole-body PET/CT have been compared using a common breast phantom [14]. In that comparative study, the breast phantom was located at the centre of each scanner, and no studies have reported on the quality of dbPET images close to the edge of the detector. However, many Japanese women have small breasts, and their mammary glands are often located near the chest wall, close to the edge of detector, even when they are in the prone position. This tendency is particularly common in young women who are less likely to have breast ptosis than older women. Therefore, it is necessary to evaluate the consequences of a shift in the position of the breast phantom away from the centre of the detector. This study aimed to confirm the image quality of dbPET at the edge of the detector by phantom and clinical studies.

Methods
This single-institution study was approved by the institutional review board and ethics committee of our institute in accordance with the Declaration of Helsinki; written informed consent was obtained from each patient for access to their data.
Ring-shaped dbPET scanner 4 The ring-shaped dbPET scanner (Elmammo, Shimadzu Corp., Kyoto, Japan) consists of 36 detector modules arranged in three contiguous rings, has a diameter of 195 mm and an axial length of 156.5 mm, and has depth-of-interaction measurement capability [15]. The transaxial effective field-ofview (FOV) is 185 × 156.5 mm 2 . Each detector block consists of a four-layered 32 × 32 array of lutetium oxyorthosilicate crystals coupled to a 64-channel positron-sensitive photomultiplier tube via a light guide. Attenuation correction was calculated using a uniform attenuation map with object boundaries obtained from emission data [16]. Scatter correction was performed using the convolution-subtraction method [17] with kernels obtained by background tail fitting. The characteristics and standard performance of this scanner have been reported in detail previously [13].
Whole-body PET/CT scanner PET/CT scans were obtained using a Biograph Horizon TrueV FDG-PET/CT system (Siemens Medical Solutions, Knoxville, TN, USA). This system has 52 detector rings consisting of 160 blocks, with each block containing an array of 13 × 13 lutetium oxyorthosilicate crystals (4 mm × 4 mm × 20 mm) covering an axial FOV of 221 mm and a transaxial FOV of 690 mm. A CT scan was performed for attenuation correction (130 kV; 15-70 mA; tube rotation time, 0.6 s per rotation; pitch, 1; a transaxial FOV, 700 mm; and section thickness, 5 mm).

Development and preparation of the breast phantom
A cylindrical breast phantom containing four plastic spheres of different diameters was used. The inner and outer diameters of the cylinder were 100 mm and 140 mm, respectively, and the height was 170 mm. The diameters of the spheres arranged inside were 5, 7.5, 10, and 16 mm. The cylinder and four spheres were filled with 18 F-FDG solution at a sphere-to-background radioactivity ratio of 8:1 in accordance with a previous study [14]. The background radioactivity at the start of data acquisition by dbPET was set to 2.46 kBq/mL. One scan was performed under each condition.

Data acquisition and image reconstruction
The breast phantom was positioned such that the spheres were precisely located in the same axial plane at 8 mm, 13 mm, 19.5 mm (depth of 1/8), 39 mm (depth of 1/4), and 78 mm (depth of 1/2, the centre of the detector) below the top edge of the detector (Fig. 1). Sphere placement at each position in the detector was confirmed visually and by measurement on the image. The dbPET images were 5 reconstructed using a three-dimensional list mode dynamic row-action maximum-likelihood algorithm with one iteration and 128 subsets, a relaxation control parameter of β = 20, and a matrix size in the axial view of 236 × 200 × 236 with a post-reconstruction smoothing Gaussian filter (1.17-mm FWHM).
Tight or just mode attenuation correction using a uniform attenuation map with object boundaries obtained from the emission data was performed on phantom or clinical dbPET images, respectively.
The scatter correction method used was the convolution-subtraction method with kernels obtained by background tailfitting [17].
The clinical PET/CT images were reconstructed using the ordered subset expectation maximisation method and the time-of-flight algorithm with four iterations and 10 subsets. The CT data were resized from a 512 × 512 matrix to a 180 × 180 matrix to match the PET data and construct CT-based transmission maps for attenuation correction of the PET data with a post-reconstruction smoothing Gaussian filter (5 mm FWHM).

Analyses of phantom image quality
Visual analyses of the phantom images were performed using syngo. via VB10 (Siemens Healthcare GmbH, Erlangen, Germany). An experienced nuclear medicine physician and two experienced PET technologists evaluated the hot spheres. Evaluations were performed using the slices displayed in the transverse image slice containing the centres of the spheres. The images were displayed in an inverse grey scale with a standardised uptake range of 0-6. The 5-mm-diameter hot sphere was visually graded as follows: 2, identifiable; 1, visualised, but similar hot spots observed elsewhere; and 0, not visualised. Spheres with visual scores ≥ 1.5 were deemed to be detectable. The final score was the mean of the scores from three readers. The visual assessment was performed based on the Japanese guideline [18]. Physical analysis was also performed using syngo. via VB10. The coefficient of variation of the background (CV background ), % background variability (%N 5mm ), % contrast (%Q H,5mm ), and contrast-to-noise ratio (Q H,5mm /N 5mm ) were calculated. The CV background was calculated by evaluation of various regions of interest (ROIs) in the transverse image slice that contained the centres of the spheres Ten ROIs with a diameter of 16 mm were placed in the background region in that slice and ± 5-mm-adjacent slices (30 ROIs in total). %Q H,5mm , %N 5mm , and 6 their ratio (%Q H,5mm /N 5mm ) were also calculated by evaluation of various ROIs. The 12 ROIs that were 5 mm in diameter were placed on the background region in that slice and ± 5-mm-adjacent slices (36 ROIs in total). %Q H,5mm and %N 5mm were used as measures for the image contrast and noise for the sphere, and their ideal values were 100% and 0%, respectively. These physical values were calculated according to a previous report [19]. All PET images were evaluated separately by two experienced nuclear medicine physicians (with 16 and 7 years of experience in interpreting PET, respectively). Of these 62 lesions, those on the chest wall side from the boundary line obtained in the phantom test were defined as the "peripheral group", and those on the nipple side were defined as the "non-peripheral group". Non-mass uptakes, other than focus and mass-like uptakes, were excluded because their quantitative reliability could not be established.

Analysis of human images
Tumours that were exactly centred in both marginal and non-marginal regions and whose volume was equally present in both regions were also excluded.
The quantitative value of PET is known to be affected by the partial volume effect [20]. To account for lesion size bias, propensity matching was performed to compare the peripheral and non-peripheral groups. The non-peripheral group was reorganised such that lesion size matched the peripheral group in a one-to-one correspondence. As a result, 23 lesions in each group (total 46 lesions) were included in the final analysis.
To evaluate lesion detectability in dbPET depending on the position of the tumour, tumour-to-background ratio (TBR) was calculated as follows. First, the smallest spheroid volume of interest (VOI) that just contained the tumour was placed on the monitor. Second, 5 (or 6) spherical VOIs with a diameter of 5 mm were placed on the top, bottom, left and right, and anterior (and the posterior for non-peripheral groups) as close as possible to the tumour as illustrated in Fig. 2. TBR was the maximum standardised uptake value (SUV max ) of the VOI on the tumour divided by an average SUV mean of the five (6) VOIs on the background. The TBRs were compared between dbPET and PET/CT images, and the TBR of dbPET was compared between the peripheral and non-peripheral groups.
Additionally, the SUV max and the SUV peak (maximum average SUV within a 1-cm 3 spherical volume) were measured and compared between groups and between devices.

Statistical analysis
A paired t-test was used to compare the TBR of dbPET and whole-body PET/CT for the peripheral and non-peripheral groups, respectively. The Mann-Whitney U test was used to test for differences in TBR on dbPET between peripheral and non-peripheral lesion groups. The correlations between SUV max and SUV peak on dbPET and on WB-PET/CT were evaluated using Pearson correlation coefficients. Statistical significance was defined as p < 0.05. Additionally, for these PET measurements, the interclass correlation coefficients (ICC) were used to evaluate the reliability between readers.

Results dbPET phantom studies
Breast phantom images of the breast phantom scanned by dbPET at the five different positions are shown in Fig. 3. In the qualitative evaluation, the visual scores recorded by a nuclear medicine physician and two nuclear medicine technologists on the dbPET images at 8 mm, 13 mm, 19.5 mm (depth of 1/8), 39 mm (depth of 1/4), and 78 mm (depth of 1/2, the centre of the detector) below the top edge of the detector were 0, 0.33, 1.67, 2, and 2, respectively (Table 1) Fig. 4A). The %N 5mm and CV background increased and %Q H,5mm and Q H,5mm /N 5mm decreased when the phantom was placed closer to the edge of the detector (Table 1, Fig. 4B). Image degradation closer to the edge of the detector was confirmed by r=0.96, p<0.0001]) ( Figure 5). The TBR of dbPET was significantly higher than that of whole-body PET/CT in both peripheral and non-peripheral groups (p<0.0001, Figure 6A). There was no significant difference in the TBR of dbPET between the peripheral and non-peripheral groups (p=0.6367, Figure   6B). Figure 7 shows representative cases of peripheral and non-peripheral breast cancer acquired by dbPET and PET/CT. The breast cancers were visualised on dbPET more obviously than on PET/CT regardless of the location of the lesion (peripheral or non-peripheral).

Discussion
In this study, we evaluated the image quality obtained at different locations within the detector for ring-type breast PET. In the phantom study, the closer to the top of the detector, the higher the CV background and %N 5mm , and the lower the %Q H,5mm and %Q H,5mm /N 5mm were. These results indicated that the quantitative measurements were almost equal except for the end of one-eighth of the detector (about 2 cm from the end of the detector).
Minoura et al. reported that the dbPET images show high levels of noise at the edge of the detector (the bottom of the detector or the chest wall side) and showed the relationship between the slice position in the dbPET image and the standard deviation of noise [21]. Our results showing that the dbPET image quality decreases at 1/8 of the detector edge are consistent with their reports. The geometric efficiency by Monte Carlo simulation at a detector depth of 1/8 was calculated as 0.2, which is considerably lower than that at the centre, which was 0.65. Usually, whole-body PET scans use overlapping acquisition beds to correct for reduced sensitivity at the detector edges. Acquisition of data in overlapped regions can improve quantitative accuracy [22,23]. However, since the dbPET scanner is fixed and cannot use overlapping acquisition to improve image quality near the edges of the detector, there are concerns that important information could not be identified. Additionally, the radioactivity of out of FOV, among which myocardial uptake may be most significant, would also significantly affect image quality. SUV max and SUV peak obtained from dbPET and PET/CT images were highly correlated in this clinical study, regardless of the tumour location. This means that these quantitative values of the tumour closer to the edge of dbPET detector were ensured.
Based on the phantom test results, the detectability of clinical dbPET images was compared between those whose lesions were located up to 2 cm from the upper edges of the detector and the other participants. Contrary to the phantom study results, there was no significant difference in TBR between the two groups. This might be because (i) the clinical dbPET images are observed in multiple directions in the tomographic image and (ii) there were cases with higher TBR than that in the prior phantom study due to the low background physiological uptake in fatty mammary glands.
Additionally, TBR in both groups was significantly higher than that in PET/CT. dbPET is a higherresolution scanner than conventional whole-body PET/CT and respiratory movements are significantly suppressed in whole-body PET/CT scans, so even if the lesion is located at the edge of the detector, dbPET may show higher detectability than PET/CT which uses overlapping acquisition. Positive and linear correlations were observed between SUV max and SUV peak of PET/CT and dbPET.
dbPET achieves higher sensitivity and resolution than whole-body PET/CT by i) reduction of respiratory movement of the breast by acquisition in the prone position, and ii) bringing the detector close to the breast. The 4-layer depth-of-interaction (DOI) detector used in dbPET can maintain sensitivity and resolution at the edges of the transverse field of view [24,25]. Although the background mammary gland shows physiological FDG uptake, the contrast of the lesion and background is higher than that of PET/CT. As a result, the 2018 edition of the Japanese Guidelines for the Practice of Breast Cancer newly describe the use of high-resolution breast PET as a supplemental modality for breasts with high density on mammography, and dbPET is expected to be applied to young women who often have high-density breasts. Both dbPET and PEM have the disadvantage that, due to their structural features, a part of the mammary gland near the chest wall is in the blind area and the lesion may be outside the field of view. However, this study suggests that if the lesion is within the field of view of dbPET, it can be detected with high probability even at the edge of the detector.
Our study had several limitations. First, the phantom was scanned only once for each position. The reproducibility of the findings would have been better if the average results of several scans under each condition were calculated. Second, the study design was retrospective, and the patient cohort was small. In this clinical study, because only histologically proven breast cancers were included, small breast cancers near the edge of the detector that are false-negative on PET may not be sufficiently evaluated. Studies with larger populations and considerations including histology and subtypes of breast cancer will be required to address these limitations.

Conclusion
In our phantom study, based on the image quality at the depth of 1/8, the quality decreased when the phantom was closer to the edge of the detector. In the clinical studies, however, lesion detectability was the same regardless of whether the lesion was close to the edge of the detector or not, and the detectability in both conditions was higher than that in PET/CT.

Funding
This research did not receive any specific grant from funding agencies in the public, or not-for-profit sectors.

11
The authors declare that they have no conflict of interest.

Ethical Approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee (include name of committee + reference number) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.This study was approved by the Ethics Committee of Kofu Neurosurgical Hospital and Yamanashi PET imaging clinic.

Informed Consent
We have obtained informed consent from all the patients included in the study.    Correlation of quantitative parameters between dbPET and whole-body PET/CT. The graphs show the correlation of SUVmax and SUVpeak of dbPET with those of whole-body PET/CT in the peripheral group (a, c) and the non-peripheral group (b, d). In both peripheral and nonperipheral groups, the SUVmax and SUVpeak of dbPET were higher than those of PET/CT with high correlation.

Figure 6
Comparisons of tumour-to-background ratios (TBRs). There were significant differences between TBRs of dbPET and whole-body PET/CT in both peripheral (p<0.0001) and nonperipheral (p<0.0001) groups (a). There was no significant difference in the TBR of dbPET between the peripheral and non-peripheral groups (p=0.6367, b). dbPET and whole-body PET/CT images, it was obvious on dbPET with higher TBR than that on whole-body PET/CT, regardless of the peripheral or non-peripheral location.