Lesion-to-background ratio threshold value of SUVmax of simultaneous [68Ga]Ga-PSMA-11 PET/MRI imaging in patients with prostate cancer

Purpose This study aimed to calculate an applicable relative ratio threshold value instead of the absolute threshold value for simultaneous 68Ga prostate-specific membrane antigen/positron emission tomography ([68Ga]Ga-PSMA-11 PET) in patients with prostate cancer (PCa). Materials and methods Our study evaluated thirty-two patients and 170 focal prostate lesions. Lesions are classified into groups according to Prostate Imaging Reporting and Data System (PI-RADS). Standardized uptake values maximum (SUVmax), corresponding lesion-to-background ratios (LBRs) of SUVmax, and LBR distributions of each group were measured based on regions of interest (ROI). We examined LBR with receiver operating characteristic analysis to determine threshold values for differentiation between multiparametric magnetic resonance imaging (mpMRI)-positive and mpMRI-negative lesions. Results We analyzed a total of 170 focal prostate lesions. Lesions number of PI-RADS 2 to 5 was 70, 16, 46, and 38. LBR of SUVmax of each PI-RADS scores was 1.5 (0.9, 2.4), 2.5 (1.6, 3.4), 3.7 (2.6, 4.8), and 6.7 (3.5, 12.7). Based on an optimal threshold ratio of 2.5 to be exceeded, lesions could be classified into MRI-positive lesion on [68Ga]Ga-PSMA PET with a sensitivity of 85.2%, a specificity of 72.0%, with the corresponding area under the receiver operating characteristic curve (AUC) of 0.83, p < 0.001. This value matches the imaging findings better. Conclusion The ratio threshold value of SUVmax, LBR, has improved clinical and research applicability compared with the absolute value of SUVmax. A higher threshold value than the background’s uptake can dovetail the imaging findings on MRI better. It reduces the bias from using absolute background uptake value as the threshold value.

• A higher threshold value than the background's uptake can dovetail the imaging findings on MRI better. • The specificity of [ 68 Ga]Ga-PSMA PET needs to be further improved.

Background
Prostate cancer (PCa) is a common malignant disease in the elderly male population. Approximately 17% of patients with early prostate cancer have metastatic

Open Access
Insights into Imaging *Correspondence: jing.zhao@charite.de 1 Institute of Radiology and Nuclear Medicine, Charité -Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Charitéplatz 1, 10117 Berlin, Germany Full list of author information is available at the end of the article disease. PCa is the second leading cause of cancer death in men in the western world [1]. Multiparametric magnetic resonance imaging (mpMRI) has been a clinical imaging tool for detecting primary PCa and guiding subsequent biopsy. MpMRI includes T2-weighted imaging (T2WI), diffusionweighted imaging (DWI), apparent diffusion coefficient (ADC), and dynamic contrast-enhanced MRI (DCE-MRI). Prostate Imaging Reporting and Data System (PI-RADS) interprets results [2,3].
PSMA is a transmembrane glycoprotein related to tumor progression and disease recurrence. PSMA overexpresses in prostate cancer cells. It is associated with PCa with higher serum prostate-specific antigen (PSA) levels and a higher Gleason score (GS) [4,5].
Positron emission tomography (PET) images are coregistered with computed tomography (CT) scans. CT is easily acquired and widely available to provide anatomical information about the localization of PSMAavid lesions. Previous studies suggest that [ 68 Ga] Ga-PSMA-11 PET/CT has a high detection rate for prostate tumors, with a sensitivity of 67-97% [6,7]. Koerber et al. [8] and Woythal et al. [7]reported that SUVmax of PCa is higher than that of non-cancerous lesions and healthy prostate tissue. Combining [ 68 Ga] Ga-PSMA-11 PET and mpMRI has the potential to improve localization accuracy and diagnostic efficiency, as Zamboglou et al. proved [9]. In both studies, experts elaborated on the advantages of PET/MRI in the diagnosis of PCa.
Nevertheless, two aspects can be further optimized. First, MRI-positive lesions may show unapparent or low uptake in PET images. MRI-negative lesions may show apparent uptake in PET images. It may misdiagnose part of MRI-negative lesions as positive if we consider all apparent uptake lesions as positive in PET images. Therefore, it is necessary to increase the threshold value, which is higher than the background SUVmax.
Second, in most publications, individual research centers adopt its threshold standard to proceed with studies. The threshold standard varies from different medical centers. Hence, each study is conducted under different execution standards. Eiber et al. [10] took SUVmax higher than the background as a threshold value to prove diagnostic accuracy improvement. Woythal et al. [7] reported the best threshold value of 3.15 with sensitivity 97%, specificity 90%, and area under curve (AUC) 0.987. Donato et al. [11] described lesions as mildly avid (SUVmax < 5), moderately avid (SUVmax > 5), or intensely avid (SUVmax > 10). Hicks et al. [12]calculated a threshold of 6.7, with sensitivity 88%; specificity 96%.
However, SUVmax is affected by a specific combination of radiotracer manufacturer, systems vendor, reconstruction techniques, uptake time, post-processing software, the time between radiotracer injection and scanning, and even the human race. Taking absolute value for research results in bias from different imaging conditions. Therefore, in our study, we used ratio value LBR to perform research.
We aimed to classify prostate lesions according to MRI morphological imaging analysis to achieve a better threshold LBR value. This LBR threshold value matches the imaging findings on MRI better. It reduces the possibility of MRI-negative lesions being misdiagnosed as positive in PET images. It reduces the bias from using absolute background uptake value as the threshold value. We also re-examined clinical follow-up information and subsequent pelvic MRI to verify whether the lesion is radiological positive or negative.

Study population
This retrospective study was approved by the institutional ethics review board (EA1/060/16), and the institutional review board waived the requirement for informed consent for this retrospective analysis.
Inclusion criteria are as follows: (1) patients with biopsy-proven PCa who underwent simultaneous [ 68 Ga] Ga-PSMA-11 PET/MRI between January 2017 and March 2020 in our department; (2) all necessary additional information could be obtained from clinical records; (3) patients underwent pelvic MRI examination at our institution for follow-up analysis. Exclusion criteria are as follows: (1) patients who underwent prostatectomy before scanning; (2) patients whose follow-up information is not adequate.
Imaging was performed with a 3.0 T PET/MRI system (SIEMENS MAGNETOM Biograph mMR, Erlangen, Germany). Every patient uses the same protocol of PET and MRI scanning. The acquisition contains two parts.
First, body PET/MRI from the vertex to mid-thigh was performed with 3 min of PET acquisition in each bed position, each 24 cm. Two six-element body matrix coils placed anteriorly were used in conjunction with two posterior spine clusters to optimize the signal-to-noise ratio (SNR) in the MRI scanner. A Dixon 3D volumetric interpolated breath-hold examination (VIBE) T1-weighted MRI sequence was performed at each bed position and used for the generation of attenuation maps and anatomic allocation of the PET results. Siemens StarVIBE overcomes motion artifacts.
The second part was a dedicated MRI scan of the pelvis, followed by the reconstruction of PET data. Reconstruction was conducted with an ordered subset expectation maximization algorithm (OSEM), with 3 iterations/21 subsets, based on an x-matrix acquisition with a 4-mm Gaussian filter and relative scatter scaling. Attenuation correction was performed using the non-enhanced MRI data. Table 1 summarizes MRI imaging parameters.

Image analysis
Image analysis was performed on a Visage 7.1 Workstation (Visage Imaging GmbH, Berlin, Germany). All mpMRI images were interpreted by a board-certified radiologist with more than fifteen years without access to the PET images, following the PI-RADS criteria, version 2 [16]. The readers classified prostate focal lesions with PI-RADS scores of 2 and 3 as MRI negative, while 4 and 5 as MRI positive. The present analysis excluded PI-RADS 1 because we do not report PI-RADS 1 lesions. T2WI was used for anatomic correlation for [ 68 Ga]Ga-PSMA-11 PET.[ 68 Ga]Ga-PSMA-11 PET scans were read by a nuclear medicine specialist with more than ten years of experience, who was not aware of the MRI results. ROI was defined as a region with an abnormal signal in MRI images or avid PSMA uptake in PET images. SUVmax is measured based on ROI. Any avid focal lesion in the prostate with uptake above prostate background not attributable to physiologic radiotracer biodistribution was considered positive in [ 68 Ga]Ga-PSMA-11 PET.
Lesions with the same or lower uptake than background were considered negative in [ 68 Ga]Ga-PSMA-11 PET. Besides, background SUVmax was measured in the nearest visually defined normal tissue adjacent to a lesion as background uptake 1.0cm 2 , a perfect circle. LBR is defined as a ratio of lesion SUVmax to background SUVmax. Readers resolved discrepancies based on a separate consensus reading. Both interpreters reviewed all imaging studies in a single session.

Statistical analysis
We classified prostate lesions into four groups according to PI-RADS from 2 to 5 and calculated the LBR of each group. Additionally, we classified LBR into four levels, including LBR ≤ 1, 1 < LBR ≤ 2, 2 < LBR ≤ 3, LBR > 3, and analyzed how does LBR of each PI-RADS group distribute.
To estimate the optimal LBR threshold, we performed ROC analysis and calculation of the AUC. Youden's index defined the optimal cutoff value. Youden's index= sensitivity + specificity − 1. In order to present the threshold's effect on sensitivity and specificity, we also calculated the sensitivity and specificity corresponding to the other six thresholds, besides the optimal threshold. Two-sided p values < 0.05 were considered statistically significant. All statistical analyses were performed using SPSS 25 for Windows (IBM Corp, Armonk, NY). The significance level was set to α < 0.05. Patient demographics and clinical characteristics are summarized using descriptive statistics. Normally distributed data are reported as mean ± SD, and non-normally distributed data are reported as medians (interquartile range, IQR Q1, Q3).

Corresponding LBR analysis
We also calculated the sensitivity and specificity corresponding to the other six thresholds, besides the optimal threshold, to present the threshold's effect on sensitivity, specificity, and Youden's index, summarized in Table 4.   We analyzed 32 patients with prostate cancer undergoing [ 68 Ga]Ga-PSMA-11 PET/MRI. The accuracy of imaging examinations plays a crucial role in diagnosing prostate focal lesions. In our study, LBR revealed the comparison of lesion uptake and background uptake in PET images. LBR ≤ 1 means that lesion uptake is lower or equal to background uptake, as well as negative in [ 68 Ga] Ga-PSMA-11 PET. LBR > 1 represents that lesion uptake is higher than background uptake, as well as positive on [ 68 Ga]Ga-PSMA-11 PET. The higher the ratio, the greater the tendency of a lesion to be PET positive. In the prostate, increased expression of PSMA receptors is not characteristic only of prostate cancer cells, but may also occur in normal prostate cells or non-cancerous lesions such as benign prostatic hyperplasia (BPH), as shown in Fig. 3. Our [ 68 Ga]Ga-PSMA-11 PET findings showed that MRInegative focal lesions might also show a certain degree of PSMA uptake. While some of MRI-positive lesions showed unapparent or mild in PSMA PET, the majority MRI-positive lesions exhibited moderate to strong PSMA avidity, as shown in Fig. 4.
Our lesion-based LBR analysis shows that higher LBR of SUVmax tends to indicate a higher likelihood of malignancy. The higher the PI-RADS score lesions group, the more significant the proportion of LBR > 3. Hence, taking prostate background SUVmax as a threshold value to identify PET positive or negative is relatively low. Our results suggest that the use of prostate background's SUVmax as a threshold value for differentiating MRI negative from MRI-positive prostate lesions can cause in a relevant number of false-positive cases. LBR is defined as a ratio of lesion SUVmax to background SUVmax. For metastasis, background SUVmax is the uptake value of nearby normal tissue, including normal bone tissue and normal soft tissue. These normal tissues usually do not show PSMA avid uptake. The background SUVmax of metastasis is relatively low. The difference between lesions and background is more pronounced. Therefore, the optimal LBR threshold of 2.5 can also be used for metastasis.
Both  [17]. Hirmas et al. reported high performance for the detection of lymph node metastasis and bone metastasis. It revealed a significantly higher concordance rate of 90%, compared to the bone scan of 75%, MRI of 73%, and CT of 60% [18]. The benefit of it is a comprehensive scanning range and high sensitivity, and [ 68 Ga]Ga-PSMA-11 PET is widely used to achieve accurate staging and posttreatment efficacy evaluation. Therefore, more extensive use of [ 68 Ga]Ga-PSMA-11 PET shortens the time of prostate metastatic lesion detection and improves clinical decision-making. . This example showed that MRI-positive lesions could be either PSMA avid uptake (yellow arrows) or unapparent PSMA uptake (white arrows). And MRI-negative lesions could be unapparent PSMA uptake (green arrows) MRI brings valuable superiority over [ 68 Ga]Ga-PSMA-11 PET/CT because of the high soft-tissue contrast and provides the advantages of functional MRI techniques, as Hoeks et al. [19] demonstrated. Some attempts have been made to provide a multimodality approach. Park et al. [20] found that [ 68 Ga]Ga-PSMA-11 PET can be used to identify prostate cancer, while MR imaging provides detailed anatomic guidance. Therefore, [ 68 Ga]Ga-PSMA-11 PET/MRI imaging provides valuable diagnostic information and may inform the need for and extent of pelvic node dissection. Domachevsky et al. [21] proved that pelvic [ 68 Ga]Ga-PSMA-11 PET/MRI is superior to whole-body [ 68 Ga]Ga-PSMA-11 PET/CT in detecting extensions of localized disease. It is mainly due to the high soft-tissue resolution of MRI, by comparing between pelvic [ 68 Ga]Ga-PSMA-11 PET/MRI and wholebody [ 68 Ga]Ga-PSMA-11 PET/CT for the initial evaluation of prostate cancer. Abd-Alazeez et al. [22] studied the added value of apparent diffusion coefficient maps and dynamic contrast-enhanced images for the detection of radio recurrent prostate cancer and proved that MRI could evaluate recurrent or residual disease.[ 68 Ga] Ga-PSMA-11 PET/MRI has also been used to detected metastasis. Kranzbühler et al. [23] reported the usage of [ 68 Ga]Ga-PSMA-11 PET/MRI-positive peritoneal metastasis in the falciform ligament in recurrent prostate cancer. In conclusion, the development of MRI technology has dramatically improved the diagnostic accuracy of prostate cancer.
Nevertheless, PI-RADS is not perfect yet. Westphalen et al. critically evaluated the PI-RADS interpretation in 26 centers and reported that the positive predictive value of PI-RADS varied widely across centers [24]. The reason is that the efficacy of PI-RADS is generally related to the personal experience of physicians in practical application. Urologists and radiologists are still working on further optimizing the scoring system. Therefore, it is necessary to use a multimodality quantitative analysis to provide more information on diagnosis. For the interpretation of PET images, a five-point ordinal scale, Likertscale can be utilized with a score of 1, meaning PCa was highly unlikely and a score of 5, meaning PCa was highly likely. For interpretation of PET and PET/MRI images in PCa lesions, we may consider LBR lower than 1 as highly unlikely, LBR between 1 and 2 as unlikely, respectively, LBR between 2 and 3 as equivocal, LBR higher than 3 as likely, respectively, and LBR higher than 4 as highly likely.
Afshar  [20]. In terms of tumor severity and evaluation of extracapsular and seminal vesicular invasion, the results of [ 68 Ga] Ga-PSMA-11 PET were encouraging. These parameters  It leads to an increased risk of urinary incontinence and erectile dysfunction after prostatectomy [28]. These factors also have a profound impact on prognosis because both extracapsular extension and seminal vesicle invasion are associated with an increased risk of recurrence and lymph node and bone metastasis. Because of the high sensitivity of PSMA, it is easier to detect hidden residual and recurrent focals [5,22,23,25,26,[29][30][31][32] Ga]Ga-PSMA-11 PET/CT as a regular examination for primary staging before performing radical prostatectomy. In this condition, there are more studies on [ 68 Ga] Ga-PSMA-11 PET/CT. These researches can take radical prostatectomy specimens as the reference standard to perform lesion by lesion study. Chen et al. retrospectively enrolled patients who underwent both MRI and PET/CT before radical prostatectomy and analyzed the molecular imaging PSMA expression score and the pathologic results [35].
Second, [ 68 Ga]Ga-PSMA PET/MRI needs a more extended scanning protocol than [ 68 Ga]Ga-PSMA-11 PET/CT. During the scanning process, the MRI device emits a harsh noise. Although technicians adopt sound insulation solutions to patients, they still cannot eliminate the interference from noise to patients. The patient needs to keep the body stable and immobile during the entire scan.
Third, although [ 68 Ga]Ga-PSMA-11 has been one of the milestone discovery in the development of nuclear medicine in recent decades, which significantly improves the accuracy of prostate cancer diagnosis and assessment. Its specificity still could be further enhanced. Optimizing the targeting specificity of molecular probes is one of the most important methods. We hope that this problem will be solved in the future.
In the end, [ 68 Ga]Ga-PSMA-11 PET/MRI is widely used for staging reevaluation with recurrent prostate cancer after radical prostatectomy or to evaluate the conditions of patients who have already been treated by non-surgical therapies. Radical prostatectomy and prostate biopsies are invasive procedures with a high risk of focal hemorrhages and infection. PCa patients are mostly elderly men, with some underlying disease or age-related diseases that are not recommended to perform a pathological examination under this situation. To some extent, histopathological examination is not often feasible due to ethical and practical reasons.

Limitations
The limitation of our retrospective analysis is that [ 68 Ga] Ga-PSMA-11 PET/MRI is not compared with full histopathology examination because our cohort patients are elderly male and were not feasible to perform RP. Therefore, this analysis is a descriptive radiological imaging features study.

Conclusion
The ratio threshold value of SUVmax, LBR, has improved clinical and research applicability compared with the absolute value of SUVmax. A higher threshold value than the background's uptake is capable of dovetailing the imaging findings on MRI better. The specificity of [ 68 Ga] Ga-PSMA PET needs to be further improved by optimizing the targeting specificity of molecular probes.