Detection of clinically silent brain lesions in [18F]FDG PET/CT study in oncological patients: analysis of over 10,000 studies

The study aimed to show that including the brain region into the standard 2-deoxy-2-[18F]fluoro-d-glucose positron emission tomography/computed tomography ([18F]FDG PET/CT) study protocol may result in detecting clinically silent brain tumours. We retrospectively analyzed the group of 10,378 from the total of 12,011 consecutive patients who underwent the torso and brain [18F]FDG PET/CT scanning, considering an ability of the method to evaluate undetected before brain tumours in patients diagnosed and treated in our institution. While collecting the database, we followed the inclusion criteria: at least 1-year of follow-up, a full medical history collected in our institution, histopathologic examination or other studies available to confirm the type of observed lesion, and the most importantly—no brain lesions reported in the patients’ medical data. In this study, performing the torso and brain [18F]FDG PET/CT imaging helped to detect clinically silent primary and metastatic brain tumours in 129 patients, and the benign lesions in 24 studied cases, in whom no suspicious brain findings were reported prior to the examination. In conclusion, including the brain region into the standard [18F]FDG PET/CT protocol can be considered helpful in detecting clinically silent malignant and benign brain tumours.

In 19 from 24 of studied patients, single brain lesion was detected with the [18F]FDG PET/CT method, and in 5 cases the brain tumour was one of a multiple evaluated malignant lesions due to advanced stage of the primary disease (i.e. lymph nodes involvement, metastatic bone tumours; Fig. 1). Figure 1 shows the primary brain tumour discovered in CUP syndrome patient.
Brain metastases. We found the total of 137 metastatic brain tumours in 105 patients (up to 3 foci in one patient). We excluded from the final analysis 32 metastatic foci, using the total tumour's volume and the highest [18F]FDG uptake value as the inclusion criteria. The studied group consisted of 65 women, 40 men (mean age ± S. www.nature.com/scientificreports/ The metastatic brain tumour was the only one evaluated malignant lesion in 39 from 105 patients (37.1%), and one of a few detected tumours in the patient's body in 13 cases (12.4%). In 50.5% of the analyzed scans, we obtained the advanced oligometastatic disease with the significant lymph nodes involvement and the secondary bone tumours observed (Fig. 2). Figure 2 shows the metastatic brain lesion, obtained with the torso and brain [18F]FDG PET/CT study. Benign brain lesions. We found the benign brain lesions in 24 patients (15.7% of all detected lesions).
Glucose metabolism activity of the brain lesions. We measured glucose metabolism activity within the analyzed groups of brain lesions, using PET-dedicated parameter of the maximal standardized uptake value (SUVmax). According to the Shapiro-Wilk test's results, the SUVmax value distribution within the groups of the primary, metastatic and the benign lesions was Gaussian with P = 0.9, P = 0.7, P = 0.1, respectively (Table 1). Among the evaluated brain tumours, we found both [18F]FDG-avid and non-[18F]FDG-avid lesions. Table 1 shows the measurements. According to the Student t-test's for the independent groups of variables, the benign and malignant lesions' metabolic activity differed significantly with P < 0.001. According to the Kruskal-Wallis test's results, the SUVmax value difference between the primary and metastatic brain lesions was insignificant (P = 0.9). The SUVmax value dataset obtained within the benign lesions group differed significantly from the malignant tumours groups (P = 0.02).
Benign and malignant brain lesions differential diagnosis. The [18F]FDG PET/CT method is a widely used imaging technique for the oncological purposes, among which the especially important seems to be  FDG uptake within the grey matter results in observing the increased SUVmax value levels within the brain. Therefore, the possibility to detect a small brain lesion and to perform the differential diagnosis seems difficult using the SUVmax value of 2.5 criterion. In this study, we performed the Receiver Operating Characteristics (ROC) analysis to calculate the SUVmax value cut-off which may differentiate between the benign and the primary brain lesions in this database. The evaluation of cut-off points distinguishing all benign and all malignant tumours, and the primary and metastatic brain tumours was omitted due to significant differences in sample-sizes between those groups of malignant foci. When compared the SUVmax value levels between groups of the benign and primary brain lesions, we found the SUVmax value cut-off suggesting the malignant pathology was 6.1 with the sensitivity and the specificity of the method of 96%, 83%, respectively (Fig. 3).

Discussion
The study aimed to indicate the importance of the incidental brain lesions detection in the oncological patients, using the torso and brain [18F]FDG PET/CT acquisition. Although, the study is not considered a method of choice in the brain diagnosis, it remains one of the most used and accessible oncological imaging techniques. Discovering the primary or secondary brain tumours seems to be crucial due to high risk of the lethal effect of a developing malignant brain lesion as well as due to several neurological and cognitive ailments observed in patients suffering from the brain tumours. Authors 12,14-16 discussed the applications and limitations of the torso and brain [18F]FDG PET/CT study, indicating a low sensitivity and the specificity of the method in detecting small tumours located within the brain. In some of the nuclear medicine departments performing the PET/CT scanning using the [18F]FDG might be the only way to support their oncological patients due to the limited accessibility to the more suitable for the brain tumours diagnosis radiopharmaceuticals, i.e. tumours' hypoxia or increased cells' proliferation indicators 17,18 .
In this study, we detected clinically silent brain tumours in 153 from examined 10,378 oncological patients (1.5% of examined population), among which nearly 90% occurred to be malignant, and 24 were benign. The available follow-up data of the remaining 10,225 studied subjects did not mention any other cases of the brain tumour being diagnosed. Detecting benign brain pathologies might not effect with the necessity of the surgical resection of the obtained foci, and according to the available medical records, none of the evaluated 24 benign lesions underwent surgical resection. However, the discovery demands follow-up studies to avoid possible sideeffects connected with the tumour's growth, including the neurological, locomotory, and cognitive disorders. Moreover, in some cases, the developing benign brain lesion may also be lethal due to, i.e., progressing destruction of the brain tissue caused by the increasing over time lesion's volume. In this database, in every condition, the regular follow-up studies were recommended. www.nature.com/scientificreports/ Most often, malignant brain foci undergo treatment as deadly. The patients in whom the malignant brain foci was detected are monitored more carefully, especially if the primary or metastatic brain focus is the only one or one of a few detected symptoms of the developing oncological disease. According to the patients' medical records, all of 129 malignant brain tumours patients received treatment (surgical resection, chemotherapy, surgery combined with radiotherapy; depending on the staging of the disease, tumour's location, patients' general health condition).
We have included into the final analysis the patients' in whom the brain lesion were not reported previously in any other available studies. In some of cases, the histopathologic data were not available due to high postsurgical risk of mortality. In this group, we used other imaging methods' results and the follow-up studies. To ensure a full access to the patients' medical records, we included in the study patients diagnosed and treated only in our institution (patients' transferred to other hospitals to continue the therapy were excluded due to insufficient follow-up data).
According to authors 12,16 , the torso and brain [18F]FDG PET/CT study can be considered useful in the benign and malignant brain tumours differential diagnosis. In this study, we used the ROC analysis to evaluate the SUVmax value cut-off, distinguishing benign and primary lesions. The SUVmax value cut-off differentiating between malignant and benign brain lesions was 6.1 with a considerably high sensitivity and specificity of the method (96%, 83%, respectively).
The oncological patients' database is heterogenous due to institution characteristics (wide range of examined diagnoses) which may be considered the limitation of the method. Moreover, the differences in the sample-sizes between the groups of the benign, primary and metastatic brain tumours limited the possibility to establish the SUVmax value cut-off, distinguishing the benign and metastatic, and the primary and metastatic tumours. Finally, the relevance of the obtained results may be considered subjective and depending on the clinician's individual assessment whether to use the torso and brain [18F]FDG PET/CT in the daily practice.
In this study, performing the torso and brain [18F]FDG PET/CT examination resulted in the incidental detection of the clinically silent brain tumours, and gave 129 patients a chance to successfully undergo the oncological therapy, which would not be possible if the 1 min of the brain scanning would be excluded from the [18F]FDG PET/CT acquisition protocol. Thus, inclusion of the brain region into the standard acquisition PET/CT protocol might be worthy consideration.

Bioethics. The study is based on the retrospective anonymized analysis approved by the Local Bioethical
Committee (Poznan University of Medical Sciences, prof. Pawel Checinski, date of approval: 30.01.2020). This study includes the original studies performed upon patients' informed consent in writing due to the standard institution protocol. The study is based on the unsponsored, single-institutional studies, using the database collected in 2011 to 2019. All data have been anonymized and the examined patients cannot be identified. All steps of the examinations have been performed in accordance with the Bioethical Committee guidelines and the Declaration of Helsinki.
Statistical analysis. In this study, we performed the comparative analysis of the PET-dedicated parameter of SUVmax, obtained withing the groups of the benign brain lesions, primary brain tumours, and the metastatic brain foci. Prior the final analysis, we used the Shapiro-Wilk normality test and divided datasets into independent groups of lesions. We followed the statistical significance level of α = 0.05 (confidence interval at the level of 95%, CI 95 ), and summarized the statistical tests' results, considering P value. When analyzed the results, we used a null and alternative hypotheses (H 0 , H a , respectively) assumptions: H 0 suggested that the true variables' distribution was normal or the evaluated differences between the calculations were statistically insignificant (P > 0.05), H a -true distribution significantly differed from Gaussian or the observed differences were significant (P < 0.05) 19 . In this study, we used the following statistical tests: the Kruskal-Wallis' , and the Student t-tests for independent variables. Furthermore, we performed the ROC analysis to establish the predictive SUVmax value cut-off distinguishing the benign lesions and the primary brain tumours glucose metabolism activity in this database (including the YI, and AUC parameters).
We used the STATISTICA, StatSoft software, version 13.3 (TIBCO Software, Palo Alto, California, USA, available upon individual license).
Database collection. In this study, we have retrospectively analyzed 12,011 consecutive patients examined with the [18F]FDG PET/CT technique in our institution. In this group, we collected the database consisting of 10,378 patients, evaluated using the torso and brain [18F]FDG PET/CT 12 studies. We included into analysis the total of 10,378 studies, following the criteria shown in the Table 2.
Study performance. The torso and brain [18F]FDG PET/CT study has been performed at 60 min (min).
The acquisition protocol included the area of skull-apex to mid-thigh (patients laid supine with arms above the head) 12 . PET imaging preceded Body Low-Dose CT using the following parameters: 150-245 miliamperseconds (mAs), 120-140 kilovoltage peak (kVp), Pitch of 0.8. PET section scanning time was 90 s (s). The scanning duration did not exceed 35 min (min), including 1 min of the brain region imaging. To evaluate lesions within the Table 2. 10,378 studies collection conditions: inclusion and exclusion criteria ( source: original data). a Range: 1-4 years of follow-up.

Inclusion criteria Exclusion criteria
Torso and brain [18F]FDG PET/CT study protocol performed in our institution to ensure a full access to the images, ranging from the skull-apex to mid-thigh Other than torso and brain imaging [18F]FDG PET/CT (dated protocols omitting the brain region, a low quality of the brain scanning, incomplete brain imaging, movement or other artefacts observed) The patients' medical records available (including the therapeutic management data) Incomplete medical records (including: studies performed in other hospitals, treatment continued in external institution) At least 1-year a of the patients' follow-up: the brain lesion's type confirmed using the histopathological examination (depending on the possibility to perform surgery) or other studies (ceCT, MRI, repeated torso and brain [18F]FDG PET/CT imaging) Insufficient follow-up, no other than torso and brain [18F]FDG PET/CT study available to confirm the lesion's type or to analyze the further patients' management No brain lesion reported in the patients' medical records Brain tumour previously detected or suspected prior to the [18F] FDG PET/CT study Clinically silent brain lesion observed Symptoms of the developing brain lesion reported (i.e. neurological or cognitive ailments)