This prospective pilot study was approved by the institutional review board of Stanford University School of Medicine where it was wholly performed, in accordance with the 1964 Declaration of Helsinki, and with all patients providing informed consent prior to enrollment. Subsequent analysis of the collected anonymized data was performed at The University of Texas MD Anderson Cancer Center through a data transfer agreement, and for which separate institutional review board approval was granted. This study was performed in the context of a phase I clinical trial (NCT03539731) and enrolled subjects ≥ 18 years old. Note that the radiation dosimetry and human biodistribution results from this trial have previously been reported [10]. Two primary groups were included in this study, with the first comprised of healthy volunteers, who underwent a single [18F]DASA-23 PET/MRI scan [10]. The second group was comprised of patients with newly-diagnosed or recurrent grade 4 Glioblastoma IDH wild-type, and grade 3 diffuse astrocytoma IDH mutant based upon the 2021 WHO Classification of Tumors of the CNS, and were required to have ≥ 1 cm3 of enhancement. The patients underwent [18F]DASA-23 PET/MRI before (baseline) and 1–10 weeks after initiation of therapy (follow-up), but with only a subset of the high-grade glioma patients undergoing the follow-up scan. Patients were required to have adequate organ function prior to radiotracer injection as demonstrated by absolute neutrophil count ≥ 1.5 × 109 cells/L, hemoglobin ≥ 9 g/dL, platelet count ≥ 100 × 109/L, bilirubin ≤ 1.5 × upper limit of normal (ULN), alanine aminotransferase and aspartate aminotransferase each ≤ 2.5 × ULN, and alkaline phosphatase ≤ 3 × ULN. Furthermore, in reproductive age females a negative serum pregnancy test was required.

PET and MRI data were simultaneously acquired with Signa PET/MRI scanner (GE Healthcare, Waukesha, WI, USA). PET data was acquired with 3D technique and in list mode with a field of view of 300 × 300 mm. PET data was reconstructed with GE Healthcare’s proprietary algorithm VUE Point FX (VPFXS) which utilized a time of flight ordered subsets expectation maximation (OSEM) technique with 28 subsets, 2 iterations, matrix of 256 × 256, in-plane gaussian filter with full-width-half-maximum (FWHM) of 4 mm spatial cut-off frequency, and a Z-axis filter with a [1 4 1] matrix kernel. The mean injected radiotracer dosage among subjects was 5.85 mCi (range 4.49–8.66 mCi), or 216.45 MBq (range 166.13–320.42 MBq). Dynamic PET acquisition was started immediately upon radiotracer injection for a duration of 60 min. Attenuation correction of the PET data was performed by using the MR zero echo time (ZTE) method [12]. Prior analysis of this dynamic data showed sufficient differential uptake between tumor and background parenchyma starting at approximately 20 min following radiotracer injection [6]. Based on this information, counts were summated for frames spanning 20–30 min to yield a single static acquisition through the brain. A standard of care MRI brain tumor protocol was concurrently acquired, and included T1 pre-contrast, T1 post-contrast, T2, T2 FLAIR, DWI, arterial spin labeling (ASL), and dynamic susceptibility contrast (DSC) imaging. However, not all patients underwent all of these sequences due to a desire to decrease scan-times in patients who had recent brain MRIs as part of their routine clinical care. Patients were asked to void their bladder before and after completion of the PET/MRI scan to reduce radiation exposure.

The healthy control cohort underwent [18F]DASA-23 PET/MR imaging at a single time point following enrollment. Patients with high-grade gliomas were imaged following recent diagnosis or recurrence as verified by resection or biopsy, and prior to initiation of treatment modalities including radiotherapy, temozolomide, lomustine, and/or bevacizumab. To evaluate the diagnostic performance of [18F]DASA-23 in the identification of high-grade gliomas, three board certified nuclear medicine physicians who were blinded to patient data were presented with the PET data, in isolation of simultaneously acquired MR imaging, of healthy controls and high-grade glioma patients. These readers were instructed to identify the presence or absence of focal uptake within the intracranial compartment that was above the background parenchyma white matter (which is notably higher than grey matter), serving as a marker of viable high-grade glioma. Disagreement among the readers was resolved by selecting the interpretation (i.e. presence or absence of focal uptake) provided by the majority of readers. From this data, the sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) of [18F]DASA-23 in identifying viable high-grade glioma were calculated, with confidence intervals derived from the Exact Binomial method [13].

The PET avid lesions in high-grade glioma patients on their baseline scan and follow-up scan (in the subset of patients who had follow-up scans) were segmented by a dual board-certified neuroradiology and nuclear medicine physician using the proprietary semi-automated PETEdge+ tool of MIM Software version 7.2 (MIM Software Inc., Cleveland, OH), and their maximum standardized uptake value (SUVmax), mean standardized uptake value (SUVmean), and metabolic tumor volume (MTV) were recorded. In brief, for this segmentation method the user selects a single point centrally within an avid lesion above background, and the software uses an SUV gradient-based method (i.e. region of interest (ROI) margins based upon degree of uptake drop-off) to automatically delineate the lesion’s boundaries [14]. The SUVmax and SUVmean of the normal contralateral white matter at the level of the centrum semiovale was also recorded to allow computation of TBR SUVmax (TBRmax) and TBR SUVmean (TBRmean), respectively. The baseline and follow-up SUVmax, TBRmax, SUVmean, TBRmean, MTV, as well as change in these values (i.e. follow-up values subtracted from baseline values), then underwent Pearson correlation analysis with each patient’s subsequent PFS from date of baseline scan as documented in the patient’s medical record. Statistical significance was defined as a 2-tailed p-value < 0.05. All statistical analyses were performed with SPSS Statistics version 24 (IBM, Armonk, New York). Additionally, an exploratory analysis based upon the results of Pearson correlation was planned, with those uptake parameters demonstrating statistically significant correlations with PFS undergoing subsequent stratification into two groups, (1) those with uptake values below median (2) those with uptake at or above median. These separate group’s PFS then underwent Kaplan-Meier survival curve analysis to evaluate for significant differences in their PFS distributions using the Log Rank test, with statistical significance set at p-value < 0.05.

In order to assess the relationship between perfusion imaging metrics and [18F]DASA-23 uptake, those patients in group 2 who received MR-based DSC and ASL imaging simultaneously with their PET acquisition had their perfusion parameters correlated with the [18F]DASA-23 TBRmax. The previously created segmented [18F]DASA-23 PET uptake ROI for each patient were coregistered with the post-processed DSC and ASL imaging series. For the ASL series, the [18F]DASA-23 segmented ROIs were used to measure the relative cerebral blood flow max (rCBFmax) within the region of [18F]DASA-23 uptake, which was divided by rCBFmax in a contralateral cerebral hemisphere ROI encompassing cortex and subcortical white matter, yielding normalized rCBFmax (nCBFmax) [15]. As the DSC imaging was confounded by prominent signal intravascularly at sites intimately associated with [18F]DASA-23 uptake, the coregistered segmented ROIs were first manually edited by a dual-board certified neuroradiologist and nuclear medicine physician to remove these sites of confounding uptake. Next, the relative cerebral blood volume mean (rCBVmean) within the edited segmented ROI was measured, followed by measuring background rCBVmean within a 1.5 cm ROI within the contralateral centrum semiovale white matter. The former value was divided by the background rCBVmean to yield a normalized rCBV (nCBV). Pearson correlation was then performed for both [18F]DASA-23 TBRmax and nCBFmax, followed by [18F]DASA-23 TBRmax and nCBV. Statistical significance was again defined as a 2-tailed p-value < 0.05.

The ASL was acquired with the single post-labeling delay pseudocontinuous technique, which is the method recommended for clinical use by the International Society for Magnetic Resonance in Medicine Perfusion Study Group and the European ASL in Dementia Consortium [16]. ASL acquisition parameters included post-labeling delay = 2025 ms, repetition time (TR) = 4854 ms, echo time (TE) = 10.7 ms, flip angle = 111°, acquisition matrix = 512 × 512, field of view (FOV) = 240 × 240 mm, and slice thickness = 4 mm. DSC was performed with a single-echo gradient EPI technique with half-dose preload bolus followed by half-dose dynamic bolus via power injector of gadobenate dimeglumine at 0.09 mmol/kg, with the latter having a 15 s delay prior to acquisition. Additional DSC acquisition parameters included TR = 1800 ms, TE = 40 ms, flip angle = 60°, acquisition matrix 96 × 96 mm, FOV = 240 × 240 mm, and slice thickness = 3.6 mm. DSC post-processing was performed with the IB Neuro platform version 21.05.1810 (Imaging Biometrics, Elm Grove, Wisconsin) and incorporated vascular leakage correction.

Finally, as an exploratory analysis, we sought to investigate how [18F]DASA-23 uptake overlapped with sites of both T1 post-contrast enhancement and non-enhancing T2-FLAIR hyperintense signal about the treatment cavity on their baseline imaging prior to therapy. Patients with high-grade gliomas who underwent T1 pre-contrast, T1 post-contrast, and T2-FLAIR sequences as part of their baseline [18F]DASA-23 PET/MR scans were evaluated as part of this analysis. The [18F]DASA-23 uptake ROIs for each of these scans were co-registered with the T1 post-contrast sequence. The T1 pre-contrast imaging was preliminarily reviewed to exclude sites of intrinsic T1 hyperintensity which could mimic enhancement on the T1 post-contrast exams. Regions of enhancement on the T1 post-contrast sequence, regardless of the presence of superimposed T2 FLAIR hyperintensity, within the [18F]DASA-23 ROIs were manually segmented and their volumes were recorded along with the overall volume of the [18F]DASA-23 ROI. Next, both the [18F]DASA-23 ROI and enhancement segmented ROI were coregistered to the T2-FLAIR sequence. This allowed for manual segmentation of T2-FLAIR hyperintensity within the [18F]DASA-23 ROI, but excluding already segmented regions of enhancement, so as to accurately record the volume of non-enhancing T2-FLAIR hyperintensity. The enhancement and T2-FLAIR hyperintensity volumes then underwent paired sample t-test to determine if these volumes differed significantly, with statistical significance defined as a 2-tailed p-value < 0.05.