Large language models show great promise for clinical decision support but generally lack medical imaging integration. Therefore, vision-language models have been proposed, but there is little evidence on their performance, particularly compared with conventional machine learning. Prostate-specific membrane antigen (PSMA)-targeted radiopharmaceutical therapy is a well-established treatment for metastatic castration-resistant prostate cancer (mCRPC), but it is still not known which patients will be good long-term responders. Therefore, this study proposes a physician-inspired visual question answering (VQA) framework that emulates the structured integration of PSMA-directed imaging and clinical data used in routine clinical assessment for overall survival prognostication. Methods: As a proof of concept, we conducted a retrospective, multicenter study across 2 academic institutions that included 110 patients with mCRPC who were treated with either [177Lu]Lu-PSMA I&T or [177Lu]Lu-PSMA-617. Our VQA approach incorporates a 3-dimensional image encoder, the domain-specific language model RadBERT, and a cross-modal transformer to process baseline and interim (after 2 cycles) PET and SPECT images, clinical data, and clinical query. The model was trained to predict 24-mo survival, and performance was compared with state-of-the-art approaches. External validation was performed with the [177Lu]Lu-PSMA-617 cohort. Results: The VQA model outperformed all state-of-the-art approaches while showing accuracy comparable to the query-free attention multimodal fusion model. Specifically, it achieved 81.8% accuracy (95% CI, 0.52-0.95) when integrating SPECT or PET images with prostate-specific antigen measurements. External validation of the VQA showed 71.9% accuracy (95% CI, 0.53-0.86) after few-shot fine-tuning, with further improvement over all state-of-the-art approaches in survival stratification. The VQA approach showed the closest calibration with the ground-truth survival distribution, with no significant differences observed (log-rank test, P = 0.76). Receiver operating characteristic analysis showed the highest area under the curve (0.75) among the evaluated models. Together, these findings suggest that the VQA approach demonstrated adaptability across different PSMA ligands while maintaining slightly improved yet comparable overall survival prognostic performance. Conclusion: By adopting a physician-inspired assessment workflow, the VQA framework shows advantages in integrating imaging and laboratory data for clinical decision support compared with conventional machine learning. The preliminary results indicate that VQA-based approaches warrant future evaluation for survival prognostication in patients with mCRPC who are receiving PSMA radiopharmaceutical therapy.

[18F]F-AraG is a clinical-stage PET tracer in immuno-oncology that accumulates in activated T cells and tissues with high mitochondrial biogenesis. Characterizing its distribution in individuals without malignancy is essential for interpreting immunometabolic states relevant to cancer biology. Methods: Twenty adults without active malignancy underwent a single [18F]F-AraG PET/CT scan, with tracer uptake assessed in relation to clinical history, medications, and demographic and physiologic factors. Results: Bilateral brown adipose tissue activation was identified in 3 participants receiving potentially brown adipose tissue-activating medications, in the absence of cold exposure or fasting. One showed elevated vertebral bone marrow uptake, previously linked to poor survival in immune checkpoint inhibitor-treated lung cancer. One of 2 participants with breast implants showed asymmetric periimplant uptake. Posterior pulmonary uptake occurred in a subset of participants, often with prior respiratory conditions and thoracic lymph node uptake. Conclusion: Whole-body [18F]F-AraG PET captures diverse immunometabolic activity in individuals without malignancy, reflecting systemic processes that may influence cancer and treatment outcomes.

Prior oncologic PET/MRI repeatability studies have been limited by single-center designs, small cohorts, or select patient populations. Our primary aim was to estimate the test-retest repeatability of uptake metrics on [18F]FDG PET/MRI studies across a variety of cancer types. A secondary aim was to assess the impact of intersession μ-map variability on repeatability. Methods: In this prospective study involving 2 institutions with different PET/MRI scanners, subjects completing standard-of-care [18F]FDG PET/CT or MRI examinations were recruited to undergo 2 research [18F]FDG PET/MRIs (i.e., test and retest) within a 1-9 d period per Quantitative Imaging Biomarker Alliance specifications. Enrollment required at least 1 measurable lesion (long axis ≥ 2 cm, known or expected SUVmax ≥ 4.0) and no active oncologic therapy. SUVmax and SUVpeak were extracted from measurable lesions (1-5 per subject). At site 1, attenuation-coefficient maps (i.e., μ-maps) were inspected for substantial differences across sessions. Additionally, at site 1, the retest μ-map was coregistered to the test μ-map, and the PET attenuation-correction process was repeated for the test session based on the retest μ-map. Intersession repeatability was assessed via Bland-Altman plots, and within-subject coefficients of variation (wCVs) were calculated. Results: At site 1, 10 subjects with 11 measurable lesions were analyzed, and the wCVs for SUVmax and SUVpeak were 12.3% and 9.6%, respectively. At site 2, 11 subjects with 30 measurable lesions were analyzed, and the wCVs for SUVmax and SUVpeak were 9.5% and 8.8%, respectively. At site 1, repeating attenuation correction of the test PET data with the retest μ-map slightly improved repeatability, reducing the wCV to 10.1% for SUVmax and to 9.1% for SUVpeak For 8 of 11 lesions at site 1, the μ-map substitution altered the test-retest change by less than 5%, with no appreciable intersession μ-map differences or artifacts. Conclusion: The test-retest repeatability of SUVmax among hypermetabolic lesions on [18F]FDG PET/MRI, using data from 2 scanner models, was similar to the Quantitative Imaging Biomarker Alliance's wCV benchmark of 10%-12% for [18F]FDG PET/CT. Intersession μ-map variations had minimal negative impact on repeatability.

B7 homolog 3 (B7H3) is widely expressed in solid tumors, but its role in the diagnosis of breast cancer is unclear. We aimed to evaluate the diagnostic performance of [68Ga]Ga-B7H3-BCH PET/CT in breast cancer patients. Methods: Thirty patients with diverse subtypes of breast cancer underwent both [68Ga]Ga-B7H3-BCH and [18F]FDG PET/CT. Immunohistochemical staining for B7H3 was performed in 25 patients. Results: B7H3 positivity was detected in 84% of samples by immunohistochemistry. [68Ga]Ga-B7H3-BCH demonstrated a higher lesion detection rate compared with [18F]FDG (96.8% vs. 92.4%) among 157 lesions. Median SUVmax was 9.5 in strongly positive lesions; bone metastases had the highest uptake (median SUVmax, 10.9). Imaging resulted in changing the stage or treatment in 8 patients. Hormone receptor-positive/human epidermal growth factor receptor 2-negative lesions showed the highest B7H3 positivity (87.5%) and uptake (median SUVmax, 10.5). Conclusion: [68Ga]Ga-B7H3-BCH PET/CT offers high sensitivity for staging advanced breast cancer and noninvasive B7H3 screening. It may influence clinical decisions and appears to reveal subtype-specific uptake, with a tendency to favor hormone receptor-positive/human epidermal growth factor receptor 2-negative disease.

Somatostatin receptor type 2 is highly expressed in meningiomas and can be quantified using [68Ga]Ga-DOTATATE PET. The Krenning score, originally developed for [111In]In-octreotide scintigraphy, remains the standard visual grading system for somatostatin receptor expression and is widely used to guide eligibility for peptide receptor radionuclide therapy. However, visual grading may not optimally translate to intracranial disease. This study evaluated the relationship between brain-specific quantitative PET parameters-SUVmax and SUV ratio normalized to the superior sagittal sinus (SUVRsss)-and the Krenning score in meningioma. Methods: This single-institution retrospective study included 44 patients (70 meningiomas) who underwent clinical vertex-to-thigh [68Ga]Ga-DOTATATE PET/CT. Quantitative parameters (SUVmax, SUVRsss, lesion size) were extracted from coregistered PET and MR images. Spearman rank correlation analysis was performed to assess associations between imaging variables and Krenning score. Differences in quantitative PET parameters across Krenning categories were evaluated using Kruskal-Wallis test and Holm-Bonferroni-corrected Mann-Whitney U test. Receiver-operating-characteristic curve analysis using the Youden J index identified optimal cutoffs for clinically relevant thresholds (Krenning score of <3 vs. ≥3 and 3 vs. 4). Results: SUVRsss and SUVmax were strongly correlated with the Krenning score (SUVRsss: Spearman ρ = 0.70; 95% CI, 0.55-0.82; SUVmax: Spearman ρ = 0.77; 95% CI, 0.64-0.85; P < 0.0001). PET parameters differed significantly across Krenning categories (all P < 0.001). SUVRsss optimal thresholds were 9.00 (area under the curve [AUC], 0.86; 95% CI, 0.74-0.96) for Krenning scores of 2 versus 3 and 29.84 (AUC, 0.98; 95% CI, 0.93-1.00) for Krenning scores of 3 versus 4. SUVmax optimal thresholds were 7.20 (AUC, 0.92; 95% CI, 0.84-0.99) and 30.70 (AUC, 0.98; 95% CI, 0.95-1.00), respectively. Conclusion: SUVRsss and SUVmax are complementary quantitative correlates of Krenning scores 2-4 in [68Ga]Ga-DOTATATE PET-visible, MRI-measurable meningiomas. Brain-only PET evaluation may provide an alternative to liver- and spleen-based visual grading and reduce the need for whole-body imaging in selected settings. Receiver-operating-characteristic curve-derived SUV thresholds correspond to Krenning scores used for eligibility for treatment with peptide receptor radionuclide therapy, supporting further evaluation of quantitative PET parameters for theranostic decision-making for patients with meningioma.

Evidence supporting the efficacy and safety of [177Lu]Lu-DOTATATE in refractory meningiomas derives mainly from small, single-center studies. The French national multicenter MELUTE study aimed to consolidate these data using a standardized regimen within a compassionate-use program approved by a national multidisciplinary tumor board. Methods: In this retrospective multicenter cohort study, patients with refractory meningiomas treated with [177Lu]Lu-DOTATATE in 18 French centers between July 2022 and December 2024 were included. Treatment consisted of 7.4 GBq every 8 wk for up to 4 cycles. Clinical data (demographics, histologic World Health Organization grade [WHO], prior treatments, number of cycles, toxicity, progression, survival status as of October 1, 2025) were collected. Baseline [68Ga]Ga-DOTATOC PET and brain MRI scans obtained before and after treatment were centrally reviewed. The primary endpoint was 6-mo progression-free survival (PFS) assessed using volumetric Response Assessment in Neuro-Oncology Group criteria. Secondary endpoints were PFS, overall survival (OS), and safety. Results: Sixty-nine patients were included (mean age, 65.1 ± 13.2 y; 52% female): 13 WHO grade 1, 42 WHO grade 2, 8 WHO grade 3, and 6 without histology. Most had prior surgery (86%) or radiotherapy (93%); 32% received prior systemic therapy. Twelve patients (17%) did not complete 4 cycles; only one discontinuation was toxicity-related. Six-month PFS was 70% (95% CI, 58.7-80.4%). Median PFS was 14.6 mo (95% CI, 13.3-18.5 mo), and median OS was 27.6 mo (95% CI, 24.2 mo to not reached). Grade 3 adverse events, mainly lymphopenia, occurred in 16%. Meningioma WHO grade independently predicted PFS, whereas higher Karnofsky performance status and corticosteroid use were protective; worse OS was associated with higher baseline tumor growth rate and prior systemic therapy. Conclusion: MELUTE demonstrates promising efficacy and favorable safety of [177Lu]Lu-DOTATATE in refractory meningiomas with limited therapeutic options.

PET is the most powerful imaging modality for studying neurotransmitters, receptors, and neuromodulation in humans, complementing the anatomic and hemodynamic information provided by MRI. However, PET temporal resolution remains constrained by sensitivity, limiting studies of fast, dynamic neurochemical processes. To address this gap, we are developing the Human Dynamic NeuroChemical Connectome, an imaging system integrating NeuroSphere PET, a novel ultra-high-sensitivity dedicated brain PET camera, with the Siemens Healthineers Magnetom Terra.X, a clinical 7-T MRI scanner. This article describes the design, engineering, and simulation-based performance estimation of NeuroSphere PET. Sensitivity maximization drove 2 core design choices: a partial spheric detector arrangement with a 32-cm inner diameter and 70% solid-angle coverage and the use of 26-mm lutetium oxyorthosilicate crystals for high stopping power (∼90% at 511 keV), coupled to silicon photomultipliers in a light-sharing, depth-of-interaction-capable configuration. The system comprises 872 detector blocks organized across 19 rings and 218 modules. At the detector level, measured performance includes energy resolution of approximately 11% full width at half maximum (FWHM), depth-of-interaction resolution of approximately 8 mm FWHM, and coincidence timing resolution of approximately 500 ps FWHM. System-level performance was evaluated using the ultrafast Monte Carlo PET simulator, accounting for realistic detector response, positron range, photon noncollinearity, and interdetector scatter. Sensitivity was estimated from a 30-cm uniform sphere, predicting a peak sensitivity of about 38% and a whole-brain average exceeding 25%. Spatial resolution and image quality were assessed through Derenzo phantom simulations evaluated at both the center of the field of view and 8 cm off-center, indicating that 1.6-mm structures can be resolved near the center of the field of view and that resolution of less than 2 mm can be maintained at the cortex. The system incorporates a 3-dimensional-printed MRI-compatible gantry, a dedicated 16-channel parallel transmit-receive radiofrequency head coil, and a custom radiofrequency shielding enclosure for simultaneous PET/MRI at 7 T. By targeting sensitivity an order of magnitude higher than that of commercial PET systems, NeuroSphere PET is expected to enable simultaneous neurochemical and hemodynamic studies with substantially improved spatiotemporal resolution, providing new insights into brain function and connectivity.

People with recent-onset diabetes often have preserved C-peptide production, which tends to decline during the first 7 y after diagnosis and stabilizes thereafter. Whether this holds true for β-cell mass is unknown. In this pooled post hoc analysis, we aimed to investigate total β-cell mass in individuals with short- and long-duration type 1 diabetes. Methods: Data of 16 individuals with short- (<7 y) and 10 with long-duration type 1 diabetes (≥7 y) who underwent [[68]Ga]Ga-NODAGA-exendin-4 PET/CT and mixed-meal tolerance test were used to determine total β-cell mass and function, respectively. Intrapancreatic fat deposition and pancreatic volume were derived from CT. Results: Total β-cell mass was higher in individuals with short-duration type 1 diabetes compared with those with long-duration type 1 diabetes (4.4 ± 1.8 vs. 2.8 ± 1.3 kBq/MBq, P = 0.012). Larger pancreatic volume, lower intrapancreatic fat deposition, and higher β-cell function were observed in individuals with short-duration type 1 diabetes compared with the long-duration group. Total β-cell mass positively correlated with β-cell function (r = 0.46, P = 0.019) and pancreatic volume (r = 0.91, P < 0.0001) and inversely correlated with body mass index (r = -0.61, P = 0.0009). Conclusion: This study offers insights in the potential of imaging markers together with function tests to monitor disease progression and may offer valuable information for the timing to restore β-cell function or to slow down β-cell destruction.

The clinical standard for neuroblastoma detection is based on targeting the norepinephrine transporter with [[123]I]MIBG, which has several challenges. Imaging the somatostatin receptor (SSTR) is a strong alternative. In contrast to standard DOTATATE, SSTR antagonist imaging with [[68]Ga]Ga-satoreotide trizoxetan ([[68]Ga]-SSO120) can potentially improve lesion detection as it targets both active and inactive SSTRs. We present the biodistribution and radiation exposure results of [[68]Ga]Ga-SSO120 PET/CT imaging in pediatric patients with neuroblastoma. Methods: Patients age 3-7 y with stage 4 neuroblastoma received 2 MBq/kg of [[68]Ga]Ga-SSO120, followed by a 60-min dynamic scan and a static PET/CT scan. SUV normalized to lean body mass (SULmax, SULmean, SULpeak) and time-activity curves were measured for suspected lesions and various tissue compartments. Tumor-to-liver and tumor-to-blood pool ratios were calculated. Radiation doses to various organs and the total-body effective dose were computed. Results: Five patients underwent dynamic [[68]Ga]Ga-SSO120 PET/CT scanning, with a total of 17 neuroblastoma lesions. The median administered activity was 2.5 MBq/kg (range, 2.1-2.8 MBq/kg). At 60 min postinjection, the median tumor SULmax and SULpeak were 4.28 (range, 1.89-18.07) and 2.71 (range, 1.23-14.77), respectively, and the median tumor-to-liver ratio was 4.05. The highest exposure was in the bladder (22.3 ± 4.3 μSv/MBq). The effective dose was 0.088 ± 0.014 mSv/MBq. Conclusion: [[68]Ga]Ga-SSO120 PET/CT was safe in pediatric patients with neuroblastoma, with a low effective radiation dose, and detected neuroblastoma lesions with a good target-to-noise ratio. Image acquisition at 30 and 60 min postinjection seemed equivalent.