Diagnostic Value of 18 F-FACBC PET/MRI in Brain Metastases.

PURPOSE: The study aims to evaluate whether combined 18 F-FACBC PET/MRI could provide additional diagnostic information compared with MRI alone in brain metastases. PATIENTS AND METHODS: Eighteen patients with newly diagnosed or suspected recurrence of brain metastases received dynamic 18 F-FACBC PET/MRI. Lesion detection was evaluated on PET and MRI scans in 2 groups depending on prior stereotactic radiosurgery (SRS group) or not (no-SRS group). SUVs, time-activity curves, and volumetric analyses of the lesions were performed. RESULTS: In the no-SRS group, 29/29 brain lesions were defined as "MRI positive." With PET, 19/29 lesions were detected and had high tumor-to-background ratios (TBRs) (D max MR , ≥7 mm; SUV max , 1.2-8.4; TBR, 3.9-25.9), whereas 10/29 lesions were undetected (D max MR , ≤8 mm; SUV max , 0.3-1.2; TBR, 1.0-2.7). In the SRS group, 4/6 lesions were defined as "MRI positive," whereas 2/6 lesions were defined as "MRI negative" indicative of radiation necrosis. All 6 lesions were detected with PET (D max MR , ≥15 mm; SUV max , 1.4-4.2; TBR, 3.6-12.6). PET volumes correlated and were comparable in size with contrast-enhanced MRI volumes but were only partially congruent (mean DSC, 0.66). All time-activity curves had an early peak, followed by a plateau or a decreasing slope. CONCLUSIONS: 18 F-FACBC PET demonstrated uptake in brain metastases from cancer of different origins (lung, gastrointestinal tract, breast, thyroid, and malignant melanoma). However, 18 F-FACBC PET/MRI did not improve detection of brain metastases compared with MRI but might detect tumor tissue beyond contrast enhancement on MRI. 18 F-FACBC PET should be further evaluated in recurrent brain metastases.

AI-driven attenuation correction for brain PET/MRI: Clinical evaluation of a dementia cohort and importance of the training group size.

INTRODUCTION: Robust and reliable attenuation correction (AC) is a prerequisite for accurate quantification of activity concentration. In combined PET/MRI, AC is challenged by the lack of bone signal in the MRI from which the AC maps has to be derived. Deep learning-based image-to-image translation networks present itself as an optimal solution for MRI-derived AC (MR-AC). High robustness and generalizability of these networks are expected to be achieved through large training cohorts. In this study, we implemented an MR-AC method based on deep learning, and investigated how training cohort size, transfer learning, and MR input affected robustness, and subsequently evaluated the method in a clinical setup, with the overall aim to explore if this method could be implemented in clinical routine for PET/MRI examinations. METHODS: A total cohort of 1037 adult subjects from the Siemens Biograph mMR with two different software versions (VB20P and VE11P) was used. The software upgrade included updates to all MRI sequences. The impact of training group size was investigated by training a convolutional neural network (CNN) on an increasing training group size from 10 to 403. The ability to adapt to changes in the input images between software versions were evaluated using transfer learning from a large cohort to a smaller cohort, by varying training group size from 5 to 91 subjects. The impact of MRI sequence was evaluated by training three networks based on the Dixon VIBE sequence (DeepDixon), T1-weighted MPRAGE (DeepT1), and ultra-short echo time (UTE) sequence (DeepUTE). Blinded clinical evaluation relative to the reference low-dose CT (CT-AC) was performed for DeepDixon in 104 independent 2-[18F]fluoro-2-deoxy-d-glucose ([18F]FDG) PET patient studies performed for suspected neurodegenerative disorder using statistical surface projections. RESULTS: Robustness increased with group size in the training data set: 100 subjects were required to reduce the number of outliers compared to a state-of-the-art segmentation-based method, and a cohort >400 subjects further increased robustness in terms of reduced variation and number of outliers. When using transfer learning to adapt to changes in the MRI input, as few as five subjects were sufficient to minimize outliers. Full robustness was achieved at 20 subjects. Comparable robust and accurate results were obtained using all three types of MRI input with a bias below 1% relative to CT-AC in any brain region. The clinical PET evaluation using DeepDixon showed no clinically relevant differences compared to CT-AC. CONCLUSION: Deep learning based AC requires a large training cohort to achieve accurate and robust performance. Using transfer learning, only five subjects were needed to fine-tune the method to large changes to the input images. No clinically relevant differences were found compared to CT-AC, indicating that clinical implementation of our deep learning-based MR-AC method will be feasible across MRI system types using transfer learning and a limited number of subjects.

Quantitative and clinical impact of MRI-based attenuation correction methods in [18F]FDG evaluation of dementia.

BACKGROUND: Positron emission tomography/magnetic resonance imaging (PET/MRI) is a promising diagnostic imaging tool for the diagnosis of dementia, as PET can add complementary information to the routine imaging examination with MRI. The purpose of this study was to evaluate the influence of MRI-based attenuation correction (MRAC) on diagnostic assessment of dementia with [18F]FDG PET. Quantitative differences in both [18F]FDG uptake and z-scores were calculated for three clinically available (DixonNoBone, DixonBone, UTE) and two research MRAC methods (UCL, DeepUTE) compared to CT-based AC (CTAC). Furthermore, diagnoses based on visual evaluations were made by three nuclear medicine physicians and one neuroradiologist (PETCT, PETDeepUTE, PETDixonBone, PETUTE, PETCT + MRI, PETDixonBone + MRI). In addition, pons and cerebellum were compared as reference regions for normalization. RESULTS: The mean absolute difference in z-scores were smallest between MRAC and CTAC with cerebellum as reference region: 0.15 ± 0.11 σ (DeepUTE), 0.15 ± 0.12 σ (UCL), 0.23 ± 0.20 σ (DixonBone), 0.32 ± 0.28 σ (DixonNoBone), and 0.54 ± 0.40 σ (UTE). In the visual evaluation, the diagnoses agreed with PETCT in 74% (PETDeepUTE), 67% (PETDixonBone), and 70% (PETUTE) of the patients, while PETCT + MRI agreed with PETDixonBone + MRI in 89% of the patients. CONCLUSION: The MRAC research methods performed close to that of CTAC in the quantitative evaluation of [18F]FDG uptake and z-scores. Among the clinically implemented MRAC methods, DixonBone should be preferred for diagnostic assessment of dementia with [18F]FDG PET/MRI. However, as artifacts occur in DixonBone attenuation maps, they must be visually inspected to assure proper quantification.

Image quality and detectability in Siemens Biograph PET/MRI and PET/CT systems-a phantom study.

BACKGROUND: The technology of modern positron emission tomography (PET) systems continuously improving, and with it the possibility to detect smaller lesions. Since first introduced in 2010, the number of hybrid PET/magnetic resonance imaging (MRI) systems worldwide is constantly increasing. It is therefore important to assess and compare the image quality, in terms of detectability, between the PET/MRI and the well-established PET/computed tomography (CT) systems. For this purpose, a PET image quality phantom (Esser) with hot spheres, ranging from 4 to 20 mm in diameter, was prepared with fluorodeoxyglucose and sphere-to-background activity concentrations of 8:1 and 4:1, to mimic clinical conditions. The phantom was scanned on a PET/MRI and a PET/CT system for both concentrations to obtain contrast recovery coefficients (CRCs) and contrast-to-noise ratios (CNRs), for a range of reconstruction settings. The detectability of the spheres was scored by three human observers for both systems and concentrations and all reconstructions. Furthermore, the impact of acquisition time on CNR and observer detectability was investigated. RESULTS: Reconstructions applying point-spread-function modeling (and time-of-flight for the PET/CT) yielded the highest CRC and CNR in general, and PET/CT demonstrated slightly higher values than PET/MRI for most sphere sizes. CNR was dependent on reconstruction settings and was maximized for 2 iterations, a pixel size of less than 2 mm and a 4 mm Gaussian filter. Acquisition times of 97 s (PET/MRI) and 150 s (PET/CT) resulted in similar total net true counts. For these acquisition times, the smallest detected spheres by the human observers in the 8:1 activity concentration was the 6-mm sphere with PET/MRI (CNR = 5.6) and the 5-mm sphere with PET/CT (CNR = 5.5). With an acquisition time of 180 s, the 5-mm sphere was also detected with PET/MRI (CNR = 5.8). The 8-mm sphere was the smallest detected sphere in the 4:1 activity concentration for both systems. CONCLUSION: In this experimental study, similar detectability was found for the PET/MRI and the PET/CT, although for an increased acquisition time for the PET/MRI.