Deep learning based low-activity PET reconstruction of [11C]PiB and [18F]FE-PE2I in neurodegenerative disorders.

PURPOSE: Positron Emission Tomography (PET) can support a diagnosis of neurodegenerative disorder by identifying disease-specific pathologies. Our aim was to investigate the feasibility of using activity reduction in clinical [18F]FE-PE2I and [11C]PiB PET/CT scans, simulating low injected activity or scanning time reduction, in combination with AI-assisted denoising. METHODS: A total of 162 patients with clinically uncertain Alzheimer's disease underwent amyloid [11C]PiB PET/CT and 509 patients referred for clinically uncertain Parkinson's disease underwent dopamine transporter (DAT) [18F]FE-PE2I PET/CT. Simulated low-activity data were obtained by random sampling of 5% of the events from the list-mode file and a 5% time window extraction in the middle of the scan. A three-dimensional convolutional neural network (CNN) was trained to denoise the resulting PET images for each disease cohort. RESULTS: Noise reduction of low-activity PET images was successful for both cohorts using 5% of the original activity with improvement in visual quality and all similarity metrics with respect to the ground-truth images. Clinically relevant metrics extracted from the low-activity images deviated < 2% compared to ground-truth values, which were not significantly changed when extracting the metrics from the denoised images. CONCLUSION: The presented models were based on the same network architecture and proved to be a robust tool for denoising brain PET images with two widely different tracer distributions (delocalized, ([11C]PiB, and highly localized, [18F]FE-PE2I). This broad and robust application makes the presented network a good choice for improving the quality of brain images to the level of the standard-activity images without degrading clinical metric extraction. This will allow for reduced dose or scan time in PET/CT to be implemented clinically.

Baseline FDG PET/CT in free breathing versus deep inspiration breath-hold for pediatric patients with mediastinal lymphoma.

INTRODUCTION: The prospective TEDDI protocol investigates the feasibility of radiotherapy delivery in deep inspiration breath-hold (DIBH) for pediatric patients. To secure optimal radiotherapy planning, a diagnostic baseline FDG PET/CT in free breathing (FB) and DIBH was acquired. The anatomical changes in the mediastinum and the effect on PET metrics between the two breathing conditions were assessed for pediatric patients with mediastinal lymphoma. MATERIAL AND METHODS: Ten patients aged 5-17 were included and had a PET/CT in FB and DIBH. Metabolic active lymphoma volumes were manually delineated with a visually based segmentation method and the PET metrics were extracted. The anatomical lymphoma, lung and heart volumes were delineated on CT. RESULTS: The lung volume increased while the heart was displaced caudally and separated from the lymphoma in DIBH compared to FB. Both the anatomical and the metabolically active lymphoma volumes appeared different regarding shape and configuration in the two breathing conditions. The image quality of the DIBH PET was equal to the FB PET regarding interpretation and delineation of lymphoma lesions. All PET metrics increased on the DIBH PET compared to the FB PET with the highest increase observed for the maximum standardized uptake value (33%, range 7-56%). CONCLUSION: Diminished respiratory motion together with anatomical changes within the lymphoma increased all PET metrics in DIBH compared to FB. The anatomical changes observed in DIBH compared to FB are expected to reduce radiation doses to the heart and lungs in pediatric patients with mediastinal lymphoma referred for radiotherapy delivery in DIBH and, thereby, reduce their risk of late effects. TRIAL REGISTRATION: The Danish Ethical Committee (H-16035870, approved November 24th 2016), the Danish Data Protection Agency (2012-58-0004, approved 1 January 2017). Registered retrospectively at clinicaltrials.gov (NCT03315546, 20 October 2017).

Toward PET/MRI as one-stop shop for radiotherapy planning in cervical cancer patients.

BACKGROUND: Radiotherapy (RT) planning for cervical cancer patients entails the acquisition of both Computed Tomography (CT) and Magnetic Resonance Imaging (MRI). Further, molecular imaging by Positron Emission Tomography (PET) could contribute to target volume delineation as well as treatment response monitoring. The objective of this study was to investigate the feasibility of a PET/MRI-only RT planning workflow of patients with cervical cancer. This includes attenuation correction (AC) of MRI hardware and dedicated positioning equipment as well as evaluating MRI-derived synthetic CT (sCT) of the pelvic region for positioning verification and dose calculation to enable a PET/MRI-only setup. MATERIAL AND METHODS: 16 patients underwent PET/MRI using a dedicated RT setup after the routine CT (or PET/CT), including eight pilot patients and eight cervical cancer patients who were subsequently referred for RT. Data from 18 patients with gynecological cancer were added for training a deep convolutional neural network to generate sCT from Dixon MRI. The mean absolute difference between the dose distributions calculated on sCT and a reference CT was measured in the RT target volume and organs at risk. PET AC by sCT and a reference CT were compared in the tumor volume. RESULTS: All patients completed the examination. sCT was inferred for each patient in less than 5 s. The dosimetric analysis of the sCT-based dose planning showed a mean absolute error (MAE) of 0.17 ± 0.12 Gy inside the planning target volumes (PTV). PET images reconstructed with sCT and CT had no significant difference in quantification for all patients. CONCLUSIONS: These results suggest that multiparametric PET/MRI can be successfully integrated as a one-stop-shop in the RT workflow of patients with cervical cancer.

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.

Combined PET/MR imaging in neurology: MR-based attenuation correction implies a strong spatial bias when ignoring bone.

AIM: Combined PET/MR systems have now become available for clinical use. Given the lack of integrated standard transmission (TX) sources in these systems, attenuation and scatter correction (AC) must be performed using the available MR-images. Since bone tissue cannot easily be accounted for during MR-AC, PET quantification can be biased, in particular, in the vicinity of the skull. Here, we assess PET quantification in PET/MR imaging of patients using phantoms and patient data. MATERIALS AND METHODS: Nineteen patients referred to our clinic for a PET/CT exam as part of the diagnostic evaluation of suspected dementia were included in our study. The patients were injected with 200MBq [(18)F]FDG and imaged with PET/CT and PET/MR in random sequence within 1h. Both, PET/CT and PET/MR were performed as single-bed acquisitions without contrast administration. PET/CT and PET/MR data were reconstructed following CT-based and MR-based AC, respectively. MR-AC was performed based on: (A) standard Dixon-Water-Fat segmentation (DWFS), (B) DWFS with co-registered and segmented CT bone values superimposed, and (C) with co-registered full CT-based attenuation image. All PET images were reconstructed using AW-OSEM, with neither resolution recovery nor time-of-flight option employed. PET/CT (D) or PET/MR (A-C) images were decay-corrected to the start time of the first examination. PET images following AC were evaluated visually and quantitatively using 10 homeomorphic regions of interest drawn on a transaxial T1w-MR image traversing the central basal ganglia. We report the relative difference (%) of the mean ROI values for (A)-(C) in reference to PET/CT (D). In a separate phantom experiment a 2L plastic bottle was layered with approximately 12mm of Gypsum plaster to mimic skull bone. The phantom was imaged on PET/CT only and standard MR-AC was performed by replacing hyperdense CT attenuation values corresponding to bone (plaster) with attenuation values of water. PET image reconstruction was performed with CT-AC (D) and CT-AC using the modified CT images corresponding to MR-AC using DWFS (A). RESULTS: PET activity values in patients following MR-AC (A) showed a substantial radial dependency when compared to PET/CT. In all patients cortical PET activity was lower than the activity in the central region of the brain (10-15%). When adding bone attenuation values to standard MR-AC (B and C) the radial gradient of PET activity values was removed. Further evaluation of PET/MR activity following MR-AC (A) relative to MR-AC (C) using the full CT for attenuation correction showed an underestimation of 25% in the cortical regions and 5-10% in the central regions of the brain. Observations in patients were replicated by observations from the phantom study. CONCLUSION: Our phantom and patient data demonstrate a spatially varying bias of the PET activity in PET/MR images of the brain when bone tissue is not accounted for during attenuation correction. This has immediate implications for PET/MR imaging of the brain. Therefore, refinements to existing MR-AC methods or alternative strategies need to be found prior to adopting PET/MR imaging of the brain in clinical routine and research.

Improved Positron Emission Tomography Quantification: Evaluation of a Maximum-Likelihood Scatter Scaling Algorithm.

Incorrect scatter scaling of positron emission tomography (PET) images can lead to halo artifacts, quantitative bias, or reconstruction failure. Tail-fitted scatter scaling (TFSS) possesses performance limitations in multiple cases. This study aims to investigate a novel method for scatter scaling: maximum-likelihood scatter scaling (MLSS) in scenarios where TFSS tends to induce artifacts or are observed to cause reconstruction abortion. [68Ga]Ga-RGD PET scans of nine patients were included in cohort 1 in the scope of investigating the reduction of halo artifacts relative to the scatter estimation method. PET scans of 30 patients administrated with [68Ga]Ga-uPAR were included in cohort 2, used for an evaluation of the robustness of MLSS in cases where TFSS-integrated reconstructions are observed to fail. A visual inspection of MLSS-corrected images scored higher than TFSS-corrected reconstructions of cohort 1. The quantitative investigation near the bladder showed a relative difference in tracer uptake of up to 94.7%. A reconstruction of scans included in cohort 2 resulted in failure in 23 cases when TFSS was used. The lesion uptake values of cohort 2 showed no significant difference. MLSS is suggested as an alternative scatter-scaling method relative to TFSS with the aim of reducing halo artifacts and a robust reconstruction process.

PET/MR imaging of the pelvis in the presence of endoprostheses: reducing image artifacts and increasing accuracy through inpainting.

PURPOSE: In combined whole-body PET/MR, attenuation correction (AC) is performed indirectly using the available MR image information and subsequent segmentation. Implant-induced susceptibility artifacts and subsequent signal voids may challenge MR-based AC (MR-AC). We evaluated the accuracy of MR-AC in PET/MR in patients with metallic endoprostheses, and propose a clinically feasible correction method. METHODS: We selected patients with uni- or bilateral endoprostheses from 61 consecutive referrals for whole-body PET/MR imaging (mMR; Siemens Healthcare). Simultaneous whole-body PET/MR imaging was performed at 120 min after injection of about 300 MBq [(18)F]FDG. MR-AC was performed using (1) original MR images and subsequent Dixon water-fat segmentation, (2) as method 1 with implant-induced signal voids filled with soft tissue, (3) as method 2 with superimposed coregistered endoprostheses from the CT scan, and (4) as method 1 with implant-induced signal voids filled with metal. Following MR-AC (methods 1-4) PET emission images were reconstructed on 344 × 344 matrices using attenuation-weighted OSEM (three iterations, 21 subsets, 4 mm gaussian). Maximum body-weight normalized standardized uptake values (SUVmax) were obtained for both hips. Mean SUV (SUVmean) in homogeneous reference regions in the gluteal muscle and bladder following MR-AC (methods 1-4) are also reported. RESULTS: In total, four patients presented with endoprostheses, unilateral in two and bilateral in two. The fraction of voxels in MR images affected by the implant was at least twice that of the voxels representing the actual implants. MR-AC using methods 2 and 3 recovered the FDG distribution pattern compared to uncorrected PET images and method 1, while method 4 resulted in severe overestimation of FDG uptake (>460 % SUVmax). When compared to method 1, relative changes in SUVmean in the reference regions from method 2 and 3 were generally small albeit not correlated with the fraction of the attenuation image affected by implant-induced artifacts. CONCLUSIONS: Endoprostheses cause PET/MR artifacts that exceed the volume occupied by the implants, and bias PET quantification. Artifacts and bias can be corrected by semiautomated inpainting with soft tissue with a single composition prior to MR-AC, thus restoring quantitative activity distribution.

Impact of Reduced Image Noise on Deauville Scores in Patients with Lymphoma Scanned on a Long-Axial Field-of-View PET/CT-Scanner.

BACKGROUND: Total body and long-axial field-of-view (LAFOV) PET/CT represent visionary innovations in imaging enabling either improved image quality, reduction in injected activity-dose or decreased acquisition time. An improved image quality may affect visual scoring systems, including the Deauville score (DS), which is used for clinical assessment of patients with lymphoma. The DS compares SUVmax in residual lymphomas with liver parenchyma, and here we investigate the impact of reduced image noise on the DS in patients with lymphomas scanned on a LAFOV PET/CT. METHODS: Sixty-eight patients with lymphoma underwent a whole-body scan on a Biograph Vision Quadra PET/CT-scanner, and images were evaluated visually with regard to DS for three different timeframes of 90, 300, and 600 s. SUVmax and SUVmean were calculated from liver and mediastinal blood pool, in addition to SUVmax from residual lymphomas and measures of noise. RESULTS: SUVmax in liver and in mediastinal blood pool decreased significantly with increasing acquisition time, whereas SUVmean remained stable. In residual tumor, SUVmax was stable during different acquisition times. As a result, the DS was subject to change in three patients. CONCLUSIONS: Attention should be drawn towards the eventual impact of improvements in image quality on visual scoring systems such as the DS.

Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI.

In the context of brain tumour response assessment, deep learning-based three-dimensional (3D) tumour segmentation has shown potential to enter the routine radiological workflow. The purpose of the present study was to perform an external evaluation of a state-of-the-art deep learning 3D brain tumour segmentation algorithm (HD-GLIO) on an independent cohort of consecutive, post-operative patients. For 66 consecutive magnetic resonance imaging examinations, we compared delineations of contrast-enhancing (CE) tumour lesions and non-enhancing T2/FLAIR hyperintense abnormality (NE) lesions by the HD-GLIO algorithm and radiologists using Dice similarity coefficients (Dice). Volume agreement was assessed using concordance correlation coefficients (CCCs) and Bland-Altman plots. The algorithm performed very well regarding the segmentation of NE volumes (median Dice = 0.79) and CE tumour volumes larger than 1.0 cm3 (median Dice = 0.86). If considering all cases with CE tumour lesions, the performance dropped significantly (median Dice = 0.40). Volume agreement was excellent with CCCs of 0.997 (CE tumour volumes) and 0.922 (NE volumes). The findings have implications for the application of the HD-GLIO algorithm in the routine radiological workflow where small contrast-enhancing tumours will constitute a considerable share of the follow-up cases. Our study underlines that independent validations on clinical datasets are key to asserting the robustness of deep learning algorithms.

DeepDixon synthetic CT for [18F]FET PET/MRI attenuation correction of post-surgery glioma patients with metal implants.

Purpose: Conventional magnetic resonance imaging (MRI) can for glioma assessment be supplemented by positron emission tomography (PET) imaging with radiolabeled amino acids such as O-(2-[18F]fluoroethyl)-L-tyrosine ([18F]FET), which provides additional information on metabolic properties. In neuro-oncology, patients often undergo brain and skull altering treatment, which is known to challenge MRI-based attenuation correction (MR-AC) methods and thereby impact the simplified semi-quantitative measures such as tumor-to-brain ratio (TBR) used in clinical routine. The aim of the present study was to examine the applicability of our deep learning method, DeepDixon, for MR-AC in [18F]FET PET/MRI scans of a post-surgery glioma cohort with metal implants. Methods: The MR-AC maps were assessed for all 194 included post-surgery glioma patients (318 studies). The subgroup of 147 patients (222 studies, 200 MBq [18F]FET PET/MRI) with tracer uptake above 1 ml were subsequently reconstructed with DeepDixon, vendor-default atlas-based method, and a low-dose computed tomography (CT) used as reference. The biological tumor volume (BTV) was delineated on each patient by isocontouring tracer uptake above a TBR threshold of 1.6. We evaluated the MR-AC methods using the recommended clinical metrics BTV and mean and maximum TBR on a patient-by-patient basis against the reference with CT-AC. Results: Ninety-seven percent of the studies (310/318) did not have any major artifacts using DeepDixon, which resulted in a Dice coefficient of 0.89/0.83 for tissue/bone, respectively, compared to 0.84/0.57 when using atlas. The average difference between DeepDixon and CT-AC was within 0.2% across all clinical metrics, and no statistically significant difference was found. When using DeepDixon, only 3 out of 222 studies (1%) exceeded our acceptance criteria compared to 72 of the 222 studies (32%) with the atlas method. Conclusion: We evaluated the performance of a state-of-the-art MR-AC method on the largest post-surgical glioma patient cohort to date. We found that DeepDixon could overcome most of the issues arising from irregular anatomy and metal artifacts present in the cohort resulting in clinical metrics within acceptable limits of the reference CT-AC in almost all cases. This is a significant improvement over the vendor-provided atlas method and of particular importance in response assessment.

Estimation of brain amyloid accumulation using deep learning in clinical [11C]PiB PET imaging.

INTRODUCTION: Estimation of brain amyloid accumulation is valuable for evaluation of patients with cognitive impairment in both research and clinical routine. The development of high throughput and accurate strategies for the determination of amyloid status could be an important tool in patient selection for clinical trials and amyloid directed treatment. Here, we propose the use of deep learning to quantify amyloid accumulation using standardized uptake value ratio (SUVR) and classify amyloid status based on their PET images. METHODS: A total of 1309 patients with cognitive impairment scanned with [11C]PIB PET/CT or PET/MRI were included. Two convolutional neural networks (CNNs) for reading-based amyloid status and SUVR prediction were trained using 75% of the PET/CT data. The remaining PET/CT (n = 300) and all PET/MRI (n = 100) data was used for evaluation. RESULTS: The prevalence of amyloid positive patients was 61%. The amyloid status classification model reproduced the expert reader's classification with 99% accuracy. There was a high correlation between reference and predicted SUVR (R2 = 0.96). Both reference and predicted SUVR had an accuracy of 97% compared to expert classification when applying a predetermined SUVR threshold of 1.35 for binary classification of amyloid status. CONCLUSION: The proposed CNN models reproduced both the expert classification and quantitative measure of amyloid accumulation in a large local dataset. This method has the potential to replace or simplify existing clinical routines and can facilitate fast and accurate classification well-suited for a high throughput pipeline.

Attenuation Correction of Long Axial Field-of-View Positron Emission Tomography Using Synthetic Computed Tomography Derived from the Emission Data: Application to Low-Count Studies and Multiple Tracers.

Recent advancements in PET/CT, including the emergence of long axial field-of-view (LAFOV) PET/CT scanners, have increased PET sensitivity substantially. Consequently, there has been a significant reduction in the required tracer activity, shifting the primary source of patient radiation dose exposure to the attenuation correction (AC) CT scan during PET imaging. This study proposes a parameter-transferred conditional generative adversarial network (PT-cGAN) architecture to generate synthetic CT (sCT) images from non-attenuation corrected (NAC) PET images, with separate networks for [18F]FDG and [15O]H2O tracers. The study includes a total of 1018 subjects (n = 972 [18F]FDG, n = 46 [15O]H2O). Testing was performed on the LAFOV scanner for both datasets. Qualitative analysis found no differences in image quality in 30 out of 36 cases in FDG patients, with minor insignificant differences in the remaining 6 cases. Reduced artifacts due to motion between NAC PET and CT were found. For the selected organs, a mean average error of 0.45% was found for the FDG cohort, and that of 3.12% was found for the H2O cohort. Simulated low-count images were included in testing, which demonstrated good performance down to 45 s scans. These findings show that the AC of total-body PET is feasible across tracers and in low-count studies and might reduce the artifacts due to motion and metal implants.

Potential Clinical Impact of LAFOV PET/CT: A Systematic Evaluation of Image Quality and Lesion Detection.

We performed a systematic evaluation of the diagnostic performance of LAFOV PET/CT with increasing acquisition time. The first 100 oncologic adult patients referred for 3 MBq/kg 2-[18F]fluoro-2-deoxy-D-glucose PET/CT on the Siemens Biograph Vision Quadra were included. A standard imaging protocol of 10 min was used and scans were reconstructed at 30 s, 60 s, 90 s, 180 s, 300 s, and 600 s. Paired comparisons of quantitative image noise, qualitative image quality, lesion detection, and lesion classification were performed. Image noise (n = 50, 34 women) was acceptable according to the current standard of care (coefficient-of-varianceref < 0.15) after 90 s and improved significantly with increasing acquisition time (PB < 0.001). The same was seen in observer rankings (PB < 0.001). Lesion detection (n = 100, 74 women) improved significantly from 30 s to 90 s (PB < 0.001), 90 s to 180 s (PB = 0.001), and 90 s to 300 s (PB = 0.002), while lesion classification improved from 90 s to 180 s (PB < 0.001), 180 s to 300 s (PB = 0.021), and 90 s to 300 s (PB < 0.001). We observed improved image quality, lesion detection, and lesion classification with increasing acquisition time while maintaining a total scan time of less than 5 min, which demonstrates a potential clinical benefit. Based on these results we recommend a standard imaging acquisition protocol for LAFOV PET/CT of minimum 180 s to maximum 300 s after injection of 3 MBq/kg 2-[18F]fluoro-2-deoxy-D-glucose.

Scanner agnostic large-scale evaluation of MS lesion delineation tool for clinical MRI.

Introduction: Patients with MS are MRI scanned continuously throughout their disease course resulting in a large manual workload for radiologists which includes lesion detection and size estimation. Though many models for automatic lesion segmentation have been published, few are used broadly in clinic today, as there is a lack of testing on clinical datasets. By collecting a large, heterogeneous training dataset directly from our MS clinic we aim to present a model which is robust to different scanner protocols and artefacts and which only uses MRI modalities present in routine clinical examinations. Methods: We retrospectively included 746 patients from routine examinations at our MS clinic. The inclusion criteria included acquisition at one of seven different scanners and an MRI protocol including 2D or 3D T2-w FLAIR, T2-w and T1-w images. Reference lesion masks on the training (n = 571) and validation (n = 70) datasets were generated using a preliminary segmentation model and subsequent manual correction. The test dataset (n = 100) was manually delineated. Our segmentation model https://github.com/CAAI/AIMS/ was based on the popular nnU-Net, which has won several biomedical segmentation challenges. We tested our model against the published segmentation models HD-MS-Lesions, which is also based on nnU-Net, trained with a more homogenous patient cohort. We furthermore tested model robustness to data from unseen scanners by performing a leave-one-scanner-out experiment. Results: We found that our model was able to segment MS white matter lesions with a performance comparable to literature: DSC = 0.68, precision = 0.90, recall = 0.70, f1 = 0.78. Furthermore, the model outperformed HD-MS-Lesions in all metrics except precision = 0.96. In the leave-one-scanner-out experiment there was no significant change in performance (p < 0.05) between any of the models which were only trained on part of the dataset and the full segmentation model. Conclusion: In conclusion we have seen, that by including a large, heterogeneous dataset emulating clinical reality, we have trained a segmentation model which maintains a high segmentation performance while being robust to data from unseen scanners. This broadens the applicability of the model in clinic and paves the way for clinical implementation.

A zero-dose synthetic baseline for the personalized analysis of [18F]FDG-PET: Application in Alzheimer's disease.

Purpose: Brain 2-Deoxy-2-[18F]fluoroglucose ([18F]FDG-PET) is widely used in the diagnostic workup of Alzheimer's disease (AD). Current tools for uptake analysis rely on non-personalized templates, which poses a challenge as decreased glucose uptake could reflect neuronal dysfunction, or heterogeneous brain morphology associated with normal aging. Overcoming this, we propose a deep learning method for synthesizing a personalized [18F]FDG-PET baseline from the patient's own MRI, and showcase its applicability in detecting AD pathology. Methods: We included [18F]FDG-PET/MRI data from 123 patients of a local cohort and 600 patients from ADNI. A supervised, adversarial model with two connected Generative Adversarial Networks (GANs) was trained on cognitive normal (CN) patients with transfer-learning to generate full synthetic baseline volumes (sbPET) (192 × 192 × 192) which reflect healthy uptake conditioned on brain anatomy. Synthetic accuracy was measured by absolute relative %-difference (Abs%), relative %-difference (RD%), and peak signal-to-noise ratio (PSNR). Lastly, we deployed the sbPET images in a fully personalized method for localizing metabolic abnormalities. Results: The model achieved a spatially uniform Abs% of 9.4%, RD% of 0.5%, and a PSNR of 26.3 for CN subjects. The sbPET images conformed to the anatomical information dictated by the MRI and proved robust in presence of atrophy. The personalized abnormality method correctly mapped the pathology of AD subjects while showing little to no anomalies for CN subjects. Conclusion: This work demonstrated the feasibility of synthesizing fully personalized, healthy-appearing [18F]FDG-PET images. Using these, we showcased a promising application in diagnosing AD, and theorized the potential value of sbPET images in other neuroimaging routines.

A convolutional neural network for total tumor segmentation in [64Cu]Cu-DOTATATE PET/CT of patients with neuroendocrine neoplasms.

BACKGROUND: Segmentation of neuroendocrine neoplasms (NENs) in [64Cu]Cu-DOTATATE positron emission tomography makes it possible to extract quantitative measures useable for prognostication of patients. However, manual tumor segmentation is cumbersome and time-consuming. Therefore, we aimed to implement and test an artificial intelligence (AI) network for tumor segmentation. Patients with gastroenteropancreatic or lung NEN with [64Cu]Cu-DOTATATE PET/CT performed were included in our training (n = 117) and test cohort (n = 41). Further, 10 patients with no signs of NEN were included as negative controls. Ground truth segmentations were obtained by a standardized semiautomatic method for tumor segmentation by a physician. The nnU-Net framework was used to set up a deep learning U-net architecture. Dice score, sensitivity and precision were used for selection of the final model. AI segmentations were implemented in a clinical imaging viewer where a physician evaluated performance and performed manual adjustments. RESULTS: Cross-validation training was used to generate models and an ensemble model. The ensemble model performed best overall with a lesion-wise dice of 0.850 and pixel-wise dice, precision and sensitivity of 0.801, 0.786 and 0.872, respectively. Performance of the ensemble model was acceptable with some degree of manual adjustment in 35/41 (85%) patients. Final tumor segmentation could be obtained from the AI model with manual adjustments in 5 min versus 17 min for ground truth method, p < 0.01. CONCLUSION: We implemented and validated an AI model that achieved a high similarity with ground truth segmentation and resulted in faster tumor segmentation. With AI, total tumor segmentation may become feasible in the clinical routine.

The effect of reduced scan time on response assessment FDG-PET/CT imaging using Deauville score in patients with lymphoma.

PURPOSE: [18F]Fluoro-deoxy-glucose positron emission tomography/computed tomography (FDG-PET/CT) is used for response assessment during therapy in Hodgkin lymphoma (HL) and non-Hodgkin lymphoma (NHL). Clinicians report the scans visually using Deauville criteria. Improved performance in modern PET/CT scanners could allow for a reduction in scan time without compromising diagnostic image quality. Additionally, patient throughput can be increased with increasing cost-effectiveness. We investigated the effects of reducing scan time of response assessment FDG-PET/CT in HL and NHL patients on Deauville score (DS) and image quality. METHODS: Twenty patients diagnosed with HL/NHL referred to a response assessment FDG-PET/CT were included. PET scans were performed in list-mode with an acquisition time of 120 s per bed position(s/bp). From PET list-mode data images with full acquisition time of 120 s/bp and shorter acquisition times (90, 60, 45, and 30 s/bp) were reconstructed. All images were assessed by two specialists and assigned a DS. We estimated the possible savings when reducing scan time using a simplified model based on assumed values/costs for our hospital. RESULTS: There were no significant changes in the visually assessed DS when reducing scan time to 90 s/bp, 60 s/bp, 45 s/bp, and 30 s/bp. Image quality of 90 s/bp images were rated equal to 120 s/bp images. Coefficient of variance values for 120 s/bp and 90 s/bp images was significantly < 15%. The estimated annual savings to the hospital when reducing scan time was 8000-16,000 €/scanner. CONCLUSION: Acquisition time can be reduced to 90 s/bp in response assessment FDG-PET/CT without compromising Deauville score or image quality. Reducing acquisition time can reduce costs to the clinic.

Low-dose PET image noise reduction using deep learning: application to cardiac viability FDG imaging in patients with ischemic heart disease.

INTRODUCTION: Cardiac [18F]FDG-PET is widely used for viability testing in patients with chronic ischemic heart disease. Guidelines recommend injection of 200-350 MBq [18F]FDG, however, a reduction of radiation exposure has become increasingly important, but might come at the cost of reduced diagnostic accuracy due to the increased noise in the images. We aimed to explore the use of a common deep learning (DL) network for noise reduction in low-dose PET images, and to validate its accuracy using the clinical quantitative metrics used to determine cardiac viability in patients with ischemic heart disease. METHODS: We included 168 patients imaged with cardiac [18F]FDG-PET/CT. We simulated a reduced dose by keeping counts at thresholds 1% and 10%. 3D U-net with five blocks was trained to de-noise full PET volumes (128 × 128 × 111). The low-dose and de-noised images were compared in Corridor4DM to the full-dose PET images. We used the default segmentation of the left ventricle to extract the quantitative metrics end-diastolic volume (EDV), end-systolic volume (ESV), and left ventricular ejection fraction (LVEF) from the gated images, and FDG defect extent from the static images. RESULTS: Our de-noising models were able to recover the PET signal for both the static and gated images in either dose-reduction. For the 1% low-dose images, the error is most pronounced for EDV and ESV, where the average underestimation is 25%. No bias was observed using the proposed DL de-noising method. De-noising minimized the outliers found for the 1% and 10% low-dose measurements of LVEF and extent. Accuracy of differential diagnosis based on LVEF threshold was highly improved after de-noising. CONCLUSION: A significant dose reduction can be achieved for cardiac [18F]FDG images used for viability testing in patients with ischemic heart disease without significant loss of diagnostic accuracy when using our DL model for noise reduction. Both 1% and 10% dose reductions are possible with clinically quantitative metrics comparable to that obtained with a full dose.

[Artificial intelligence for diagnostic imaging].

Artificial intelligence (AI) is a computer-based system, which in diagnostic imaging can improve patient flow, optimise image processing, shorten scan time, reduce radiation dose and be used as decision aid in image interpretation. In this review, we argue that AI algorithms should be based on evidence with initial hypothesis, then a choice of algorithm and development with training on the initial data set; afterwards the algorithms should be tested on a new representative dataset, and finally it should be tested in a prospective study. If the AI is evidence-based and can solve a task better or cheaper than the usual methodology, it can be implemented.