A convolutional neural network for fully automated blood SUV determination to facilitate SUR computation in oncological FDG-PET.
Ontology highlight
ABSTRACT: Purpose
The standardized uptake value (SUV) is widely used for quantitative evaluation in oncological FDG-PET but has well-known shortcomings as a measure of the tumor's glucose consumption. The standard uptake ratio (SUR) of tumor SUV and arterial blood SUV (BSUV) possesses an increased prognostic value but requires image-based BSUV determination, typically in the aortic lumen. However, accurate manual ROI delineation requires care and imposes an additional workload, which makes the SUR approach less attractive for clinical routine. The goal of the present work was the development of a fully automated method for BSUV determination in whole-body PET/CT.Methods
Automatic delineation of the aortic lumen was performed with a convolutional neural network (CNN), using the U-Net architecture. A total of 946 FDG PET/CT scans from several sites were used for network training (N = 366) and testing (N = 580). For all scans, the aortic lumen was manually delineated, avoiding areas affected by motion-induced attenuation artifacts or potential spillover from adjacent FDG-avid regions. Performance of the network was assessed using the fractional deviations of automatically and manually derived BSUVs in the test data.Results
The trained U-Net yields BSUVs in close agreement with those obtained from manual delineation. Comparison of manually and automatically derived BSUVs shows excellent concordance: the mean relative BSUV difference was (mean ± SD) = (- 0.5 ± 2.2)% with a 95% confidence interval of [- 5.1,3.8]% and a total range of [- 10.0, 12.0]%. For four test cases, the derived ROIs were unusable (< 1 ml).Conclusion
CNNs are capable of performing robust automatic image-based BSUV determination. Integrating automatic BSUV derivation into PET data processing workflows will significantly facilitate SUR computation without increasing the workload in the clinical setting.
SUBMITTER: Nikulin P
PROVIDER: S-EPMC8041711 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Quantitative assessment of radio- and chemotherapy response with 18F-FDG whole-body PET has attracted increasing interest in recent years. In most published work, SUV has been used for this purpose. In the context of therapy response assessment, the reliability of lesion SUVs, notably their test-retest stability, thus becomes crucial. However, a recent study demonstrated substantial test-retest variability (TRV) in SUVs. The purpose of the present study was to investigate whether the tumor-to-blood SUV ratio (SUR) can improve TRV in tracer uptake. Methods: 73 patients with advanced non-small cell lung cancer from the prospective multicenter trials ACRIN 6678 (n = 34) and MK-0646-008 (n = 39) were included in this study. All patients underwent two 18F-FDG PET/CT investigations on two different days (time difference, 3.6 ± 2.1 d; range, 1-7 d) before therapy. For each patient, up to 7 tumor lesions were evaluated. For each lesion, SUVmax and SUVpeak were determined. Blood SUV was determined as the mean value of a 3-dimensional aortic region of interest that was delineated on the attenuation CT image and transferred to the PET image. SURs were computed as the ratio of tumor SUV to blood SUV and were uptake time-corrected to 75 min after injection. TRV was quantified as 1.96 multiplied by the root-mean-square deviation of the fractional paired differences in SUV and SUR. The combined effect of blood normalization and uptake time correction was inspected by considering RTRV (TRVSUR/TRVSUV), a ratio reflecting the reduction in the TRV in SUR relative to SUV. RTRV was correlated with the group-averaged-value difference (?) in CFmean (?CFmean) of the quantity ?CF = |CF - 1|, where CF is the numeric factor that converts individual ratios of paired SUVs into corresponding SURs. This correlation analysis was performed by successively increasing a threshold value ?CFmin and computing ?CFmean and RTRV for the remaining subgroup of patients/lesions with ?CF ? ?CFminResults: The group-averaged TRVSUV and TRVSUR were 32.1 and 29.0, respectively, which correspond to a reduction of variability in SUR by an RTRV factor of 0.9 in comparison to SUV. This rather marginal improvement can be understood to be a consequence of the atypically low intrasubject variability in blood SUV and uptake time and the accordingly small ?CF values in the investigated prospective study groups. In fact, subgroup analysis with increasing ?CFmin thresholds revealed a pronounced negative correlation (Spearman ? = -0.99, P < 0.001) between RTRV and ?CFmean, where RTRV ? 0.4 in the ?CFmin = 20% subgroup, corresponding to a more than 2-fold reduction of TRVSUR compared with TRVSUVConclusion: Variability in blood SUV and uptake time has been identified as a causal factor in the TRV in lesion SUV. Therefore, TRV in lesion uptake measurements can be reduced by replacing SUV with SUR as the uptake measure. The improvement becomes substantial for the level of variability in blood SUV and uptake time typically observed in the clinical context.
| S-EPMC5666644 | biostudies-literature
Project description:BACKGROUND:Today, the standardized uptake value (SUV) is essentially the only means for quantitative evaluation of static [18F-]fluorodeoxyglucose (FDG) positron emission tomography (PET) investigations. However, the SUV approach has several well-known shortcomings which adversely affect the reliability of the SUV as a surrogate of the metabolic rate of glucose consumption. The standard uptake ratio (SUR), i.e., the uptake time-corrected ratio of tumor SUV to image-derived arterial blood SUV, has been shown in the first clinical studies to overcome most of these shortcomings, to decrease test-retest variability, and to increase the prognostic value in comparison to SUV. However, it is unclear, to what extent the SUR approach is vulnerable to observer variability of the additionally required blood SUV (BSUV) determination. The goal of the present work was the investigation of the interobserver variability of image-derived BSUV. METHODS:FDG PET/CT scans from 83 patients (72 male, 11 female) with non-small cell lung cancer (N = 46) or head and neck cancer (N = 37) were included. BSUV was determined by 8 individuals, each applying a dedicated delineation tool for the BSUV determination in the aorta. Two of the observers applied two further tools. Altogether, five different delineation tools were used. With each used tool, delineation was performed for the whole patient group, resulting in 12 distinct observations per patient. Intersubject variability of BSUV determination was assessed using the fractional deviations for the individual patients from the patient group average and was quantified as standard deviation (SD is), 95% confidence interval, and range. Interobserver variability of BSUV determination was assessed using the fractional deviations of the individual observers from the observer-average for the considered patient and quantified as standard deviations (SD p, SD d) or root mean square (RMS), 95% confidence interval, and range in each patient, each observer, and the pooled data respectively. RESULTS:Interobserver variability in the pooled data amounts to RMS = 2.8% and is much smaller than the intersubject variability of BSUV (SD is= 16%). Averaged over the whole patient group, deviations of individual observers from the observer average are very small and fall in the range [ -?0.96, 1.05]%. However, interobserver variability partly differs distinctly for different patients, covering a range of [0.7, 7.4]% in the investigated patient group. CONCLUSION:The present investigation demonstrates that the image-based manual determination of BSUV in the aorta is sufficiently reproducible across different observers and delineation tools which is a prerequisite for accurate SUR determination. This finding is in line with the already demonstrated superior prognostic value of SUR in comparison to SUV in the first clinical studies.
| S-EPMC6399366 | biostudies-literature
Project description:ObjectivesThe objective of this study was to assess the feasibility and accuracy of a fully automated artificial intelligence (AI) powered coronary artery calcium scoring (CACS) method on ungated CT in oncologic patients undergoing 18F-FDG PET/CT.MethodsA total of 100 oncologic patients examined between 2007 and 2015 were retrospectively included. All patients underwent 18F-FDG PET/CT and cardiac SPECT myocardial perfusion imaging (MPI) by 99mTc-tetrofosmin within 6 months. CACS was manually performed on non-contrast ECG-gated CT scans obtained from SPECT-MPI (i.e., reference standard). Additionally, CACS was performed using a cloud-based, user-independent tool (AI-CACS) on ungated CT scans from 18F-FDG-PET/CT examinations. Agatston scores from the manual CACS and AI-CACS were compared.ResultsOn a per-patient basis, the AI-CACS tool achieved a sensitivity and specificity of 85% and 90% for the detection of CAC. Interscore agreement of CACS between manual CACS and AI-CACS was 0.88 (95% CI: 0.827, 0.918). Interclass agreement of risk categories was 0.8 in weighted Kappa analysis, with a reclassification rate of 44% and an underestimation of one risk category by AI-CACS in 39% of cases. On a per-vessel basis, interscore agreement of CAC scores ranged from 0.716 for the circumflex artery to 0.863 for the left anterior descending artery.ConclusionsFully automated AI-CACS as performed on non-contrast free-breathing, ungated CT scans from 18F-FDG-PET/CT examinations is feasible and provides an acceptable to good estimation of CAC burden. CAC load on ungated CT is, however, generally underestimated by AI-CACS, which should be taken into account when interpreting imaging findings.
| S-EPMC9406755 | biostudies-literature
Project description:Cardiovascular resonance (CMR) imaging is a standard imaging modality for assessing cardiovascular diseases (CVDs), the leading cause of death globally. CMR enables accurate quantification of the cardiac chamber volume, ejection fraction and myocardial mass, providing information for diagnosis and monitoring of CVDs. However, for years, clinicians have been relying on manual approaches for CMR image analysis, which is time consuming and prone to subjective errors. It is a major clinical challenge to automatically derive quantitative and clinically relevant information from CMR images.Deep neural networks have shown a great potential in image pattern recognition and segmentation for a variety of tasks. Here we demonstrate an automated analysis method for CMR images, which is based on a fully convolutional network (FCN). The network is trained and evaluated on a large-scale dataset from the UK Biobank, consisting of 4,875 subjects with 93,500 pixelwise annotated images. The performance of the method has been evaluated using a number of technical metrics, including the Dice metric, mean contour distance and Hausdorff distance, as well as clinically relevant measures, including left ventricle (LV) end-diastolic volume (LVEDV) and end-systolic volume (LVESV), LV mass (LVM); right ventricle (RV) end-diastolic volume (RVEDV) and end-systolic volume (RVESV).By combining FCN with a large-scale annotated dataset, the proposed automated method achieves a high performance in segmenting the LV and RV on short-axis CMR images and the left atrium (LA) and right atrium (RA) on long-axis CMR images. On a short-axis image test set of 600 subjects, it achieves an average Dice metric of 0.94 for the LV cavity, 0.88 for the LV myocardium and 0.90 for the RV cavity. The mean absolute difference between automated measurement and manual measurement is 6.1 mL for LVEDV, 5.3 mL for LVESV, 6.9 gram for LVM, 8.5 mL for RVEDV and 7.2 mL for RVESV. On long-axis image test sets, the average Dice metric is 0.93 for the LA cavity (2-chamber view), 0.95 for the LA cavity (4-chamber view) and 0.96 for the RA cavity (4-chamber view). The performance is comparable to human inter-observer variability.We show that an automated method achieves a performance on par with human experts in analysing CMR images and deriving clinically relevant measures.
| S-EPMC6138894 | biostudies-other
Project description:BackgroundGastroenteropancreatic neuroendocrine tumors most commonly metastasize to the liver; however, high normal background 68Ga-DOTATATE activity and high image noise make metastatic lesions difficult to detect. The purpose of this study is to develop a rapid, automated and highly specific method to identify 68Ga-DOTATATE PET/CT hepatic lesions using a 2D U-Net convolutional neural network.MethodsA retrospective study of 68Ga-DOTATATE PET/CT patient studies (n = 125; 57 with 68Ga-DOTATATE hepatic lesions and 68 without) was evaluated. The dataset was randomly divided into 75 studies for the training set (36 abnormal, 39 normal), 25 for the validation set (11 abnormal, 14 normal) and 25 for the testing set (11 abnormal, 14 normal). Hepatic lesions were physician annotated using a modified PERCIST threshold, and boundary definition by gradient edge detection. The 2D U-Net was trained independently five times for 100,000 iterations using a linear combination of binary cross-entropy and dice losses with a stochastic gradient descent algorithm. Performance metrics included: positive predictive value (PPV), sensitivity, F1 score and area under the precision-recall curve (PR-AUC). Five different pixel area thresholds were used to filter noisy predictions.ResultsA total of 233 lesions were annotated with each abnormal study containing a mean of 4 ± 2.75 lesions. A pixel filter of 20 produced the highest mean PPV 0.94 ± 0.01. A pixel filter of 5 produced the highest mean sensitivity 0.74 ± 0.02. The highest mean F1 score 0.79 ± 0.01 was produced with a 20 pixel filter. The highest mean PR-AUC 0.73 ± 0.03 was produced with a 15 pixel filter.ConclusionDeep neural networks can automatically detect hepatic lesions in 68Ga-DOTATATE PET. Ongoing improvements in data annotation methods, increasing sample sizes and training methods are anticipated to further improve detection performance.
| S-EPMC8487415 | biostudies-literature
Project description:PurposeThe identification of pathological mediastinal lymph nodes is an important step in the staging of lung cancer, with the presence of metastases significantly affecting survival rates. Nodes are currently identified by a physician, but this process is time-consuming and prone to errors. In this paper, we investigate the use of artificial intelligence-based methods to increase the accuracy and consistency of this process.MethodsWhole-body 18F-labelled fluoro-2-deoxyglucose ([18F]FDG) positron emission tomography/computed tomography ([18F]FDG-PET/CT) scans (Philips Gemini TF) from 134 patients were retrospectively analysed. The thorax was automatically located, and then slices were fed into a U-Net to identify candidate regions. These regions were split into overlapping 3D cubes, which were individually predicted as positive or negative using a 3D CNN. From these predictions, pathological mediastinal nodes could be identified. A second cohort of 71 patients was then acquired from a different, newer scanner (GE Discovery MI), and the performance of the model on this dataset was tested with and without transfer learning.ResultsOn the test set from the first scanner, our model achieved a sensitivity of 0.87 (95% confidence intervals [0.74, 0.94]) with 0.41 [0.22, 0.71] false positives/patient. This was comparable to the performance of an expert. Without transfer learning, on the test set from the second scanner, the corresponding results were 0.53 [0.35, 0.70] and 0.24 [0.10, 0.49], respectively. With transfer learning, these metrics were 0.88 [0.73, 0.97] and 0.69 [0.43, 1.04], respectively.ConclusionModel performance was comparable to that of an expert on data from the same scanner. With transfer learning, the model can be applied to data from a different scanner. To our knowledge it is the first study of its kind to go directly from whole-body [18F]FDG-PET/CT scans to pathological mediastinal lymph node localisation.
| S-EPMC8803782 | biostudies-literature
Project description:Blood glucose is routinely measured prior to 18F-fluorodeoxyglucose (FDG) administration in positron emission tomography (PET) imaging to identify hyperglycemia that may affect image quality. In this study we explore the effects of blood glucose levels upon semi-quantitative standardized uptake value (SUV) measurements of target organs and tissues of interest and in particular address the relationship of blood glucose to FDG accumulation in the brain and liver.436 FDG PET/CT consecutive studies performed for oncology staging in 229 patients (226 male) at the Ann Arbor Veterans Administration Healthcare System were reviewed. All patients had blood glucose measured (112.4±34.1mg/dL) prior to injection of 466.2±51.8MBq (12.6±1.4mCi) of FDG. SUV measurements of brain, aortic arch blood-pool, liver, and spleen were obtained at 64.5±10.2min' post-injection.We found a negative inverse relationship of brain SUV with increasing plasma glucose, levels for both absolute and normalized (either to blood-pool or liver) values. Higher blood glucose levels had a mild effect upon liver and blood-pool SUV. By contrast, spleen SUV was independent of blood glucose, but demonstrated the greatest variability (deviation on linear regression). In contrast to other tissues, liver and spleen SUV normalized to blood-pool SUV were not dependent upon blood glucose levels.The effects of hyperglycemia upon FDG uptake in brain and liver, over a range of blood glucose values generally considered acceptable for clinical PET imaging, may have measurable effects on semi-quantitative image analysis.
| S-EPMC5553544 | biostudies-other
Project description:FDG-PET-derived textural features describing intra-tumor heterogeneity are increasingly investigated as imaging biomarkers. As part of the process of quantifying heterogeneity, image intensities (SUVs) are typically resampled into a reduced number of discrete bins. We focused on the implications of the manner in which this discretization is implemented. Two methods were evaluated: (1) R(D), dividing the SUV range into D equally spaced bins, where the intensity resolution (i.e. bin size) varies per image; and (2) R(B), maintaining a constant intensity resolution B. Clinical feasibility was assessed on 35 lung cancer patients, imaged before and in the second week of radiotherapy. Forty-four textural features were determined for different D and B for both imaging time points. Feature values depended on the intensity resolution and out of both assessed methods, R(B) was shown to allow for a meaningful inter- and intra-patient comparison of feature values. Overall, patients ranked differently according to feature values–which was used as a surrogate for textural feature interpretation–between both discretization methods. Our study shows that the manner of SUV discretization has a crucial effect on the resulting textural features and the interpretation thereof, emphasizing the importance of standardized methodology in tumor texture analysis.
| S-EPMC4525145 | biostudies-literature
Project description:PURPOSE:To investigate the use and efficiency of 3-D deep learning, fully convolutional networks (DFCN) for simultaneous tumor cosegmentation on dual-modality nonsmall cell lung cancer (NSCLC) and positron emission tomography (PET)-computed tomography (CT) images. METHODS:We used DFCN cosegmentation for NSCLC tumors in PET-CT images, considering both the CT and PET information. The proposed DFCN-based cosegmentation method consists of two coupled three-dimensional (3D)-UNets with an encoder-decoder architecture, which can communicate with the other in order to share complementary information between PET and CT. The weighted average sensitivity and positive predictive values denoted as Scores, dice similarity coefficients (DSCs), and the average symmetric surface distances were used to assess the performance of the proposed approach on 60 pairs of PET/CTs. A Simultaneous Truth and Performance Level Estimation Algorithm (STAPLE) of 3 expert physicians' delineations were used as a reference. The proposed DFCN framework was compared to 3 graph-based cosegmentation methods. RESULTS:Strong agreement was observed when using the STAPLE references for the proposed DFCN cosegmentation on the PET-CT images. The average DSCs on CT and PET are 0.861 ± 0.037 and 0.828 ± 0.087, respectively, using DFCN, compared to 0.638 ± 0.165 and 0.643 ± 0.141, respectively, when using the graph-based cosegmentation method. The proposed DFCN cosegmentation using both PET and CT also outperforms the deep learning method using either PET or CT alone. CONCLUSIONS:The proposed DFCN cosegmentation is able to outperform existing graph-based segmentation methods. The proposed DFCN cosegmentation shows promise for further integration with quantitative multimodality imaging tools in clinical trials.
| S-EPMC6527327 | biostudies-literature
Project description:We present a composite procedure for the quantum-chemical computation of spin-spin-coupled 1 H?NMR spectra for general, flexible molecules in solution that is based on four main steps, namely conformer/rotamer ensemble (CRE) generation by the fast tight-binding method GFN-xTB and a newly developed search algorithm, computation of the relative free energies and NMR parameters, and solving the spin Hamiltonian. In this way the NMR-specific nuclear permutation problem is solved, and the correct spin symmetries are obtained. Energies, shielding constants, and spin-spin couplings are computed at state-of-the-art DFT levels with continuum solvation. A few (in)organic and transition-metal complexes are presented, and very good, unprecedented agreement between the theoretical and experimental spectra was achieved. The approach is routinely applicable to systems with up to 100-150 atoms and may open new avenues for the detailed (conformational) structure elucidation of, for example, natural products or drug molecules.
| S-EPMC5698732 | biostudies-literature