There are currently no data published regarding the proportion of nuclear medicine centers using SPECT or SPECT/CT rather than planar ventilation/perfusion (V/Q) imaging in patients with suspected acute pulmonary embolism (PE). Furthermore, the reporting criteria used for interpretation of both planar and SPECT V/Q scans are variable and data are lacking regarding which criteria are commonly used in various centers. The aim of this study was to assess current practices regarding the performance and interpretation of lung scintigraphy across 3 different countries. Methods: A short online survey composed of simple multiple-choice questions was distributed to nuclear medicine departments in Australia, Canada, and France during the period April to December 2014. The survey covered image acquisition, interpretation criteria for SPECT and planar images, and use of pseudoplanar images and radiopharmaceuticals. Information was initially solicited by 2 sets of e-mails, which pointed to the survey internet link. Departments were subsequently contacted by telephone. A single response per department was consolidated. Results: Three hundred thirty-one responses were collected (Australia, 74; Canada, 48; and France, 209). Twenty-eight percent of centers indicated use of V/Q planar imaging alone whereas 72% of centers included some form of SPECT in their acquisition protocol for evaluation of PE, specifically V/Q SPECT in 36%, V/Q SPECT/CT in 29%, Q SPECT/CT in 2%, and both V/Q planar and SPECT in 5%, with a strong variability among countries. The most commonly used criteria for SPECT interpretation were the those of the European Association of Nuclear Medicine (60%). Criteria used for planar interpretation were heterogeneous (European Association of Nuclear Medicine criteria, 35%; Prospective Investigation of Pulmonary Embolism Diagnosis study, 29%; no standardized criteria, 21%). Sixty-three percent of departments used pseudoplanar images in addition to SPECT images. Conclusion: In the 3 countries surveyed, SPECT has largely replaced planar imaging for evaluation of PE, with almost half of the SPECT studies incorporating a CT acquisition. Criteria used for interpretation are inconsistent, especially for planar imaging. It has been more than 50 y since lung scintigraphy was first described for the diagnosis of pulmonary embolism (PE). Since that time, the examination has evolved greatly with respect to equipment, radiopharmaceuticals, and imaging algorithms, along with perception among nuclear medicine and referring physicians.Although performed 25 y ago, using imaging equipment and ventilation agents that would currently be considered obsolete in most centers, the Prospective Investigation of Pulmonary Embolism Diagnosis study (PIOPED) study remains the landmark accuracy study in the eyes of many clinicians (1). The diagnostic performance of planar ventilation/perfusion (V/Q) scanning was then insufficient to allow a binary reporting approach (PE or no PE). Therefore, probabilistic reporting criteria were proposed, resulting...
Playing text-based games requires skill in processing natural language and in planning. Although a key goal for agents solving this task is to generalize across multiple games, most previous work has either focused on solving a single game or has tackled generalization with rule-based heuristics. In this work, we investigate how structured information in the form of a knowledge graph (KG) can facilitate effective planning and generalization. We introduce a novel transformer-based sequence-to-sequence model that constructs a "belief" KG from raw text observations of the environment, dynamically updating this belief graph at every game step as it receives new observations. To train this model to build useful graph representations, we introduce and analyze a set of graph-related pre-training tasks. We demonstrate empirically that KG-based representations from our model help agents to converge faster to better policies for multiple text-based games, and further, enable stronger zero-shot performance on unseen games. Experiments on unseen games show that our best agent outperforms text-based baselines by 21.6%.
The labelling of training examples is a costly task in a supervised classification. Active learning strategies answer this problem by selecting the most useful unlabelled examples to train a predictive model. The choice of examples to label can be seen as a dilemma between the exploration and the exploitation over the data space representation. In this paper, a novel active learning strategy manages this compromise by modelling the active learning problem as a contextual bandit problem. We propose a sequential algorithm named Active Thompson Sampling (ATS), which, in each round, assigns a sampling distribution on the pool, samples one point from this distribution, and queries the oracle for this sample point label. Experimental comparison to previously proposed active learning algorithms show superior performance on a real application dataset.
The aim of the study was to evaluate the interest of quantitative bone SPECT-CT in the preoperative assessment of knee osteoarthritis (OA) before unicompartmental knee arthroplasty (UKA).Patients eligible for UKA were prospectively included in 2 centers and underwent a preoperative SPECT-CT. Images were reconstructed with an OSEM, an OSCGM (allowing SUV quantification) and an enhanced OSCGM (containing uptakes to bones) algorithms. Visual analysis and quantification (SUVmax) were performed for each compartment (medial compartment [MC], lateral compartment [LC], and patellofemoral compartment [PFC]). Clinical data were preoperatively assessed. The gold standard was the per-operative OA staging (International Cartilage Repair Society [ICRS] scale). Spearman's correlation coefficient was used for correlations. Sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), and accuracy of SPECT-CT were assessed.One hundred three patients (50 women, 53 men, mean age = 64.5 ± 10.3 y/o, 120 preoperative knees) were analyzed. There was no correlation between SUVmax and clinical data. There was a correlation between ICRS staging and SUVmax with both OSCGM (MC [rs = 0.25], LC [rs = 0.51], and PFC [rs = 0.27]), and enhanced OSCGM, except in the PFC (MC [rs = 0.22], LC [rs = 0.62], and PFC [rs = 0.03]). The Se, Sp, PPV, NPV, and accuracy of SPECT-CT were, respectively, 0.99, 0.67, 0.98, 0.80, 0.97 for the MC; 0.50, 0.85, 0.42, 0.89, 0.79 for the LC; and 0.23, 0.86, 0.50, 0.64, 0.62 for the PFC.Bone SPECT-CT SUVmax is correlated with per-operative OA staging. Despite the low sensitivity of SPECT-CT in the LC, its high specificity in the LC should prompt the surgeon to be vigilant before UKA surgery.
BackgroundSPECT/CT has been shown to increase the diagnostic performance of bone scintigraphy for staging of malignancies. A systematic double-bed SPECT/CT of the trunk may allow further improvement. However, this would be balanced by higher dosimetry and longer acquisition time. The objective was to assess the incremental diagnostic utility of a systematic double-bed SPECT/CT acquisition for bone scintigraphy in initial staging of cancer patients, especially compared with the usual approach consisting in a whole body planar scan (WBS) plus one single-bed targeted SPECT/CT.MethodsOne hundred two consecutive patients referred for bone scintigraphy for initial staging of malignancy were analyzed. All patients underwent a double-bed SPECT/CT acquisition of the trunk. Images were interpreted by two nuclear medicine physicians in a 3-step procedure. Firstly, only WBS planar images were used; secondly, one additional single-bed SPECT/CT chosen based on planar images was used; finally, WBS planar and double-bed SPECT/CT images were interpreted. Lesions were classified as benign, equivocal or suspicious for metastasis. A per-lesion, per-anatomical region and per-patient analysis was performed.ResultsIn a per-lesion analysis, the number of equivocal and suspicious lesions was 91 and 241 using WBS planar images, 17 and 259 using a single-bed SPECT/CT acquisition and 11 and 269 using double-bed SPECT/CT images, respectively. In a per-patient analysis, the diagnostic conclusion was negative, equivocal or suspicious for malignancy in 35, 53 and 14 patients using WB planar images, 77, 6 and 19 patients using an additional single-bed SPECT/CT and 76, 7 and 19 using double-bed SPECT/CT images, respectively.Seventeen lesions unseen on WBS images were interpreted as suspicious (n = 12) or equivocal (n = 5) on double-bed SPECT/CT images. Six lesions unseen on “WBS + targeted single-bed SPECT/CT” were interpreted as suspicious on double-bed SPECT/CT, with no shift in the metastatic status of patients.ConclusionA systematic double-bed SPECT/CT acquisition has a limited incremental diagnostic value over an oriented single-bed SPECT/CT in terms of specificity and conclusiveness of bone scintigraphy in the initial staging of cancer patients. However, it slightly improved the sensitivity of the test by detecting unseen lesions on WBS, which may be of value for initial staging of cancer.
In this paper, an original framework to model human-machine spoken dialogues is proposed to deal with co-adaptation between users and Spoken Dialogue Systems in non-cooperative tasks. The conversation is modeled as a Stochastic Game: both the user and the system have their own preferences but have to come up with an agreement to solve a non-cooperative task. They are jointly trained so the Dialogue Manager learns the optimal strategy against the best possible user. Results obtained by simulation show that non-trivial strategies are learned and that this framework is suitable for dialogue modeling.
No abstract
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.