Neural superstatistics for Bayesian estimation of dynamic cognitive models.

Mathematical models of cognition are often memoryless and ignore potential fluctuations of their parameters. However, human cognition is inherently dynamic. Thus, we propose to augment mechanistic cognitive models with a temporal dimension and estimate the resulting dynamics from a superstatistics perspective. Such a model entails a hierarchy between a low-level observation model and a high-level transition model. The observation model describes the local behavior of a system, and the transition model specifies how the parameters of the observation model evolve over time. To overcome the estimation challenges resulting from the complexity of superstatistical models, we develop and validate a simulation-based deep learning method for Bayesian inference, which can recover both time-varying and time-invariant parameters. We first benchmark our method against two existing frameworks capable of estimating time-varying parameters. We then apply our method to fit a dynamic version of the diffusion decision model to long time series of human response times data. Our results show that the deep learning approach is very efficient in capturing the temporal dynamics of the model. Furthermore, we show that the erroneous assumption of static or homogeneous parameters will hide important temporal information.

Duration discrimination: A diffusion decision modeling approach.

The human ability to discriminate the duration of two subsequently presented stimuli is often studied with tasks that involve a comparison between a standard stimulus (with fixed duration) and comparison stimuli (with varying durations). The performance in such tasks is influenced by the presentation order of these successively presented stimuli. The so-called Type A effect refers to the impact of presentation order on the point of subjective equality. The Type B effect describes effects of presentation order on the just-noticeable-difference. Cognitive models that account for these context effects assume that participants' duration estimation is influenced by the history of previously encountered stimuli. For example, the internal reference model assumes that the magnitude of a "typical" stimulus is represented by an internal reference. This internal reference evolves throughout an experiment and is updated on every trial. Different recent models have in common that they describe how the internal reference is computed but are agnostic to the decision process itself. In this study, we develop a new model that incorporates the mechanisms of perceptual discrimination models into a diffusion model. The diffusion model focuses on the dynamics of the decision process itself and accounts for choice and response times based on a set of latent cognitive variables. We show that our model accurately predicts the accuracy and response time distribution in a classical duration discrimination task. Further, model parameters were sensitive to the Type A and B effect. The proposed model opens up new opportunities for studying human discrimination performance (e.g., individual differences).

Incidence and outcomes of unstable angina compared with non-ST-elevation myocardial infarction.

OBJECTIVE: Assess the relative incidence and compare characteristics and outcome of unstable angina (UA) and non-ST-elevation myocardial infarction (NSTEMI). DESIGN: Two independent prospective multicentre diagnostic studies (Advantageous Predictors of Acute Coronary Syndromes Evaluation [APACE] and High-Sensitivity Troponin in the Evaluation of Patients With Acute Coronary Syndrome [High-STEACS]) enrolling patients with acute chest discomfort presenting to the emergency department. Central adjudication of the final diagnosis was done by two independent cardiologists using all clinical information including serial measurements of high-sensitivity cardiac troponin (hs-cTn). All-cause death and future non-fatal MI were assessed at 30 days and 1 year. RESULTS: 8992 patients were enrolled at 11 centres. UA was adjudicated in 8.9%(95% CI 8.0 to 9.7) and 2.8% (95% CI 2.3 to 3.3) patients in APACE and High-STEACS, respectively, and NSTEMI in 15.1% (95% CI 14.0 to 16.2) and 13.4% (95% CI 12.4 to 14.3). Coronary artery disease was pre-existing in 73% and 76% of patients with UA. At 30 days, all-cause mortality in UA was substantially lower as compared with NSTEMI (0.5% vs 3.7%, p=0.002 in APACE, 0.7% vs 7.4%, p=0.004 in High-STEACS). Similarly, at 1 year in UA all-cause mortality was 3.3% (95% CI 1.2 to 5.3) vs 10.4% (95% CI 7.9 to 12.9) in APACE, and 5.1% (95% CI 0.7 to 9.5) vs 22.9% (95% CI 19.3 to 26.4) in High-STEACS, and similar to non-cardiac chest pain (NCCP). In contrast, future non-fatal MI in APACE was comparable in UA and NSTEMI (11.2%, 95% CI 7.8 to 14.6 and 7.9%, 95% CI 5.7 to 10.2), and higher than in NCCP (0.6%, 95% CI 0.2 to 1.0). CONCLUSIONS: The relative incidence and mortality of UA is substantially lower than that of NSTEMI, while the rate of future non-fatal MI is similar.

Diagnostic and prognostic value of QRS duration and QTc interval in patients with suspected myocardial infarction.

BACKGROUND: While prolongation of QRS duration and QTc interval during acute myocardial infarction (AMI) has been reported in animals, limited data is available for these readily available electrocardiography (ECG) markers in humans. METHODS: Diagnostic and prognostic value of QRS duration and QTc interval in patients with suspected AMI in a prospective diagnostic multicentre study were prospectively assessed. Digital 12-lead ECGs were recorded at presentation. QRS duration and QTc interval were automatically calculated in a blinded fashion. Final diagnosis was adjudicated by two independent cardiologists. The prognostic endpoint was all-cause mortality during 24 months of follow-up. RESULTS: Among 4042 patients, AMI was the final diagnosis in 19% of patients. Median QRS duration and median QTc interval were significantly greater in patients with AMI compared to those with other final diagnoses (98 ms [IQR 88-108] vs. 94 ms [IQR 86-102] and 436 ms [IQR 414-462] vs. 425 ms [IQR 407-445], p < 0.001 for both comparisons). The diagnostic value of both ECG signatures however was only modest (AUC 0.56 and 0.60). Cumulative mortality rates after 2 years were 15.9% vs. 5.6% in patients with a QRS > 120 ms compared to a QRS duration ≤ 120 ms (p < 0.001), and 11.4% vs. 4.3% in patients with a QTc > 440 ms compared to a QRS duration ≤ 440 ms (p < 0.001). After adjustment for age and important ECG and clinical parameters, the QTc interval but not QRS duration remained an independent predictor of mortality. CONCLUSIONS: Prolongation of QRS duration > 120 ms and QTc interval > 440 ms predict mortality in patients with suspected AMI, but do not add diagnostic value.

Combining High-Sensitivity Cardiac Troponin I and Cardiac Troponin T in the Early Diagnosis of Acute Myocardial Infarction.

BACKGROUND: Combining 2 signals of cardiomyocyte injury, cardiac troponin I (cTnI) and T (cTnT), might overcome some individual pathophysiological and analytical limitations and thereby increase diagnostic accuracy for acute myocardial infarction with a single blood draw. We aimed to evaluate the diagnostic performance of combinations of high-sensitivity (hs) cTnI and hs-cTnT for the early diagnosis of acute myocardial infarction. METHODS: The diagnostic performance of combining hs-cTnI (Architect, Abbott) and hs-cTnT (Elecsys, Roche) concentrations (sum, product, ratio, and a combination algorithm) obtained at the time of presentation was evaluated in a large multicenter diagnostic study of patients with suspected acute myocardial infarction. The optimal rule-out and rule-in thresholds were externally validated in a second large multicenter diagnostic study. The proportion of patients eligible for early rule-out was compared with the European Society of Cardiology 0/1 and 0/3 hour algorithms. RESULTS: Combining hs-cTnI and hs-cTnT concentrations did not consistently increase overall diagnostic accuracy as compared with the individual isoforms. However, the combination improved the proportion of patients meeting criteria for very early rule-out. With the European Society of Cardiology 2015 guideline recommended algorithms and cut-offs, the proportion meeting rule-out criteria after the baseline blood sampling was limited (6% to 24%) and assay dependent. Application of optimized cut-off values using the sum (9 ng/L) and product (18 ng2/L2) of hs-cTnI and hs-cTnT concentrations led to an increase in the proportion ruled-out after a single blood draw to 34% to 41% in the original (sum: negative predictive value [NPV] 100% [95% confidence interval (CI), 99.5% to 100%]; product: NPV 100% [95% CI, 99.5% to 100%]) and in the validation cohort (sum: NPV 99.6% [95% CI, 99.0-99.9%]; product: NPV 99.4% [95% CI, 98.8-99.8%]). The use of a combination algorithm (hs-cTnI <4 ng/L and hs-cTnT <9 ng/L) showed comparable results for rule-out (40% to 43% ruled out; NPV original cohort 99.9% [95% CI, 99.2-100%]; NPV validation cohort 99.5% [95% CI, 98.9-99.8%]) and rule-in (positive predictive value [PPV] original cohort 74.4% [95% Cl, 69.6-78.8%]; PPV validation cohort 84.0% [95% Cl, 79.7-87.6%]). CONCLUSIONS: New strategies combining hs-cTnI and hs-cTnT concentrations may significantly increase the number of patients eligible for very early and safe rule-out, but do not seem helpful for the rule-in of acute myocardial infarction. CLINICAL TRIAL REGISTRATION: URL (APACE): https://www.clinicaltrial.gov . Unique identifier: NCT00470587. URL (ADAPT): www.anzctr.org.au . Unique identifier: ACTRN12611001069943.

Diagnostic value of the cardiac electrical biomarker, a novel ECG marker indicating myocardial injury, in patients with symptoms suggestive of non-ST-elevation myocardial infarction.

BACKGROUND: The cardiac electrical biomarker (CEB) is a novel electrocardiographic (ECG) marker quantifying the dipolar activity of the heart with higher levels indicating myocardial injury. METHODS: We prospectively enrolled 1097 patients presenting with suspected non-ST-elevation myocardial infarction (NSTEMI) to the emergency department (ED). Digital 12-lead ECGs were recorded at presentation and the CEB values were calculated in a blinded fashion. The final diagnosis was adjudicated by two independent cardiologists. The prognostic endpoint was all-cause mortality during 2 years of follow-up. RESULTS: NSTEMI was the final diagnosis in 14% of patients. CEB levels were higher in patients with NSTEMI compared to other causes of chest pain (median 44 (IQR 21-98) vs. 30 (IQR 16-61), p < .001). A weak but significant correlation between levels of high-sensitivity cardiac troponin T (hs-cTnT) at admission to the ED and the CEB was found (r = .23, p < .001). The use of the CEB in addition to conventional ECG criteria improved the diagnostic accuracy for the diagnosis of NSTEMI as quantified by the area under the receiver operating characteristics curve from 0.66 to 0.71 (p < .001) and the sensitivity improved from 43% to 79% (p < .001). CONCLUSION: In conclusion, the CEB, an ECG marker of myocardial injury, significantly improves the accuracy and sensitivity of the ECG for the diagnosis of NSTEMI.

Direct Comparison of 2 Rule-Out Strategies for Acute Myocardial Infarction: 2-h Accelerated Diagnostic Protocol vs 2-h Algorithm.

BACKGROUND: We compared 2 high-sensitivity cardiac troponin (hs-cTn)-based 2-h strategies in patients presenting with suspected acute myocardial infarction (AMI) to the emergency department (ED): the 2-h accelerated diagnostic protocol (2h-ADP) combining hs-cTn, electrocardiogram, and a risk score, and the 2-h algorithm exclusively based on hs-cTn concentrations and their absolute changes. METHODS: Analyses were performed in 2 independent diagnostic cohorts [European Advantageous Predictors of Acute Coronary Syndrome Evaluation (APACE) study, Australian-New Zealand 2-h Accelerated Diagnostic Protocol to Assess patients with chest Pain symptoms using contemporary Troponins as the only biomarker (ADAPT) study] employing hs-cTnT (Elecsys) and hs-cTnI (Architect). The final diagnosis was adjudicated by 2 independent cardiologists. RESULTS: AMI was the final diagnosis in 16.5% (95% CI, 14.6%-18.6%) of the 1372 patients in APACE, and 12.6% (95% CI, 10.7%-14.7%) of 1153 patients in ADAPT. The negative predictive value (NPV) and sensitivity for AMI were very high and comparable with both strategies using either hs-cTnT or hs-cTnI in both cohorts (all statistical comparisons nonsignificant). The percentage of patients triaged toward rule-out was significantly lower with the 2h-ADP (36%-43%) vs the 2-h algorithm (55%-68%) with both assays and in both cohorts (P < 0.001). The sensitivity of the 2h-ADP was higher for 30-day major adverse cardiovascular events. CONCLUSIONS: Both algorithms provided very high and comparable safety as quantified by the NPV and sensitivity for AMI and major adverse cardiac events (MACE) at 30 days in patients triaged toward rule-out, although sensitivity for MACE at 30 days was lower with both algorithms in cohort 2. Although the 2-h algorithm was more efficacious, not all patients ruled out for AMI by this algorithm were appropriate candidates for early discharge. The 2h-ADP seems superior in the selection of patients for early discharge from the ED. CLINICAL TRIAL REGISTRATION: APACE: http://clinicaltrials.gov/show/NCT00470587ADAPT: Australia-New Zealand Clinical Trials Registry ACTRN12611001069943.

Early diagnosis of acute myocardial infarction in patients with mild elevations of cardiac troponin.

BACKGROUND: The early diagnosis of acute myocardial infarction (AMI) in patients with mild elevations of high-sensitivity cardiac troponin (hs-cTn) is a challenge. It is unclear whether copeptin, a marker of endogenous stress, or 1h-hs-cTn changes are better suited to address this important unmet clinical need. METHODS: We prospectively enrolled patients presenting with symptoms suggestive of AMI to the emergency department (ED). Two independent cardiologists adjudicated the final diagnosis. Mild hs-cTn elevations were defined as 26.2 ng/L (99th percentile) to 75 ng/L for hs-cTnI, and 14 ng/L (99th percentile) to 50 ng/L (biological-equivalent to 75 ng/L for hs-cTnI) for hs-cTnT. RESULTS: Among 1356 patients, 80 (6%) had mild hs-cTnI elevations at presentation. Within this group, AMI was the final diagnosis in 39 patients (49%). The diagnostic accuracy for the diagnosis of AMI as quantified by the area under the receiver operating characteristic curve (AUC) was 0.51 (95% CI 0.39-0.64) for hs-cTnI at presentation, 0.58 (95% CI 0.45-0.71) for copeptin at presentation, and 0.78 (95% CI 0.68-0.88) for 1h-hs-cTnI changes, which was significantly higher as compared to copeptin (p = 0.02) or hs-cTnI alone (p < 0.001). The additional use of 1h-hs-cTnI changes, but not of copeptin, improved diagnostic accuracy of hs-cTnI at presentation (AUC 0.80, 95% CI 0.70-0.90; p = 0.002 for comparison). Similar findings regarding copeptin and 1h-hs-cTnT/I changes were obtained for mild hs-cTnT elevations. CONCLUSIONS: About 6-22% of patients presenting with suggestive AMI to the ED have mild hs-cTnT/I elevations at presentation. In contrast to copeptin, the addition of 1h-hs-cTn changes substantially improves the early diagnosis of AMI.

Effects of vascular photodynamic therapy in a newly adapted experimental rat aortic aneurysm model.

The hypothesis driving this study was that photodynamic therapy (PDT) may limit abdominal aortic aneurysm growth due to matrix changes. The aortas of 12 rats were incubated with elastase using a newly modified experimental aneurysm model (3.5 mg/ml). Rats were allocated to an elastase-only group (n = 6) to study the elastase-induced aneurysm growth and an elastase ± PDT group to evaluate if PDT limited aneurysm growth (n = 6). PDT was performed with the photosensitizer methylene blue, and thermoneutral laser light (660 nm) was applied (120 J/cm(2), 100 mW/cm(2)) using a diode laser. Four untreated rats served as controls. The arteries were analysed after 4 weeks based on histology, immunohistochemistry and morphometry. This modified rat elastase model led to reproducible aneurysm development with no elastase-induced mortality compared with control animals (circumference, controls: 2.9 ± 0.2 vs. elastase: 5.5 ± 0.9 mm; P < 0.01). PDT after elastase incubation did not inhibit inflammatory cell infiltration. No significant change in the circumference was observed between elastase incubation and PDT treatment after elastase incubation (circumference, elastase: 5.5 ± 0.9 vs. elastase and PDT: 6.1 ± 0.8 mm; P < 0.01). Despite a PDT-induced resistance to protease digestion, PDT did not reduce aortic dilatation in the elastase-treated rat aorta. These findings suggest that PDT may not be a useful modality to prevent aneurysm growth.