Human verifications: Computable with truth values outside logic.

Cognitive scientists treat verification as a computation in which descriptions that match the relevant situation are true, but otherwise false. The claim is controversial: The logician Gödel and the physicist Penrose have argued that human verifications are not computable. In contrast, the theory of mental models treats verification as computable, but the two truth values of standard logics, true and false, as insufficient. Three online experiments (n = 208) examined participants' verifications of disjunctive assertions about a location of an individual or a journey, such as: 'You arrived at Exeter or Perth'. The results showed that their verifications depended on observation of a match with one of the locations but also on the status of other locations (Experiment 1). Likewise, when they reached one destination and the alternative one was impossible, their use of the truth value: could be true and could be false increased (Experiment 2). And, when they reached one destination and the only alternative one was possible, they used the truth value, true and it couldn't have been false, and when the alternative one was impossible, they used the truth value: true but it could have been false (Experiment 3). These truth values and those for falsity embody counterfactuals. We implemented a computer program that constructs models of disjunctions, represents possible destinations, and verifies the disjunctions using the truth values in our experiments. Whether an awareness of a verification's outcome is computable remains an open question.

A deep learning model for the classification of atrial fibrillation in critically ill patients.

BACKGROUND: Atrial fibrillation (AF) is the most common cardiac arrhythmia in the intensive care unit and is associated with increased morbidity and mortality. New-onset atrial fibrillation (NOAF) is often initially paroxysmal and fleeting, making it difficult to diagnose, and therefore difficult to understand the true burden of disease. Automated algorithms to detect AF in the ICU have been advocated as a means to better quantify its true burden. RESULTS: We used a publicly available 12-lead ECG dataset to train a deep learning model for the classification of AF. We then conducted an external independent validation of the model using continuous telemetry data from 984 critically ill patients collected in our institutional database. Performance metrics were stratified by signal quality, classified as either clean or noisy. The deep learning model was able to classify AF with an overall sensitivity of 84%, specificity of 89%, positive predictive value (PPV) of 55%, and negative predictive value of 97%. Performance was improved in clean data as compared to noisy data, most notably with respect to PPV and specificity. CONCLUSIONS: This model demonstrates that computational detection of AF is currently feasible and effective. This approach stands to improve the efficiency of retrospective and prospective research into AF in the ICU by automating AF detection, and enabling precise quantification of overall AF burden.

Quantifying deep neural network uncertainty for atrial fibrillation detection with limited labels.

Atrial fibrillation (AF) is the most common arrhythmia found in the intensive care unit (ICU), and is associated with many adverse outcomes. Effective handling of AF and similar arrhythmias is a vital part of modern critical care, but obtaining knowledge about both disease burden and effective interventions often requires costly clinical trials. A wealth of continuous, high frequency physiological data such as the waveforms derived from electrocardiogram telemetry are promising sources for enriching clinical research. Automated detection using machine learning and in particular deep learning has been explored as a solution for processing these data. However, a lack of labels, increased presence of noise, and inability to assess the quality and trustworthiness of many machine learning model predictions pose challenges to interpretation. In this work, we propose an approach for training deep AF models on limited, noisy data and report uncertainty in their predictions. Using techniques from the fields of weakly supervised learning, we leverage a surrogate model trained on non-ICU data to create imperfect labels for a large ICU telemetry dataset. We combine these weak labels with techniques to estimate model uncertainty without the need for extensive human data annotation. AF detection models trained using this process demonstrated higher classification performance (0.64-0.67 F1 score) and improved calibration (0.05-0.07 expected calibration error).

How poetry evokes emotions.

Poetry evokes emotions. It does so, according to the theory we present, from three sorts of simulation. They each can prompt emotions, which are communications both within the brain and among people. First, models of a poem's semantic contents can evoke emotions as do models that occur in depictions of all kinds, from novels to perceptions. Second, mimetic simulations of prosodic cues, such as meter, rhythm, and rhyme, yield particular emotional states. Third, people's simulations of themselves enable them to know that they are engaged with a poem, and an aesthetic emotion can occur as a result. The three simulations predict certain sorts of emotion, e.g., prosodic cues can evoke basic emotions of happiness, sadness, anger, and anxiety. Empirical evidence corroborates the theory, which we relate to other accounts of poetic emotions.

The critical care data exchange format: a proposed flexible data standard for combining clinical and high-frequency physiologic data in critical care.

Objective.To develop a standardized format for exchanging clinical and physiologic data generated in the intensive care unit. Our goal was to develop a format that would accommodate the data collection pipelines of various sites but would not require dataset-specific schemas or ad-hoc tools for decoding and analysis.Approach.A number of centers had independently developed solutions for storing clinical and physiologic data using Hierarchical Data Format-Version 5 (HDF5), a well-supported standard already in use in multiple other fields. These individual solutions involved design choices that made the data difficult to share despite the underlying common framework. A collaborative process was used to form the basis of a proposed standard that would allow for interoperability and data sharing with common analysis tools.Main Results.We developed the HDF5-based critical care data exchange format which stores multiparametric data in an efficient, self-describing, hierarchical structure and supports real-time streaming and compression. In addition to cardiorespiratory and laboratory data, the format can, in future, accommodate other large datasets such as imaging and genomics. We demonstated the feasibility of a standardized format by converting data from three sites as well as the MIMIC III dataset.Significance.Individual approaches to the storage of multiparametric clinical data are proliferating, representing both a duplication of effort and a missed opportunity for collaboration. Adoption of a standardized format for clinical data exchange will enable the development of a digital biobank, facilitate the external validation of machine learning models and be a powerful tool for sharing multiparametric, high frequency patient level data in multisite clinical trials. Our proposed solution focuses on supporting standardized ontologies such as LOINC allowing easy reading of data regardless of the source and in so doing provides a useful method to integrate large amounts of existing data.

Generalizable deep temporal models for predicting episodes of sudden hypotension in critically ill patients: a personalized approach.

The vast quantities of data generated and collected in the Intensive Care Unit (ICU) have given rise to large retrospective datasets that are frequently used for observational studies. The temporal nature and fine granularity of much of the data collected in the ICU enable the pursuit of predictive modeling. In particular, forecasting acute hypotensive episodes (AHE) in intensive care patients has been of interest to researchers in critical care medicine. Given an advance warning of an AHE, care providers may be prompted to search for evolving disease processes and help mitigate negative clinical outcomes. However, the conventionally adopted definition of an AHE does not account for inter-patient variability and is restrictive. To reflect the wider trend of global clinical and research efforts in precision medicine, we introduce a patient-specific definition of AHE in this study and propose deep learning based models to predict this novel definition of AHE in data from multiple independent institutions. We provide extensive evaluation of the models by studying their accuracies in detecting patient-specific AHEs with lead-times ranging from 10 min to 1 hour before the onset of the event. The resulting models achieve AUROC values ranging from 0.57-0.87 depending on the lead time of the prediction. We demonstrate the generalizability and robustness of our approach through the use of independent multi-institutional data.

Toxoplasmosis chorioretinitis mimicking acute retinal necrosis associated with local corticosteroid.

BACKGROUND: The cases discussed highlight the atypical presentation and diagnostic dilemmas of toxoplasmosis with fulminant retinal necrosis and the potentially devastating visual outcomes of toxoplasma chorioretinitis following local corticosteroid exposure. CASE PRESENTATION: We report a series of three patients who presented with toxoplasmosis mimicking severe acute retinal necrosis. Patients were between 59 and 77 years old and had been exposed to local corticosteroids preceding our evaluation. All patients demonstrated diffuse retinal whitening with severe vision loss on presentation. Polymerase chain reaction testing (PCR) was diagnostic in two patients, and histopathologic examination of a vitrectomy specimen was diagnostic in one patient. All cases of retinitis resolved with anti-parasitic medication; however, visual acuity failed to improve in all patients due to disease severity and presentation. CONCLUSIONS: Local corticosteroid injection may trigger or exacerbate toxoplasmosis chorioretinitis, leading to fulminant retinal necrosis and severe vision loss. Toxoplasma chorioretinitis should be considered in the differential diagnosis of patients presenting with clinical features of acute retinal necrosis, particularly following local corticosteroid injection regardless of their baseline systemic immune status. Diagnostic vitrectomy may be helpful in patients in whom PCR testing is negative and ocular toxoplasmosis is suspected.

Prediction of Patient-specific Acute Hypotensive Episodes in ICU Using Deep Models.

Forecasting acute hypotensive episodes (AHE) in intensive care patients has been of recent interest to researchers in the healthcare domain. Advance warning of an impending AHE may give care providers additional information to help mitigate the negative clinical impact of a serious event such as an AHE or prompt a search for an evolving disease process. However, the currently accepted definition of AHE is restrictive does not account for inter-patient variability. In this paper, we propose a novel definition of an AHE based on patient-specific features of blood pressure recordings. Next, we utilize a deep learning-based method to predict the onset of an AHE from multiple physiological readings for different definitions of the prediction task including variable input and gap lengths. Using a cohort of 538 patients, our model was able to successfully predict the onset of an AHE with an accuracy and AUC score of 0.80 and 0.87 respectively. Compared to a baseline logistic regression model, our model outperforms the baseline in most of the definitions of the prediction task.

Predictive Modeling using Intensive Care Unit Data: Considerations for Data Pre-processing and Analysis.

The vast quantities of data generated and collected in the Intensive Care Unit (ICU) have given rise to large retrospective datasets that are frequently used for observational studies. The temporal nature and fine granularity of much of the data collected in the ICU enable the pursuit of predictive modeling, an increasingly common topic in ICU literature. Since patient conditions can rapidly change in the ICU, predicting the onset of events that are indicative of deteriorating patient state has potential clinical utility. However, the development of predictive modeling applications using ICU data requires a number of considerations to maximize prospective performance and clinical utility. In this study, we discuss the challenges encountered and considerations taken by using the prediction of acute hypotensive episodes as an example.

Facts and Possibilities: A Model-Based Theory of Sentential Reasoning.

This article presents a fundamental advance in the theory of mental models as an explanation of reasoning about facts, possibilities, and probabilities. It postulates that the meanings of compound assertions, such as conditionals (if) and disjunctions (or), unlike those in logic, refer to conjunctions of epistemic possibilities that hold in default of information to the contrary. Various factors such as general knowledge can modulate these interpretations. New information can always override sentential inferences; that is, reasoning in daily life is defeasible (or nonmonotonic). The theory is a dual process one: It distinguishes between intuitive inferences (based on system 1) and deliberative inferences (based on system 2). The article describes a computer implementation of the theory, including its two systems of reasoning, and it shows how the program simulates crucial predictions that evidence corroborates. It concludes with a discussion of how the theory contrasts with those based on logic or on probabilities.

On selecting evidence to test hypotheses: A theory of selection tasks.

How individuals choose evidence to test hypotheses is a long-standing puzzle. According to an algorithmic theory that we present, it is based on dual processes: individuals' intuitions depending on mental models of the hypothesis yield selections of evidence matching instances of the hypothesis, but their deliberations yield selections of potential counterexamples to the hypothesis. The results of 228 experiments using Wason's selection task corroborated the theory's predictions. Participants made dependent choices of items of evidence: the selections in 99 experiments were significantly more redundant (using Shannon's measure) than those of 10,000 simulations of each experiment based on independent selections. Participants tended to select evidence corresponding to instances of hypotheses, or to its counterexamples, or to both. Given certain contents, instructions, or framings of the task, they were more likely to select potential counterexamples to the hypothesis. When participants received feedback about their selections in the "repeated" selection task, they switched from selections of instances of the hypothesis to selection of potential counterexamples. These results eliminated most of the 15 alternative theories of selecting evidence. In a meta-analysis, the model theory yielded a better fit of the results of 228 experiments than the one remaining theory based on reasoning rather than meaning. We discuss the implications of the model theory for hypothesis testing and for a well-known paradox of confirmation. (PsycINFO Database Record

Formulation of a Peribulbar Block for Prolonged Postoperative Pain Management in Vitreoretinal Surgery: A Randomized Clinical Trial.

PURPOSE: To evaluate postoperative pain level using a supplemental peribulbar injection at the conclusion of retinal surgery. DESIGN: Prospective, parallel-assigned, single-masked, randomized clinical trial. PARTICIPANTS: Fifty-eight patients undergoing scleral buckle, vitrectomy, or combined surgery. METHODS: In a single academic institutional practice, 58 patients undergoing scleral buckle, vitrectomy, or combined surgery were enrolled. Exclusion criteria included those with a risk for glaucoma, a pre-existing chronic pain disorder, among others. Patients were assigned randomly to receive a postoperative peribulbar formulation of either bupivacaine, triamcinolone acetonide, and cefazolin (group A) or bupivacaine, balanced salt solution, and cefazolin (group B). The postoperative pain score and ocular motility were assessed by a masked observer on the first postoperative day. MAIN OUTCOME MEASURES: The primary outcome measure was the postoperative pain score. Secondary outcome measures included oral analgesic use, ocular motility, and intraocular pressure (IOP). RESULTS: The mean pain scores were 2.8±2.9 for group A and 3.8±2.6 for group B (P = 0.095). Pain was absent in 28% of group A patients versus 14% of group B patients (P = 0.11). Group A required less narcotic pain medication (hydroxycodone: group A, 0.7±3 mg vs. group B, 3±6 mg; P = 0.05; oxycodone: group A, 7±7 mg vs. 9±13 mg; P = 0.2) than group B. Motility was full in group B and limited in group A (P ≤ 0.001), with no differences in mean IOP measurements at any point after surgery. CONCLUSIONS: We did not demonstrate a statistically significant reduction in mean postoperative pain scores. However, patients in group A required less hydroxycodone use and had greater akinesia, suggesting prolonged neural blockade.

A Priori True and False Conditionals.

The theory of mental models postulates that meaning and knowledge can modulate the interpretation of conditionals. The theory's computer implementation implied that certain conditionals should be true or false without the need for evidence. Three experiments corroborated this prediction. In Experiment 1, nearly 500 participants evaluated 24 conditionals as true or false, and they justified their judgments by completing sentences of the form, It is impossible that A and ___ appropriately. In Experiment 2, participants evaluated 16 conditionals and provided their own justifications, which tended to be explanations rather than logical justifications. In Experiment 3, the participants also evaluated as possible or impossible each of the four cases in the partitions of 16 conditionals: A and C, A and not-C, not-A and C, not-A and not-C. These evaluations corroborated the model theory. We consider the implications of these results for theories of reasoning based on logic, probabilistic logic, and suppositions.

Vogt-Koyanagi-Harada-like syndrome after CTLA-4 inhibition with ipilimumab for metastatic melanoma.

Cytotoxic T-lymphocyte-associated antigen is a naturally occurring inhibitor of T-cell costimulation. Monoclonal antibody inhibition of cytotoxic T-lymphocyte-associated antigen with ipilimumab blocks this negative regulator of costimulation, promoting T-cell activation and survival, and leads to melanoma regression. Findings of the Vogt-Koyanagi-Harada (VKH) syndrome, an uveomeningitic syndrome that features neurological, auditory, ophthalmologic, and cutaneous involvement because of autoimmune targeting of melanocytic antigen, have rarely been described in association with melanoma immunotherapy. We describe a case of VKH-like syndrome in a 45-year-old HLA-A02-positive patient with metastatic melanoma treated with ipilimumab. Disruption of immune tolerance by ipilimumab led to melanoma remission while also inciting systemic and ophthalmic autoimmunity toward melanocytic antigen. These observations provide insight into the pathophysiology of the VKH syndrome, and the balance between tumor-associated tolerance and autoimmunity.

Toxoplasma chorioretinitis diagnosed by histopathology in a patient with AIDS.

PURPOSE: To demonstrate the diagnostic difficulties in cases of retinal necrosis in immunocompromised patients including the potential for false-negative anterior segment sampling and also to emphasize the utility of diagnostic vitrectomy with histopathologic examination. METHODS: This patient's chart was thoroughly reviewed to present salient features that are relevant to any ophthalmologist attempting to diagnose and treat chorioretinitis. A 38-year-old man with HIV/AIDS who presented with bilateral retinal necrosis. Thorough workup, including multiple samples of anterior chamber fluid for polymerase chain reaction, was negative. RESULTS: Diagnostic vitrectomy revealed a toxoplasma cyst. Triple therapy stabilized retinitis, although vision did not improve. CONCLUSION: This case reminds the clinician to consider a broad differential diagnosis for retinal necrosis in immunocompromised hosts and, when serologic and anterior chamber samples are negative, to consider diagnostic vitrectomy for polymerase chain reaction and histopathologic examination.

Naive Probability: Model-Based Estimates of Unique Events.

We describe a dual-process theory of how individuals estimate the probabilities of unique events, such as Hillary Clinton becoming U.S. President. It postulates that uncertainty is a guide to improbability. In its computer implementation, an intuitive system 1 simulates evidence in mental models and forms analog non-numerical representations of the magnitude of degrees of belief. This system has minimal computational power and combines evidence using a small repertoire of primitive operations. It resolves the uncertainty of divergent evidence for single events, for conjunctions of events, and for inclusive disjunctions of events, by taking a primitive average of non-numerical probabilities. It computes conditional probabilities in a tractable way, treating the given event as evidence that may be relevant to the probability of the dependent event. A deliberative system 2 maps the resulting representations into numerical probabilities. With access to working memory, it carries out arithmetical operations in combining numerical estimates. Experiments corroborated the theory's predictions. Participants concurred in estimates of real possibilities. They violated the complete joint probability distribution in the predicted ways, when they made estimates about conjunctions: P(A), P(B), P(A and B), disjunctions: P(A), P(B), P(A or B or both), and conditional probabilities P(A), P(B), P(B|A). They were faster to estimate the probabilities of compound propositions when they had already estimated the probabilities of each of their components. We discuss the implications of these results for theories of probabilistic reasoning.

Causal reasoning with mental models.

This paper outlines the model-based theory of causal reasoning. It postulates that the core meanings of causal assertions are deterministic and refer to temporally-ordered sets of possibilities: A causes B to occur means that given A, B occurs, whereas A enables B to occur means that given A, it is possible for B to occur. The paper shows how mental models represent such assertions, and how these models underlie deductive, inductive, and abductive reasoning yielding explanations. It reviews evidence both to corroborate the theory and to account for phenomena sometimes taken to be incompatible with it. Finally, it reviews neuroscience evidence indicating that mental models for causal inference are implemented within lateral prefrontal cortex.

Kinematic mental simulations in abduction and deduction.

We present a theory, and its computer implementation, of how mental simulations underlie the abductions of informal algorithms and deductions from these algorithms. Three experiments tested the theory's predictions, using an environment of a single railway track and a siding. This environment is akin to a universal Turing machine, but it is simple enough for nonprogrammers to use. Participants solved problems that required use of the siding to rearrange the order of cars in a train (experiment 1). Participants abduced and described in their own words algorithms that solved such problems for trains of any length, and, as the use of simulation predicts, they favored "while-loops" over "for-loops" in their descriptions (experiment 2). Given descriptions of loops of procedures, participants deduced the consequences for given trains of six cars, doing so without access to the railway environment (experiment 3). As the theory predicts, difficulty in rearranging trains depends on the numbers of moves and cars to be moved, whereas in formulating an algorithm and deducing its consequences, it depends on the Kolmogorov complexity of the algorithm. Overall, the results corroborated the use of a kinematic mental model in creating and testing informal algorithms and showed that individuals differ reliably in the ability to carry out these tasks.