Identifying regulation with adversarial surrogates.

Homeostasis, the ability to maintain a relatively constant internal environment in the face of perturbations, is a hallmark of biological systems. It is believed that this constancy is achieved through multiple internal regulation and control processes. Given observations of a system, or even a detailed model of one, it is both valuable and extremely challenging to extract the control objectives of the homeostatic mechanisms. In this work, we develop a robust data-driven method to identify these objectives, namely to understand: "what does the system care about?". We propose an algorithm, Identifying Regulation with Adversarial Surrogates (IRAS), that receives an array of temporal measurements of the system and outputs a candidate for the control objective, expressed as a combination of observed variables. IRAS is an iterative algorithm consisting of two competing players. The first player, realized by an artificial deep neural network, aims to minimize a measure of invariance we refer to as the coefficient of regulation. The second player aims to render the task of the first player more difficult by forcing it to extract information about the temporal structure of the data, which is absent from similar "surrogate" data. We test the algorithm on four synthetic and one natural data set, demonstrating excellent empirical results. Interestingly, our approach can also be used to extract conserved quantities, e.g., energy and momentum, in purely physical systems, as we demonstrate empirically.

iCVS-Inferring Cardio-Vascular hidden States from physiological signals available at the bedside.

Intensive care medicine is complex and resource-demanding. A critical and common challenge lies in inferring the underlying physiological state of a patient from partially observed data. Specifically for the cardiovascular system, clinicians use observables such as heart rate, arterial and venous blood pressures, as well as findings from the physical examination and ancillary tests to formulate a mental model and estimate hidden variables such as cardiac output, vascular resistance, filling pressures and volumes, and autonomic tone. Then, they use this mental model to derive the causes for instability and choose appropriate interventions. Not only this is a very hard problem due to the nature of the signals, but it also requires expertise and a clinician's ongoing presence at the bedside. Clinical decision support tools based on mechanistic dynamical models offer an appealing solution due to their inherent explainability, corollaries to the clinical mental process, and predictive power. With a translational motivation in mind, we developed iCVS: a simple, with high explanatory power, dynamical mechanistic model to infer hidden cardiovascular states. Full model estimation requires no prior assumptions on physiological parameters except age and weight, and the only inputs are arterial and venous pressure waveforms. iCVS also considers autonomic and non-autonomic modulations. To gain more information without increasing model complexity, both slow and fast timescales of the blood pressure traces are exploited, while the main inference and dynamic evolution are at the longer, clinically relevant, timescale of minutes. iCVS is designed to allow bedside deployment at pediatric and adult intensive care units and for retrospective investigation of cardiovascular mechanisms underlying instability. In this paper, we describe iCVS and inference system in detail, and using a dataset of critically-ill children, we provide initial indications to its ability to identify bleeding, distributive states, and cardiac dysfunction, in isolation and in combination.

Development and validation of a multivariable prediction model in pediatric liver transplant patients for predicting intensive care unit length of stay.

BACKGROUND: Liver transplantation is the life-saving treatment for many end-stage pediatric liver diseases. The perioperative course, including surgical and anesthetic factors, have an important influence on the trajectory of this high-risk population. Given the complexity and variability of the immediate postoperative course, there would be utility in identifying risk factors that allow prediction of adverse outcomes and intensive care unit trajectories. AIMS: The aim of this study was to develop and validate a risk prediction model of prolonged intensive care unit length of stay in the pediatric liver transplant population. METHODS: This is a retrospective analysis of consecutive pediatric isolated liver transplant recipients at a single institution between April 1, 2013 and April 30, 2020. All patients under the age of 18 years receiving a liver transplant were included in the study (n = 186). The primary outcome was intensive care unit length of stay greater than 7 days. RESULTS: Recipient and donor characteristics were used to develop a multivariable logistic regression model. A total of 186 patients were included in the study. Using multivariable logistic regression, we found that age < 12 months (odds ratio 4.02, 95% confidence interval 1.20-13.51, p = .024), metabolic or cholestatic disease (odds ratio 2.66, 95% confidence interval 1.01-7.07, p = .049), 30-day pretransplant hospital admission (odds ratio 8.59, 95% confidence interval 2.27-32.54, p = .002), intraoperative red blood cells transfusion >40 mL/kg (odds ratio 3.32, 95% confidence interval 1.12-9.81, p = .030), posttransplant return to the operating room (odds ratio 11.45, 95% confidence interval 3.04-43.16, p = .004), and major postoperative respiratory event (odds ratio 32.14, 95% confidence interval 3.00-343.90, p < .001) were associated with prolonged intensive care unit length of stay. The model demonstrates a good discriminative ability with an area under the receiver operative curve of 0.888 (95% confidence interval, 0.824-0.951). CONCLUSIONS: We develop and validate a model to predict prolonged intensive care unit length of stay in pediatric liver transplant patients using risk factors from all phases of the perioperative period.

Tell me something interesting: Clinical utility of machine learning prediction models in the ICU.

In recent years, extensive resources are dedicated to the development of machine learning (ML) based clinical prediction models for intensive care unit (ICU) patients. These models are transforming patient care into a collaborative human-AI task, yet prediction of patient-related events is mostly treated as a standalone goal, without considering clinicians' roles, tasks or workflow in depth. We conducted a mixed methods study aimed at understanding clinicians' needs and expectations from such systems, informing the design of machine learning based prediction models. Our findings identify several areas of focus where clinicians' needs deviate from current practice, including desired prediction targets, timescales stemming from actionability requirements, and concerns regarding the evaluation and trust in these algorithms. Based on our findings, we suggest several design implications for ML-based prediction tools in the ICU.

Acute hemodialysis therapy in neonates with inborn errors of metabolism.

BACKGROUND: Inborn errors of metabolism (IEM), including organic acidemias and urea cycle defects, are characterized by systemic accumulation of toxic metabolites with deleterious effect on the developing brain. While hemodialysis (HD) is most efficient in clearing IEM-induced metabolic toxins, data regarding its use during the neonatal period is scarce. METHODS: We retrospectively summarize our experience with HD in 20 neonates with IEM-induced metabolic intoxication (seven with maple syrup urine disease, 13 with primary hyperammonia), over a 16-year period, between 2004 and 2020. All patients presented with IEM-induced neurologic deterioration at 48 h to 14 days post-delivery, and were managed with HD in a pediatric intensive care setting. HD was performed through an internal jugular acute double-lumen catheter (6.5-7.0 French), using an AK-200S (Gambro, Sweden) dialysis machine and tubing, with F3 or FXpaed (Fresenius, Germany) dialyzers. RESULTS: Median (interquartile range) age and weight at presentation were 5 (3-8) days and 2830 (2725-3115) g, respectively. Two consecutive HD sessions decreased the mean leucine levels from 2281 ± 631 to 179 ± 91 µmol/L (92.1% reduction) in MSUD patients, and the mean ammonia levels from 955 ± 444 to 129 ± 55 µmol/L (86.5% reduction), in patients with hyperammonemia. HD was uneventful in all patients, and led to marked clinical improvement in 17 patients (85%). Three patients (15%) died during the neonatal period, and four died during long-term follow-up. CONCLUSIONS: Taken together, our results indicate that HD is safe, effective, and life-saving for most neonates with severe IEM-induced metabolic intoxication, when promptly performed by an experienced and multidisciplinary team. A higher resolution version of the Graphical abstract is available as Supplementary information.

Same baby, different care: variations in practice between neonatologists and pediatric intensivists.

The aim of the study was to identify and explore areas in neonatal care in which significant differences in clinical care exist, among neonatal intensive care (NICU) and pediatric intensive care (PICU) physicians. A questionnaire presenting three common scenarios in neonatal critical care-severe pneumonia, post-cardiac-surgery care, and congenital diaphragmatic hernia (CDH) was electronically sent to all PICU and NICU physicians in Israel. The survey was completed by 110 physicians. Significant differences were noted between NICU and PICU physicians' treatment choices. A non-cuffed endotracheal tube, initial high-frequency ventilation, and lower tidal volumes when applying synchronized-intermittent-mechanical-ventilation were selected more often by NICU physicians. For sedation/analgesia, NICU physicians treated as needed or by continuous infusion of a single agent, while PICU physicians more often chose to continuously infuse ≥ 2 medications. Fentanyl, midazolam, and muscle relaxants were chosen more often by PICU physicians. Morphine administration was similar for both groups. Treating CDH with pulmonary hypertension and systemic hypotension, NICU physicians more often began treatment with high dose dopamine and/or dobutamine, while PICU physicians chose low-dose adrenalin and/or milrinone. For vascular access NICU physicians chose umbilical lines most often, while PICU physicians preferred other central sites. CONCLUSION: Our study identified major differences in respiratory and hemodynamic care, sedation and analgesia, and vascular access between NICU and PICU physicians, resulting from field-specific consensus guidelines and practice traditions. We suggest to establish joint committees from both professions, aimed at finding the optimal treatment for this vulnerable population - be it in the NICU or in the PICU. WHAT IS KNOWN: • Variability in neonatal care between the neonatal and the pediatric intensive care units has been previously described. WHAT IS NEW: • This scenario-based survey study identified major differences in respiratory and hemodynamic care, sedation and analgesia, and vascular access between neonatologists and pediatric intensivists, resulting from lack of evidence-based literature to guide neonatal care, field-specific consensus guidelines, and practice traditions. • These findings indicate a need for joint committees, combining the unique skills and literature from both professions, to conduct clinical trials focusing on these specific areas of care, aimed at finding the optimal treatment for this vulnerable population - be it in the neonatal or the pediatric intensive care unit.

Methylene Blue Administration for Distributive Shock States in Critically Ill Children.

BACKGROUND: Methylene blue (MB), an inhibitor of nitric oxide synthesis and its effects is a potentially effective treatment against distributive shock states such as septic shock and vasoplegic syndrome. MB has been shown to alleviate vasoplegia and promote an increase in blood pressure. It may reduce mortality. However, in the pediatric population, there are few case reports and only one controlled study on administration of MB use for vasoplegia, sepsis, or shock in general. OBJECTIVES: To summarize the experience of administering MB for vasoplegic shock in a tertiary care pediatric intensive care unit. METHODS: A retrospective chart review of seven pediatric cases treated with MB for vasoplegic shock was conducted. MB was administered as a bolus followed by continuous infusion. The authors measured blood pressure, vasopressor, and inotropic support. Patient outcome was monitored. RESULTS: The authors observed a favorable hemodynamic response with an increase in blood pressure and a reduction in vasopressor and inotropic support needed following MB administration in six patients. No side effects were observed. Three patients eventually died one to two days later, secondary to their underlying disease. CONCLUSIONS: This case series adds to the small body of evidence in the pediatric population supporting the use of MB for distributive shock states and emphasizes the need for larger, randomized trials evaluating its role in vasoplegic shock treatment.

Temporal Variability in the Sampling of Vital Sign Data Limits the Accuracy of Patient State Estimation.

OBJECTIVES: Physiologic signals are typically measured continuously in the critical care unit, but only recorded at intermittent time intervals in the patient health record. Low frequency data collection may not accurately reflect the variability and complexity of these signals or the patient's clinical state. We aimed to characterize how increasing the temporal window size of observation from seconds to hours modifies the measured variability and complexity of basic vital signs. DESIGN: Retrospective analysis of signal data acquired between April 1, 2013, and September 30, 2015. SETTING: Critical care unit at The Hospital for Sick Children, Toronto. PATIENTS: Seven hundred forty-seven patients less than or equal to 18 years old (63,814,869 data values), within seven diagnostic/surgical groups. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Measures of variability (SD and the absolute differences) and signal complexity (multiscale sample entropy and detrended fluctuation analysis [expressed as the scaling component α]) were calculated for systolic blood pressure, heart rate, and oxygen saturation. The variability of all vital signs increases as the window size increases from seconds to hours at the patient and diagnostic/surgical group level. Significant differences in the magnitude of variability for all time scales within and between groups was demonstrated (p < 0.0001). Variability correlated negatively with patient age for heart rate and oxygen saturation, but positively with systolic blood pressure. Changes in variability and complexity of heart rate and systolic blood pressure from time of admission to discharge were found. CONCLUSIONS: In critically ill children, the temporal variability of physiologic signals supports higher frequency data capture, and this variability should be accounted for in models of patient state estimation.

Distributions and Behavior of Vital Signs in Critically Ill Children by Admission Diagnosis.

OBJECTIVES: Define the distributions of heart rate and intraarterial blood pressure in children at admission to an ICU based on admission diagnosis and examine trends in these physiologic signs over 72 hours from admission (or to discharge if earlier). DESIGN: A retrospective analysis of continuously acquired signals. SETTING: A quaternary and primary referral children's hospital with a general PICU and cardiac critical care unit. PATIENTS: One thousand two hundred eighty-nine patients less than 18 years old were analyzed. Data from individual patient admissions were divided into 19 groups by primary admission diagnosis or surgical procedure. INTERVENTIONS: None. MEASUREMENT AND MAIN RESULTS: Distributions at admission are dependent on patient age and admission diagnosis (p < 10(-6)). Heart rate decreases over time, whereas arterial blood pressure is relatively stable, with differences seen in the directions and magnitude of these trends when analyzed by diagnosis group (p < 10(-6)). Multiple linear regression analysis shows that patient age, diagnosis group, and physiologic vital sign value at admission explain 50-63% of the variation observed for that physiologic signal at 72 hours (or at discharge if earlier) with admission value having the greatest influence. Furthermore, the variance of either heart rate or arterial blood pressure for the individual patient is smaller than the variance measured at the level of the group of patients with the same diagnosis. CONCLUSIONS: This is the first study reporting distributions of continuously measured physiologic variables and trends in their behavior according to admission diagnosis in critically ill children. Differences detected between and within diagnostic groups may aid in earlier recognition of outliers as well as allowing refinement of patient monitoring strategies.

Aiming for Relevance.

Vital signs are crucial in intensive care units (ICUs). They are used to track the patient's state and to identify clinically significant changes. Predicting vital sign trajectories is valuable for early detection of adverse events. However, conventional machine learning metrics like RMSE often fail to capture the true clinical relevance of such predictions. We introduce novel vital sign prediction performance metrics that align with clinical contexts, focusing on deviations from clinical norms, overall trends, and trend deviations. These metrics are derived from empirical utility curves obtained in a previous study through interviews with ICU clinicians. We validate the metrics' usefulness using simulated and real clinical datasets (MIMIC and eICU). Furthermore, we employ these metrics as loss functions for neural networks, resulting in models that excel in predicting clinically significant events. This research paves the way for clinically relevant machine learning model evaluation and optimization, promising to improve ICU patient care.

Making machine learning matter to clinicians: model actionability in medical decision-making.

Machine learning (ML) has the potential to transform patient care and outcomes. However, there are important differences between measuring the performance of ML models in silico and usefulness at the point of care. One lens to use to evaluate models during early development is actionability, which is currently undervalued. We propose a metric for actionability intended to be used before the evaluation of calibration and ultimately decision curve analysis and calculation of net benefit. Our metric should be viewed as part of an overarching effort to increase the number of pragmatic tools that identify a model's possible clinical impacts.

Descriptive characteristics of continuous oximetry measurement in moderate to severe covid-19 patients.

Non-invasive oxygen saturation (SpO2) is a central vital sign used to shape the management of COVID-19 patients. Yet, there have been no report quantitatively describing SpO2 dynamics and patterns in COVID-19 patients using continuous SpO2 recordings. We performed a retrospective observational analysis of the clinical information and 27 K hours of continuous SpO2 high-resolution (1 Hz) recordings of 367 critical and non-critical COVID-19 patients hospitalised at the Rambam Health Care Campus, Haifa, Israel. An absolute SpO2 threshold of 93% most efficiently discriminated between critical and non-critical patients, regardless of oxygen support. Oximetry-derived digital biomarker (OBMs) computed per 1 h monitoring window showed significant differences between groups, notably the cumulative time below 93% SpO2 (CT93). Patients with CT93 above 60% during the first hour of monitoring, were more likely to require oxygen support. Mechanical ventilation exhibited a strong effect on SpO2 dynamics by significantly reducing the frequency and depth of desaturations. OBMs related to periodicity and hypoxic burden were markedly affected, up to several hours before the initiation of the mechanical ventilation. In summary, OBMs, traditionally used in the field of sleep medicine research, are informative for continuous assessment of disease severity and response to respiratory support of hospitalised COVID-19 patients. In conclusion, OBMs may improve risk stratification and therapy management of critical care patients with respiratory impairment.

The truth Hertz-synchronization of electroencephalogram signals with physiological waveforms recorded in an intensive care unit.

Objective.The ability to synchronize continuous electroencephalogram (cEEG) signals with physiological waveforms such as electrocardiogram (ECG), invasive pressures, photoplethysmography and other signals can provide meaningful insights regarding coupling between brain activity and other physiological subsystems. Aligning these datasets is a particularly challenging problem because device clocks handle time differently and synchronization protocols may be undocumented or proprietary.Approach.We used an ensemble-based model to detect the timestamps of heartbeat artefacts from ECG waveforms recorded from inpatient bedside monitors and from cEEG signals acquired using a different device. Vectors of inter-beat intervals were matched between both datasets and robust linear regression was applied to measure the relative time offset between the two datasets as a function of time.Main Results.The timing error between the two unsynchronized datasets ranged between -84 s and +33 s (mean 0.77 s, median 4.31 s, IQR25-4.79 s, IQR75 11.38s). Application of our method improved the relative alignment to within ± 5ms for more than 61% of the dataset. The mean clock drift between the two datasets was 418.3 parts per million (ppm) (median 414.6 ppm, IQR25 411.0 ppm, IQR75 425.6 ppm). A signal quality index was generated that described the quality of alignment for each cEEG study as a function of time.Significance.We developed and tested a method to retrospectively time-align two clinical waveform datasets acquired from different devices using a common signal. The method was applied to 33,911h of signals collected in a paediatric critical care unit over six years, demonstrating that the method can be applied to long-term recordings collected under clinical conditions. The method can account for unknown clock drift rates and the presence of discontinuities caused by clock resynchronization events.

Machine learning to support triage of children at risk for epileptic seizures in the pediatric intensive care unit.

Objective.Epileptic seizures are relatively common in critically-ill children admitted to the pediatric intensive care unit (PICU) and thus serve as an important target for identification and treatment. Most of these seizures have no discernible clinical manifestation but still have a significant impact on morbidity and mortality. Children that are deemed at risk for seizures within the PICU are monitored using continuous-electroencephalogram (cEEG). cEEG monitoring cost is considerable and as the number of available machines is always limited, clinicians need to resort to triaging patients according to perceived risk in order to allocate resources. This research aims to develop a computer aided tool to improve seizures risk assessment in critically-ill children, using an ubiquitously recorded signal in the PICU, namely the electrocardiogram (ECG).Approach.A novel data-driven model was developed at a patient-level approach, based on features extracted from the first hour of ECG recording and the clinical data of the patient.Main results.The most predictive features were the age of the patient, the brain injury as coma etiology and the QRS area. For patients without any prior clinical data, using one hour of ECG recording, the classification performance of the random forest classifier reached an area under the receiver operating characteristic curve (AUROC) score of 0.84. When combining ECG features with the patients clinical history, the AUROC reached 0.87.Significance.Taking a real clinical scenario, we estimated that our clinical decision support triage tool can improve the positive predictive value by more than 59% over the clinical standard.

An integration engineering framework for machine learning in healthcare.

Background and Objectives: Machine Learning offers opportunities to improve patient outcomes, team performance, and reduce healthcare costs. Yet only a small fraction of all Machine Learning models for health care have been successfully integrated into the clinical space. There are no current guidelines for clinical model integration, leading to waste, unnecessary costs, patient harm, and decreases in efficiency when improperly implemented. Systems engineering is widely used in industry to achieve an integrated system of systems through an interprofessional collaborative approach to system design, development, and integration. We propose a framework based on systems engineering to guide the development and integration of Machine Learning models in healthcare. Methods: Applied systems engineering, software engineering and health care Machine Learning software development practices were reviewed and critically appraised to establish an understanding of limitations and challenges within these domains. Principles of systems engineering were used to develop solutions to address the identified problems. The framework was then harmonized with the Machine Learning software development process to create a systems engineering-based Machine Learning software development approach in the healthcare domain. Results: We present an integration framework for healthcare Artificial Intelligence that considers the entirety of this system of systems. Our proposed framework utilizes a combined software and integration engineering approach and consists of four phases: (1) Inception, (2) Preparation, (3) Development, and (4) Integration. During each phase, we present specific elements for consideration in each of the three domains of integration: The Human, The Technical System, and The Environment. There are also elements that are considered in the interactions between these domains. Conclusion: Clinical models are technical systems that need to be integrated into the existing system of systems in health care. A systems engineering approach to integration ensures appropriate elements are considered at each stage of model design to facilitate model integration. Our proposed framework is based on principles of systems engineering and can serve as a guide for model development, increasing the likelihood of successful Machine Learning translation and integration.

Timing errors and temporal uncertainty in clinical databases-A narrative review.

A firm concept of time is essential for establishing causality in a clinical setting. Review of critical incidents and generation of study hypotheses require a robust understanding of the sequence of events but conducting such work can be problematic when timestamps are recorded by independent and unsynchronized clocks. Most clinical models implicitly assume that timestamps have been measured accurately and precisely, but this custom will need to be re-evaluated if our algorithms and models are to make meaningful use of higher frequency physiological data sources. In this narrative review we explore factors that can result in timestamps being erroneously recorded in a clinical setting, with particular focus on systems that may be present in a critical care unit. We discuss how clocks, medical devices, data storage systems, algorithmic effects, human factors, and other external systems may affect the accuracy and precision of recorded timestamps. The concept of temporal uncertainty is introduced, and a holistic approach to timing accuracy, precision, and uncertainty is proposed. This quantitative approach to modeling temporal uncertainty provides a basis to achieve enhanced model generalizability and improved analytical outcomes.

Analyzing Continuous Physiologic Data to Find Hemodynamic Signatures Associated With New Brain Injury After Congenital Heart Surgery.

Continuous data capture technology is becoming more common. Establishing analytic approaches for continuous data could aid in understanding the relationship between physiology and clinical outcomes. OBJECTIVES: Our objective was to design a retrospective analysis for continuous physiologic measurements and their relationship with new brain injury over time after cardiac surgery. DESIGN SETTING AND PARTICIPANTS: Retrospective cohort study in the Cardiac Critical Care Unit at the Hospital for Sick Children in patients after repair of transposition of the great arteries (TGA) or single ventricle (SV) lesions. MAIN OUTCOMES AND MEASURES: Continuously acquired physiologic measurements for up to 72 hours after cardiac surgery were analyzed for association with new brain injury by MRI. Distributions of heart rate (HR), systolic blood pressure (BP), and oxygen saturation (Spo2) for SV and TGA were analyzed graphically and with descriptive statistics over postoperative time for data-driven variable selection. Mixed-effects regression analyses characterized relationships between HR, BP, and Spo2 and new brain injury over time while accounting for variation between patients, measurement heterogeneity, and missingness. RESULTS: Seventy-seven patients (60 TGA; 17 SV) were included. New brain injury was seen in 26 (34%). In SV patients, with and without new brain injury, respectively, in the first 24 hours after cardiac surgery, the median (interquartile range) HR was 172.0 beats/min (bpm) (169.7-176.0 bpm) versus 159.6 bpm (145.0-167.0 bpm); systolic BP 74.8 (67.9-78.5 mm Hg) versus 68.9 mm Hg (61.6-70.9 mm Hg). Higher postoperative HR (parameter estimate, 19.4; 95% CI, 7.8-31; p = 0.003 and BP, 8.6; 1.3-15.8; p = 0.024) were associated with new brain injury in SV patients. The strength of this relationship decreased with time. CONCLUSIONS AND RELEVANCE: Retrospective analysis of continuous physiologic measurements can provide insight into changes in postoperative physiology over time and their relationship with new brain injury. This technique could be applied to assess relationships between physiologic data and many patient interventions or outcomes.

Dynamics of excitability over extended timescales in cultured cortical neurons.

Although neuronal excitability is well understood and accurately modeled over timescales of up to hundreds of milliseconds, it is currently unclear whether extrapolating from this limited duration to longer behaviorally relevant timescales is appropriate. Here we used an extracellular recording and stimulation paradigm that extends the duration of single-neuron electrophysiological experiments, exposing the dynamics of excitability in individual cultured cortical neurons over timescales hitherto inaccessible. We show that the long-term neuronal excitability dynamics is unstable and dominated by critical fluctuations, intermittency, scale-invariant rate statistics, and long memory. These intrinsic dynamics bound the firing rate over extended timescales, contrasting observed short-term neuronal response to stimulation onset. Furthermore, the activity of a neuron over extended timescales shows transitions between quasi-stable modes, each characterized by a typical response pattern. Like in the case of rate statistics, the short-term onset response pattern that often serves to functionally define a given neuron is not indicative of its long-term ongoing response. These observations question the validity of describing neuronal excitability based on temporally restricted electrophysiological data, calling for in-depth exploration of activity over wider temporal scales. Such extended experiments will probably entail a different kind of neuronal models, accounting for the unbounded range, from milliseconds up.

Tradeoffs and constraints on neural representation in networks of cortical neurons.

Neural representation is pivotal in neuroscience. Yet, the large number and variance of underlying determinants make it difficult to distinguish general physiologic constraints on representation. Here we offer a general approach to the issue, enabling a systematic and well controlled experimental analysis of constraints and tradeoffs, imposed by the physiology of neuronal populations, on plausible representation schemes. Using in vitro networks of rat cortical neurons as a model system, we compared the efficacy of different kinds of "neural codes" to represent both spatial and temporal input features. Two rate-based representation schemes and two time-based representation schemes were considered. Our results indicate that, by large, all representation schemes perform well in the various discrimination tasks tested, indicating the inherent redundancy in neural population activity; Nevertheless, differences in representation efficacy are identified when unique aspects of input features are considered. We discuss these differences in the context of neural population dynamics.