Search

Mechanical ventilation (MV) is a core treatment for patients suffering from respiratory disease and failure. However, MV settings are not standardized due to significant inter- and intra- patient variability in response to care, leading to variability in outcome. There is thus a need to personalize MV settings. This research significantly extends a single compartment lung mechanics model with physiologically relevant basis functions, and uses it to identify patient-specific lung mechanics and predict response to changes in MV settings. Nonlinear evolution of pulmonary elastance over positive end expiratory pressure (PEEP) is modelled by a newly proposed, physiologically relevant and simplified compensatory function to enable prediction of pulmonary response for both volume-controlled ventilation (VCV) and pressure-controlled ventilation (PCV), and identified as patient-specific using each patient’s data at a baseline PEEP. Predictions at higher PEEP levels test the validity of the proposed models based on errors in predicted peak inspiratory pressure (PIP) in two VCV trials and volume (PIV) in one PCV trial. A total of 210 prediction cases over 36 patients (22 VCV; 14 PCV) yielded absolute predicted PIP errors within 1.0 cmH2O (2.3%) and 3.3 cmH2O (7.3%) for 90% cases in VCV, while predicted PIV errors are within 0.073 L (16.8%) for 85% cases in PCV. In conclusion, a novel deterministic virtual patient model is presented, able to offer accurate prediction of pulmonary response across a wide range of PEEP changes for the two main MV modes used clinically, enabling predictive decision support in real-time to safely personalize and optimize care.

Background Mechanical Ventilation (MV) is an important intensive care therapy. It is often used in neonatal intensive care units (NICUs) to treat respiratory distress syndrome. This paper uses model-based methods in a first in-depth attempt to quantify the underlying lung mechanics in NICU patients. Methods Up to 24 h of airway flow and pressure data were recorded in 10 mechanically ventilated patients. A single compartment model with added term for the pressure drop across the endotracheal tube is used to identify breath-specific elastance and resistance. Result The model was fitted to 422,475 (79%) breaths of 535,428, with the remainder comprising a range of asynchronies. Elastance was median 1.62 [IQR: 0.85–2.25] cmH2O/ml and resistance 5.22 [0.00–33.85] cmH2O.s/ml. Patients treated with surfactant therapy had significantly lower specific elastance (adjusting for weight) than those without (p ≤ 0.01). A decrease in elastance with increasing weight was also noted. Conclusion The single compartment model was successfully fit with low error. The subgroup cohorts showed expected trends and further validates the identified model values. There was significant breath-to-breath variability in elastance within and between patients.

Background The use of single point laser Doppler vibrometer is well established for evaluating displacement or validation in many applications such as scientific research, defence, manufacturing, and biomedicine with other techniques. In this research, breast shaped fabricated phantoms are used in laser Doppler vibrometery to validate the optical flow motion measuring algorithm used in a mechanical vibration based breast cancer screening system. Methods Three different silicone phantoms were used: healthy; 10 mm inclusion and 20 mm inclusion. The overall goal was to use single point laser Doppler vibrometer data to validate digital image-based elasto-tomography (DIET) motion data from optical flow tracking at frequencies of 16 Hz, 24 Hz, 32 Hz, and 40 Hz. Summary of findings Results show excellent validation with errors less than 6% for healthy phantom, and errors less than 8% for 10 mm inclusion and 20 mm inclusion. Conclusions The overall results from the optical flow motion algorithm used with DIET and a laser Doppler vibrometer at every frequency show the optical flow algorithm captures surface motion of breast shaped silicone phantoms with good accuracy. The optical flow algorithm is thus suitable and robust enough for use in clinical breast screening. Finally, the errors presented quantify explicitly the error of this algorithm verses this laser-based gold standard.

Background Insulin therapy for glycaemic control (GC) in critically ill patients may improve outcomes by reducing hyperglycaemia and glycaemic variability, which are both associated with increased morbidity and mortality. However, initial positive results have proven difficult to repeat or achieve safely. STAR (Stochastic TARgeted) is a model-based glycaemic control protocol using a risk-based dosing approach. STAR uses a 2D stochastic model to predict distributions of likely future changes in model-based insulin sensitivity (SI) based on its current value, and determines the optimal intervention. Objectives This study investigates the impact of a new 3D stochastic model on the ability to predict more accurate future SI distributions, which would allow more aggressive insulin dosing and improved glycaemic control. Methods The new 3D stochastic model is built using both current SI and its prior variation to predict future SI distribution from 68,629 h of clinical data (819 GC episodes). The 5th–95th percentile range of predicted SI distribution are calculated and compared with the 2D model. Results Results show the 2D model is over-conservative compared to the 3D case for more than 77% of the data, predominantly where SI is stable (|%ΔSI| ≤ 25%). These formerly conservative prediction ranges are now ∼30% narrower with the 3D model, which safely enables more aggressive insulin dosing for these patient hours. In addition, distributions of predicted SI within the 5th–95th percentile range are much closer to the ideal value of 90% for more patients with the 3D model. Conclusions The new 3D model better characterises patient specific metabolic variability and patient specific response to insulin, allowing more optimal insulin dosing to increase performance and safety.

In the intensive care unit (ICU), stress-induced insulin resistance leading to hyperglycemia is commonplace. If safe and effective glycemic control (GC) can be provided, a significant reduction in the negative effects of dysglycemia can be achieved. The Intensive Control Insulin-Nutrition-Glucose (ICING) model has worked particularly well in guiding patient-specific GC. The current method to identify patient- and time- specific insulin sensitivity (SI) with this model employs a discrete stepwise hourly function, which is effective, but not physiologically representative at hourly transitions. 2nd Order B-spline basis functions (BF) are investigated as a more continuous and physiologically relevant alternative for two different cohorts of patient data (Benchmark, 20 patients; Stochastic TARgeted (STAR) sub-cohort, 72 patients). Various knot-widths (KW) of 2nd Order B-spline BFs were investigated and compared to the currently used step function BF in terms of physiological relevance, identifiability, robustness to false measurements, and susceptibility to noise to ensure the most physiologically realistic representation of SI. The 180 min KW 2nd Order B-spline BFs provided the most physiologically realistic fit to the blood glucose (BG) measurements in both cohorts (Standard Deviation of relative fitting error: Benchmark 8.7% vs. 9.35% BG meter, STAR sub-cohort 6.0% vs. 6.0% BG meter), while also having significantly less susceptibility to false BG measurements compared to the current BF. The 180 min 2nd Order B-spline BF provides a smooth second order continuous, more physiologically representative, and identifiable model of the dynamics of SI, and a more realistic resulting BG solution. The outcome provides a promising new addition to the ICING model and STAR GC protocol, while also opening up several new potential continuous time analyses of SI dynamics, not previously possible.

Background Glycaemic control (GC) in critical care can reduce mortality and improve clinical outcomes. Model based GC allows personalised and effective predictive control. Improving accuracy and precision of patient specific models, parameter identification, and resulting blood glucose predictions would enable more effective model-based GC algorithms. Methods Glycaemic data from 30 critically ill patient episodes was used to fit a model of glucose dynamics. In this model, insulin sensitivity (SI) is identified through time using b-spline basis functions. The number of basis functions, M, relative to the number of data points in each data set, N, determines the level of parameterisation of the model. By fitting the model to all 30 data sets, the identified SI profiles could be used to build stochastic maps of SI changes over time. These maps show how SI is expected to change based on the current SI and the overall behaviour of this cohort. Thus, for a given time step into the future, SI prediction distributions can be found. Stochastic maps were made for varying basis function degree (d) (d ∊ {0, 1, 2}), and ratio of M to N (M:N ∊ {0.3, 0.65, 0.85, 1}). Using small subsets of each data set, many trials were carried out to compare SI prediction distributions to the distribution of the true SI values at the next measurement, for each combination of M and d. The main aim was to observe how the model parameterisation affected the predictive ability of the model. Outcomes from an Akaike Information Criterion (AIC) analysis were compared to the prediction analysis. AIC is a method of model comparison that assesses which of the given models provides the best trade-off between goodness-of-fit and model complexity, based on the expected measurement noise. For each basis function order, an AIC analysis was used to select the best M:N ratio of the four options tested. Results Increasing the parameterisation of the model resulted in lower model-data residuals and thus wider stochastic maps, as SI changed at a faster rate to enable the model to more closely fit to the data. Therefore, increasing M:N resulted in wider prediction distributions. In all cases, when M:N = 1, the prediction distributions were too wide, and when M:N = 0.3 the prediction distributions were too narrow. Increasing the basis function order resulted in tighter prediction distributions, and allowed more accurate predictions to be made with a higher M:N ratio. The ratios that gave the most accurate predictions were M:N = 0.65 when d = 0, M:N = 0.85 when d = 1, and M:N = 1 when d = 2. In contrast, the AIC analysis found that an M:N ratio of approximately 0.45 was optimal in all cases. Conclusions This study presented a new strategy for observing how the level of parameterisation and basis function order affects accuracy of future predictions of SI variability. Applications based on the findings of this research with a larger cohort could lead to improved predictive capability in GC algorithms. Accurate predictions would ultimately allow model-based GC algorithms to be implemented more effectively, increasing clinician confidence and improving patient outcomes.

Background and objective Cancerous tumors exhibit considerably higher stiffness than healthy breast tissues. This research develops a hysteresis loop analysis detection method for breast cancer diagnosis, which is validated using silicone phantom tests to provide an initial proof of concept of the overall DIET diagnostic approach. Methods Steady state sinusoidal displacements induced by mechanical actuation are captured using the DIET system for 4 silicone breast phantoms (1 homogeneous “healthy”; 3 with 10–20 mm stiffer inclusion “tumors” at different locations). Hysteresis loops for each measured reference point on the breast surface (N≈14,500/phantom) are constructed from the measured displacements and a calculated mass normalized restoring force. The distribution of nominal elastic stiffness over breast surface is estimated using a hypothesis test and regression analysis. A small motion threshold, Δ d , is used to reduce errors in identified elastance and its values determined from sensitivity analysis. Results Sensitivity analysis shows displacement reconstruction errors from very small motions can be eliminated using a threshold Δ d = 6 μ m to clearly reduce errors in identified nominal elastance. Results for all 4 phantoms indicate stiffer inclusions induce a greater number of reference points with stiffness ≥2 × higher in the segment with the inclusion than the nominal stiffness in the “healthy” segments in the x , y and total net 2D displacement directions. Therefore, the presence of a stiffer “tumor” inclusion is automatically detected and located by comparing the nominal stiffness to a calculated threshold, k t . Inclusion segments are finally localized algorithmically using an inclusion indicator, K index . Conclusions The overall results validate the capability of the proposed method to accurately detect and locate a stiff inclusion in typical, representative silicone breast phantoms without misidentifying other regions or a healthy no-inclusion phantom in the non-invasive DIET system. The methods are entirely generalizable to human clinical data from a DIET system, and the results justify such clinical trials.

Background and motivation Athletes experience substantial physiological impacts after exhaustive events, including post-exercise hyperglycemia. Objective measurement of athlete recovery is a major goal in sports training to optimize performance. Continuous glucose monitors (CGMs) provide frequent, minimally invasive blood glucose measurements to potentially observe the metabolic impact of a major endurance exercise event and subsequent recovery. Methods Ten sub-elite athletes were fitted with CGMs. Blood glucose levels (BGLs), exercise, and nutrition were monitored for 4–6 days. Athletes performed an endurance exercise test to exhaustion after 1–2 days of monitoring. Validated State and Substate pattern analyses quantified glycemic level and variability before and after the exercise test, including basal glucose levels and glycemic responses to carbohydrate intake. Results Glycemic variability and glycemic response to carbohydrate intake increased on the day of the test and returned to normal the next day. Overnight glucose levels remained elevated up to 3–4 days post-test. These results capture the metabolic impact of the test and subsequent recovery. Conclusions and significance The initial hyperglycemic period and lasting metabolic effects of an endurance test to exhaustion are quantifiable in commercially available CGM devices. The results encourage further investigation into using CGMs to monitor athletic recovery after intense exercise events.

Background Spontaneous breathing (SB) effort is common in invasively ventilated patients, as they recover or at low sedation levels. SB effort makes estimation of patient-specific lung mechanics difficult as it introduces unknown inputs. Currently, there is no proven, non-invasive method to quantify SB effort. This study develops non-invasive model-based metrics for SB. Methods Lung mechanics are identified using a proven model to identify time-varying elastance over inspiration using retrospective data from 22 patients on neutrally adjusted ventilatory assist (NAVA) ventilation. Two methods are explored to quantify negative elastance values associated with breathing effort: a zero-crossing and a trapezoidal method. Area under the resulting positive and negative portions of the time-varying elastance curve is used to quantify SB. Results The median [interquartile range (IQR)] of AUC quantification of SB for the zero crossing method was −3.29 [−4.80 to −2.50] cmH2O/l, and −1.90 [−2.36 to −1.66] cmH2O/l for the trapezoidal method. AUC values between the two methods were highly correlated, indicating consistency of both model-based methods to quantify SB. Conclusion Two model-based methods for SB quantification provide consistent metrics for SB. These methods provide computationally simple and cost effective methods to estimate SB effort, which may have clinical value in assessing patient-ventilator synchrony, choice of ventilation mode, or need for weaning.

Sepsis is highly correlated with mortality and morbidity. Sepsis is a clinical condition demarcated as the existence of infection and systemic inflammatory response syndrome, SIRS. Confirmation of infection requires a blood culture test, which requires incubation, and thus results take at least 48 h for a syndrome that requires early direct treatment. Since sepsis has a strong inflammatory component, it is hypothesized that metabolic markers affected by inflammation, such as insulin sensitivity, might provide a metric for more rapid, real-time diagnosis. This study uses clinical data from 30 sepsis patients (7624 h in ICU) of whom 60% are male. Median age and median Apache II score are 63 years and 19, respectively. Model-identified insulin sensitivity (SI) profiles were obtained for each patient, and insulin sensitivity and its hourly changes were correlated with modified hourly sepsis scores (SSH1). SI profiles and values were similar across the cohort. The sepsis score is highly variable and changes rapidly. The modified hourly sepsis score, SSH1, shows a better relation with insulin sensitivity due to less fluctuation in the SIRS element. Median SI and median ΔSI of the cohort is 0.4193e-3 and 0.004253e-3 L/mU.min, respectively. Additionally, median SI are 4.392 × 10−4 L/mU min (SSH1 = 0), 4.153 × 10−4 L/mU min (SSH1 = 1), 3.752 × 10−4 L/mU min (SSH1 = 2) and 2.353 × 10−4 L/mU min (SSH1 = 3). Significant relationship between insulin sensitivity across different SSH1 groups was observed (p < 0.05) even when corrected for multiple comparisons. CDF of SI indicates that insulin sensitivity is more significant when comparing an hourly sepsis score at a very distinguished level.

Advancements in health technologies are crucial to support healthcare professionals, improve patient outcomes, and best utilize increasingly scarce and under-demand healthcare resources. This research presents an initial proof-of-concept study of simple, non-invasive monitoring techniques used in Mechanical Ventilation (MV), which is the primary therapy for Acute Respiratory Distress Syndrome (ARDS). The high levels of inter-patient variability seen in patients with ARDS have resulted in much speculation about the ideal method of determining ventilation settings, such as tidal volume (Vt) and Positive End Expiratory Pressure (PEEP). One of the oldest and simplest methods is acoustic sensing of recruitment and lung condition. This project involves using a digital recording stethoscope to monitor the acoustic output of patients in the Intensive Care Unit (ICU) during mechanical lung ventilation. During lung recruitment, ‘crackles’ can be heard within the chest cavity with a stethoscope. These crackles vary significantly, depending on the status of the patient’s respiratory system and are used as an indicator of the level of alveolar recruitment. This preliminary, proof-of-concept study focused on crackle detection and involved gathering sound samples from patients in the Christchurch Hospital ICU with evidence of crackles in the chest cavity. Frequency based analysis showed that crackles can be detected as emissions with higher power levels between 100 and 300 Hz (subject to patient variability). The ability to non-invasively record, detect and quantify the intensity of crackles could provide immediate feedback to clinicians and, in the long term, aid in the optimization of ventilator therapy.

Glycaemia of outpatients with diabetes is very difficult to model due to sparse, low quality data, inter- and intra- patient variability and a myriad of other factors that have significant, but intermittent effects. For model-based control strategies, such factors can contribute non-random grey noise that can confound patient-specific models, and reduce prediction confidence. Incorporating such factors in glycaemic models would significantly improve control if the data available allows practical identifiability of these model parameters. This review compares and analyses models that capture the glycaemic grey-noise impact of nutrition, stress and illness, exercise and circadian rhythms are compared and considered in the context of practical application to model-based outpatient diabetes management. Candidate models to capture glycaemia in outpatients with diabetes must be considered in the context of the data needed to identify the models, the ability of the model to adapt to the patient state, and the practical identifiability of the models for a particular data quality. In particular, the outpatient environment presents challenges for acquiring quality data and gold-standard methods of measurement are frequently infeasible. Only models that can be practically identified using the type and quality of data available in an outpatient setting should be considered, thus informing model development. Furthermore, the candidate models should also be capable of capturing inter- and intra- patient variability in the heterogeneous metabolism of individuals with diabetes. Finally, practically identifiable models need to also be identifiable over a clinically acceptable time period so the models are useful in context for managing diabetes.

Severe sepsis and septic shock are common causes of intensive care unit (ICU) admission, length of stay, mortality and cost. Sepsis is very difficult to diagnose in real-time and early treatment can show positive benefits. Current sepsis scores are hierarchically calculated, creating significant case-control and/or misclassification biases in developing real-time diagnostic classifiers based on real-time clinical data. This study examines the impact on kernel density classifier performance of a continuous sepsis score derived from a hierarchical score and based on current well-accepted metrics. Clinical data and model-based insulin sensitivity from 36 patients in the Christchurch Hospital ICU (6071 h) is used to segregate severe sepsis and septic shock from mild sepsis and SIRS using a modified continuous sepsis score using a kernel density estimator (KDE) method. Best case resubstitution, worst case bootstrap, and expected case .632 bootstrap estimates are calculated. Diagnostic quality is assessed using likelihood ratios (LHRs), ROC curves, diagnostic odds-ratios (DOR), sensitivity and specificity. Results were compared to those using the hierarchical sepsis score and the same KDE. The continuous sepsis score yielded smoother, more realistic time-varying trajectories for sepsis evolution. The classifier provided very good results, potentially able to change clinical decisions, but did not reach the best diagnostic accuracy for all performance metrics. Using several metrics avoided assessment bias, and likelihood ratios provided the best overall assessment relative to clinical decision making processes. Finally, quantifying infection presence is a source of case-control and misclassification bias in classifier development, showing clear need for improved sepsis scores.

Mechanical ventilation (MV) is the primary way to treat patients with acute respiratory distress syndrome (ARDS) in the intensive care unit (ICU). Positive-end-expiratory-pressure (PEEP) is often applied in MV to maximise gas exchange and prevent further lung damage. However, optimal MV with patient-specific PEEP level is subjective, based on clinician experience, thus increasing the risk variability, and cost of care. In this study, a single compartment lung mechanics model is employed to predict pressure outcomes at higher PEEP levels using data from lower PEEP levels. Particularly, a hysteresis loop analysis (HLA) algorithm is applied to both the dynostatic curve and pressure-volume (PV) loop to identify the lung elastance at lower PEEP levels. Pulmonary resistance can then be calculated with measured airway pressure and identified elastance. Finally, the elastance and resistance at higher PEEP levels are predicted based on model fitting techniques to obtain the pressure with changes of PEEP. The result of pressure fitting and prediction show 95% of the airway pressure curve errors are within 3% across all the 8 data sets, while peak-inspiratory-pressure (PIP) prediction errors are within 1% for all data sets. The overall approach is readily automated, thus providing an initial critical information in determining optimal MV and PEEP settings. Such a model could also be used to validate the prediction accuracy of other more complex models and improve the confidence of their clinical application. Most importantly, the ability to predict PIP and overall pressure trajectories provide a means to safely titrate PEEP to optimal levels without the necessary to build an accurate enough mathematical model which could cost plenty of time and effort.

Objective Safe, effective glycaemic control (GC) requires accurate prediction of future patient insulin sensitivity (SI), balancing the risk of hyper- and hypo-glycaemia. The stochastic targeted (STAR) protocol combines a clinically validated metabolic model and SI metric with a risk-based stochastic approach to optimise patient specific insulin and feed rates. Validated virtual trials comparing a novel 3D stochastic model for prediction of future patient SI using current patient SI and current blood glucose (BG) to an existing 2D stochastic model for SI prediction were conducted. Methods The virtual trials involved 1477 retrospective patients across two hospitals and two GC protocols. They were conducted using five-fold cross-validation to build each stochastic model, ensuring independent test data. Results The 3D stochastic model shifted BG from the 4.4–8.0 mmol/L target band towards the lower 4.4–6.5 mmol/L band, providing a decrease from 12.31 % to 11.19 % in hyperglycaemic hours (BG > 8.0 mmol/L), but only a 0.24 % increase, from 1.01 % to 1.25 %, in light hypoglycaemic hours (BG Conclusions The 3D stochastic model provided greater personalisation and better realised STAR’s design philosophy of minimising hyperglycaemic events for an acceptable clinical risk of 5.0 % BG

Critically ill patients often experience stress-induced hyperglycaemia, which results in increased morbidity and mortality. Glycaemic control (GC) can be implemented in the intensive care unit (ICU) to safely manage hyperglycaemia. Two protocols SPRINT and STAR, have been implemented in the Christchurch ICU, and have been successful in treating hyperglycaemia while decreasing the risk of hypoglycaemia. This paper presents a new GC protocol that implements the probabilistic, stochastic forecasting methods of STAR, while formalizing the control methodology using model predictive control (MPC) theory to improve the ability to tune the dynamic response of the controller. This Stochastic Model Predictive (STOMP) controller predicts the response to a given insulin/nutrition intervention, and attributes weighted penalty values to several key performance metrics. The controller thus chooses an intervention at each hour that minimizes the sum of these penalties over a prediction window of 6 h, which is twice as long as the 3-h window used in STAR. Clinically validated virtual trials were used to evaluate the relative performance of STOMP. Results showed STOMP was able to obtain results very similar to STAR with both protocols maintaining approximately 85% of time within 4.4–8.0 mmol/L glycaemic band, and only 4–5 patients of the 149 patient STAR cohort having blood glucose (BG)

A minimal recruitment model can be used to guide mechanical ventilation PEEP selection for acute respiratory distress syndrome (ARDS) patients. However, implementation of this model requires a specific clinical protocol and is computationally expensive, and thus not suitable for bedside application. This work aims to improve the performance and bedside utility of the minimal recruitment model through simplifying the model, and improving the identification algorithm without compromising the model's physiological relevance to the disease. Identifying the model requires fitting of 8 unique parameters to pressure–volume data at multiple levels of positive end-expiratory pressure (PEEP). A minimal algorithm is proposed to improve the model's performance. The algorithm utilises a non-linear least-squares solver to estimate a global set of the parameters to a pressure–volume curve at one PEEP level. These global parameters were then subsequently used to estimate other parameters at different pressure–volume curves. The accuracy and computational performance of the minimal algorithm is compared to the grid search algorithm for 2 ARDS patient cohorts. The median [IQR] absolute percentage curve fitting error over all patients for grid search algorithm is 1.40% [0.55–3.75], and for the minimal algorithm is 2.43% [0.83–8.09] ( p R 2 > 0.9. The model fitting error for the minimal algorithm is higher than the grid search algorithm. However the model was able capture similar trends in physiologically relevant parameters without the loss of important clinical information. The minimal algorithm is less computationally intensive than the grid search algorithm, whilst still providing a means of selecting PEEP with only a small increase in model fitting error. The minimal algorithm is able to improve computational performance while maintaining the ability to capture physiological parameters as the grid search algorithm. The significant reduction in computational time encourages its clinical application at the bedside for decision making.

Magnetic Resonance Elastography (MRE) is able to identify mechanical properties of biological tissues in vivo based on underlying assumptions of the model used for inversion. Models, such as the linearly elastic or viscoelastic (VE), can be used with a single input frequency data and can produce a reasonable estimate of identified parameters associated with mechanical properties. However, more complex models, such as the Rayleigh damping (RD) model, are not identifiable given single frequency data without significant a priori information under certain conditions, thus limiting diagnostic potential. To overcome this limitation, two approaches have been postulated: simultaneous inversion across multiple input frequencies and a parametric approach, when only single frequency data is available. This research compares simultaneous multi-frequency (MF) RD reconstructions using both zero-order and power-law (PL) models with parametric reconstructions for a series of tissue-simulating phantoms, made of tofu and gelatine materials, tested at 4 frequencies (50 Hz, 75 Hz, 100 Hz and 125 Hz) that are commonly applied in clinical MRE examinations. Results indicate that accurate delineation of RD based properties and concomitant damping ratio ( ξ d ) using MF inversion is still a challenging task. Specific results showed that the real shear modulus ( μ R ) can be reconstructed well, while imaginary components representing attenuation ( μ I and ρ I ) had much lower quality. However, overall trends correlate well with the expected higher damping levels within the saturated tofu material compared to stiff gelatine in both phantoms. Depending on the phantom configuration, measured μ R values within the tofu and gelatine materials ranged from 4.77 to 7 kPa and 15.5 to 16.3 kPa, respectively, while damping levels were 11–19% and 3.1–4.3%, as expected. Correlation of the μ R and ξ d values with previously reported result measured by independent mechanical testing and VE based MRE is acceptable, ranging from 48 to 60%. Both PL and zero-order models produced similar qualitative and quantitate results, thus no significant advantage of the PL model was noted to account for dispersion characteristics of these types of materials. The relatively narrow range of frequencies used in this study limited practical identifiability and can thus produce a potentially false assurance of identifiability of the model parameters. We conclude that application of multiple input frequencies over a wide range, as well as selection of an appropriate model that can accurately account for dispersion characteristics of given materials are required for achieving robust practical identifiability of the RD model in time-harmonic MRE.

Background The pressure dependent recruitment model (PRM) is a comprehensive mathematical description of pulmonary mechanics in acute respiratory distress syndrome (ARDS). However, previous investigations of the PRM implied that the number of model parameters may cause inaccurate parameter estimation. Methods PRM models were evaluated for 12 ARDS patients that underwent a low-flow recruitment manoeuvre. The identified parameter set formed the basis of a parameter reduction investigation of the PRM. The parameter reduction investigation measured the mean cohort residual error (ψ) yielded by each possible combination of identified parameter set with the non-identified parameter values set to a priori population constants. Results Reducing the five variable PRM to a particular three variable model configuration produced a limited increase in model fit to data residuals (ψ5 = 22.68, ψ3 = 29.21 mbar). The reduced model evaluates airway-resistance, compliance and distension as model variables and uses population values for alveoli opening pressure and the ratio of open alveoli at end expiratory. Conclusions The reduced PRM model captures all major pressure–volume response features in the ARDS patients. Reduced parameterisation allows more robust parameter identification and thus more reliable parameter estimates that may prove more useful in a clinical setting.

Background The 30-min quick dynamic insulin sensitivity test (DISTq30) uses two blood-glucose measurements from a dynamic insulin sensitivity test protocol to provide low-cost, real-time insulin sensitivity ( SI ) measurements. However, the DISTq30 clinical protocol contains a potentially redundant 10-min period between glucose and insulin boluses that occur at t = 0 and t = 10 min. Methods A proposed protocol (DISTq20) reduces the DISTq30 test duration to 20 min and administers a combined 10 g glucose and 1 U insulin bolus at t = 0. The proposed protocol was evaluated against the clinically validated DISTq30 in a Monte Carlo analysis. 313 clinical responses to the dynamic insulin sensitivity and secretion tests (DISST) from three different studies were used to provide realistic parameter value sets. These values were used to create realistic in-silico responses to DISTq20 and DISTq30 protocols. Each simulated response was ‘sampled’ at the appropriate times and SI was identified 200 times in a Monte Carlo analysis with added random assay error for each protocol and parameter set. In a second analysis, the DISTq20 response was simulated with 0–50% inhibition of the first phase insulin response to assess robustness to this potential effect. Results Simulated noise had a very similar effect on DISTq30 and DISTq20 SI values ( R = 0.99). DISTq20 overestimated DISTq30 SI by a median 1.7% (IQR −4.3% to 7.3%). The second analysis showed that DISTq20 results were robust to variance in first phase insulin secretion ( R = 0.97). DISTq20 and DISTq30 both had a median CV of 7.9%. Conclusions Inconsequential differences between SI values found by the DISTq30 and the DISTq20 in-silico indicate that the DISTq20 may produce similar clinical results to the DISTq30. Further analysis showed that the identification method was robust to the assumption of zero insulin suppression.

Introduction Current successful AGC (Accurate Glycemic Control) protocols require extra clinical effort and are impractical in less acute wards where patients are still susceptible to stress-induced hyperglycemia. Long-acting insulin Glargine has the potential to be used in a low effort controller. However, potential variability in efficacy and length of action prevent direct in-hospital use in an AGC framework for less acute wards. Method Clinically validated virtual trials based on data from stable ICU patients from the SPRINT cohort who would be transferred to such an approach are used to develop a 24-h AGC protocol robust to different Glargine potencies (1.0×, 1.5× and 2.0× regular insulin) and initial dose sizes (dose = total insulin over prior 12, 18 and 24 h). Glycemic control in this period is provided only by varying nutritional inputs. Performance is assessed as %BG in the 4.0–8.0 mmol/L band and safety by %BG Results The final protocol consisted of Glargine bolus size equal to insulin over the previous 18 h. Compared to SPRINT there was a 6.9–9.5% absolute decrease in mild hypoglycemia (%BG Conclusion Based on current understanding of Glargine behaviour, a robust protocol for a 24–48 clinical trial has been designed to safely investigate possible differences in efficacy and kinetics of Glargine in a critically ill population. This protocol is a first step towards developing a Glargine-based protocol for less acute wards. Ensuring robustness to variability in Glargine efficacy directly affects the performance and safety that can be obtained.

Hyperglycaemia is a common complication of prematurity and stress in neonatal intensive care units (NICUs). It has been linked to worsened outcomes and mortality. There is currently no universally accepted best practice glycaemic control method, with many protocols lacking patient specificity or relying heavily on ad hoc clinical judgment from clinical staff who may be caring or overseeing care for several patients at once. The result is persistent hypoglycaemia and poor control. This research presents the virtual trial design and optimisation of a stochastic targeted (STAR) approach to improve performance and reduce hypoglycaemia. Clinically validated virtual trials based on NICU patient data (N = 61 patients, 7006 h) are used to develop and optimise a STAR protocol that improves on current STAR-NICU performance and reduce hypoglycaemia. Five approaches are used to maximise the stochastic range of BG outcomes within 4.0–8.0 mmol/L, and are designed based on an overall cohort risk to provide clinically specified risk (5%) of BG above or below a clinically specified level. The best protocol placed the 5th percentile BG outcome for an intervention on 4.0 mmol/L band. The optimised protocol increased %BG in the 4.0–8.0 mmol/L band by 3.5% and the incidence of BG

Though well studied, acute respiratory distress syndrome (ARDS) is still not fully understood. Mechanical ventilation (MV) has been a key treatment for ARDS. However, the optimal ventilation settings of basic MV parameters are still significantly debated. Only recently, were low tidal volumes shown to have lower mortality rates among ARDS patients. Despite over two decades of study, no standardisation of therapy or approach to MV appears on the horizon. This problem is likely due to the heterogeneity of the ARDS patient and ARDS affected lung. Currently, external MV parameters are set to try and treat an internal condition. There is no way to determine if more harm than good is being done. Hence, there is a tradeoff in between the risk of and benefit. What is required a method to assess that tradeoff and thus the potential risk. The use of positive end expiratory pressure (PEEP) and tidal volume has been identified as key ventilation parameters when treating ARDS patients. Although the impact of both parameters has been studied extensively, only the use of low tidal volumes has been conclusively determined. In contrast, the application of PEEP is still widely disputed. This review discusses two different approaches to ventilation management and the impact on optimal PEEP. The first approach examines the use of imaging techniques to determine regional lung mechanics. In the past, computed tomography (CT) was seen as a way to optimise PEEP, but the risks associated with it have limited it to a research tool. Newer methods such as lung ultrasound and electrical impedance tomography (EIT) seem to offer a less riskier approach to assessing regional mechanics. The second approach examines model-based approaches to ventilation management. Models that take ventilation data and depict a physical picture offer the potential to assess the risks on a patient-specific basis. Models offer the benefit of creating an approach to a highly heterogeneously and patient-specific problem in a non-invasive manner. Given the added dynamic of a patient's evolution over time, a highly patient-specific approach is typical and also what is required. Although both approaches can potentially be used to help with clinical decision making with regard to PEEP, they both pose advantages and disadvantages. The use of a given approach will depend on the individual needs of each clinic. Although not currently deployed in the clinic, model-based methods represent a novel methodology in treating ARDS patients. Thus, model-based approaches represent a “state of possible” rather than currently practiced methods, and require further clinical validation before justifying their use in the clinic.

Tight glycemic control has been shown to reduce mortality by 29–45% in critical care. Targeted glycemic control in critical care patients can be achieved by frequent fitting and prediction of a patient's modelled insulin sensitivity index, S I . This parameter can vary significantly in the critically ill due to the evolution of their condition and drug therapy. A three-dimensional stochastic model of S I variability is constructed using 18 long-term retrospective critical care patients’ data. Given S I for an hour, the stochastic model returns the probability distribution of S I for the next hour. Consequently, the resulting glycemic distribution 1 h following a known insulin and/or nutrition intervention can be derived. Knowledge of this distribution enables more accurate predictions for glycemic control with pre-determined likelihood based on confidence intervals. Clinical control data from eight independent critical care glycemic control trials were re-evaluated using the stochastic model. The stochastic model successfully captures the identified S I variation trend, accounting for 84% of measurements over time within the 0.90 confidence band, and 45% with a 0.50 confidence. Incorporating the stochastic model into the numerical glucose–insulin dynamics model, a virtual cohort was generated, imitating typical glucose–insulin dynamics in a critically ill population. Control trial simulations on this virtual cohort showed that the 0.90 confidence intervals cover 88% of measurements, and the 0.5 confidence intervals cover 46%. These results indicate that the stochastic model provides first order estimate of insulin sensitivity, S I , variation and resulting glycemic variation in critical care.