Next-generation, personalised, model-based critical care medicine: a state-of-the art review of in silico virtual patient models, methods, and cohorts, and how to validation them

Envisioning the Future of Personalized Medicine: Role and Realities of Digital Twins

Digital twins have emerged as a groundbreaking concept in personalized medicine, offering immense potential to transform health care delivery and improve patient outcomes. It is important to highlight the impact of digital twins on personalized medicine across the understanding of patient health, risk assessment, clinical trials and drug development, and patient monitoring. By mirroring individual health profiles, digital twins offer unparalleled insights into patient-specific conditions, enabling more accurate risk assessments and tailored interventions. However, their application extends beyond clinical benefits, prompting significant ethical debates over data privacy, consent, and potential biases in health care. The rapid evolution of this technology necessitates a careful balancing act between innovation and ethical responsibility. As the field of personalized medicine continues to evolve, digital twins hold tremendous promise in transforming health care delivery and revolutionizing patient care. While challenges exist, the continued development and integration of digital twins hold the potential to revolutionize personalized medicine, ushering in an era of tailored treatments and improved patient well-being. Digital twins can assist in recognizing trends and indicators that might signal the presence of diseases or forecast the likelihood of developing specific medical conditions, along with the progression of such diseases. Nevertheless, the use of human digital twins gives rise to ethical dilemmas related to informed consent, data ownership, and the potential for discrimination based on health profiles. There is a critical need for robust guidelines and regulations to navigate these challenges, ensuring that the pursuit of advanced health care solutions does not compromise patient rights and well-being. This viewpoint aims to ignite a comprehensive dialogue on the responsible integration of digital twins in medicine, advocating for a future where technology serves as a cornerstone for personalized, ethical, and effective patient care.

Can Machine Learning Personalize Cardiovascular Therapy in Sepsis?

Large randomized trials in sepsis have generally failed to find effective novel treatments. This is increasingly attributed to patient heterogeneity, including heterogeneous cardiovascular changes in septic shock. We discuss the potential for machine learning systems to personalize cardiovascular resuscitation in sepsis. While the literature is replete with proofs of concept, the technological readiness of current systems is low, with a paucity of clinical trials and proven patient benefit. Systems may be vulnerable to confounding and poor generalization to new patient populations or contemporary patterns of care. Typical electronic health records do not capture rich enough data, at sufficient temporal resolution, to produce systems that make actionable treatment suggestions. To resolve these issues, we recommend a simultaneous focus on technical challenges and removing barriers to translation. This will involve improving data quality, adopting causally grounded models, prioritizing safety assessment and integration into healthcare workflows, conducting randomized clinical trials and aligning with regulatory requirements.

Preserving multi-dimensional information: A hypersphere method for parameter space analysis

Background

Physiological modelling often involves models described by large numbers of variables and significant volumes of clinical data. Mathematical interpretation of such models frequently necessitates analysing data points in high-dimensional spaces. Existing algorithms for analysing high-dimensional points either lose important dimensionality or do not describe the full position of points. Hence, there is a need for an algorithm which preserves this information.

Methods

The most-distant uncovered point (MDUP) hypersphere method is a binary classification approach which defines a collection of equidistant N-dimensional points as the union of hyperspheres. The method iteratively generates hyperspheres at the most distant point in the interest region not yet contained within any hypersphere, until the entire region of interest is defined by the union of all generated hyperspheres. This method is tested on a 7-dimensional space with up to 35.8 million points representing feasible and infeasible spaces of model parameters for a clinically validated cardiovascular system model.

Results

For different numbers of input points, the MDUP hypersphere method tends to generate large spheres away from the boundary of feasible and infeasible points, but generates the greatest number of relatively much smaller spheres around the boundary of the region of interest to fill this space. Runtime scales quadratically, in part because the current MDUP implementation is not parallelised.

Conclusions

The MDUP hypersphere method can define points in a space of any dimension using only a collection of centre points and associated radii, making the results easily interpretable. It can identify large continuous regions, and in many cases capture the general structure of a region in only a relative few hyperspheres. The MDUP method also shows promise for initialising optimisation algorithm starting conditions within pre-defined feasible regions of model parameter spaces, which could improve model identifiability and the quality of optimisation results.

Obstructive respiratory disease simulation device

Respiratory disease is a major contributor to healthcare costs, as well as increasing morbidity and early mortality. The device presented is used to simulate the effects of Chronic Obstructive Pulmonary Disease (COPD) in healthy people. The intended use is to provide data equivalent to COPD data measured from those who are ill for initial validation of respiratory mechanics models. It would thus eliminate the need to test unhealthy and/or fragile subjects, or the need for invasive or costly equipment based test methods. The device is used in conjunction with an open-access venturi-based flow sensor, to measure pressure, flow, and breath tidal volume. The device simulates the pressure and flow profiles of a person who has COPD including the non-linear increased resistance to end-exhalation and gas trapping. To achieve this non-linearity, a combination of high and low resistance outlets is used. Thus, the simulator allows the collection of patient-specific COPD-like breathing data in a non-invasive manner from healthy subjects. The device is low-cost with the majority of the parts 3D printed using a Prusa mini 3D printer and PLA filament.

Pulmonary response prediction through personalized basis functions in a virtual patient model

BACKGROUND AND OBJECTIVE

Recruitment maneuvers with subsequent positive-end-expiratory-pressure (PEEP) have proven effective in recruiting lung volume and preventing alveoli collapse. However, determining a safe, effective, and patient-specific PEEP is not standardized, and this more optimal PEEP level evolves with patient condition, requiring personalised monitoring and care approaches to maintain optimal ventilation settings.

METHODS

This research examines 3 physiologically relevant basis function sets (exponential, parabolic, cumulative) to enable better prediction of elastance evolution for a virtual patient or digital twin model of MV lung mechanics, including novel elements to model and predict distension elastance. Prediction accuracy and robustness are validated against recruitment maneuver data from 18 volume-controlled ventilation (VCV) patients at 7 different baseline PEEP levels (0 to 12 cmH2O) and 14 pressure-controlled ventilation (PCV) patients at 4 different baseline PEEP levels (6 to 12 cmH2O), yielding 623 and 294 prediction cases, respectively. Predictions were made up to 12 cmH2O of added PEEP ahead, covering 6 × 2 cmH2O PEEP steps.

RESULTS

The 3 basis function sets yield median absolute peak inspiratory pressure (PIP) prediction error of 1.63 cmH2O for VCV patients, and median peak inspiratory volume (PIV) prediction error of 0.028 L for PCV patients. The exponential basis function set yields a better trade-off of overall performance across VCV and PCV prediction than parabolic and cumulative basis function sets from other studies. Comparing predicted and clinically measured distension prediction in VCV demonstrated consistent, robust high accuracy with R2 = 0.90-0.95.

CONCLUSIONS

The results demonstrate recruitment mechanics are best captured by an exponential basis function across different mechanical ventilation modes, matching physiological expectations, and accurately capture, for the first time, distension mechanics to within 5-10 % accuracy. Enabling the risk of lung injury to be predicted before changing ventilator settings. The overall outcomes significantly extend and more fully validate this digital twin or virtual mechanical ventilation patient model.

Development and usability testing of a patient digital twin for critical care education: a mixed methods study

Background

Digital twins are computerized patient replicas that allow clinical interventions testing in silico to minimize preventable patient harm. Our group has developed a novel application software utilizing a digital twin patient model based on electronic health record (EHR) variables to simulate clinical trajectories during the initial 6 h of critical illness. This study aimed to assess the usability, workload, and acceptance of the digital twin application as an educational tool in critical care.

Methods

A mixed methods study was conducted during seven user testing sessions of the digital twin application with thirty-five first-year internal medicine residents. Qualitative data were collected using a think-aloud and semi-structured interview format, while quantitative measurements included the System Usability Scale (SUS), NASA Task Load Index (NASA-TLX), and a short survey.

Results

Median SUS scores and NASA-TLX were 70 (IQR 62.5-82.5) and 29.2 (IQR 22.5-34.2), consistent with good software usability and low to moderate workload, respectively. Residents expressed interest in using the digital twin application for ICU rotations and identified five themes for software improvement: clinical fidelity, interface organization, learning experience, serious gaming, and implementation strategies.

Conclusion

A digital twin application based on EHR clinical variables showed good usability and high acceptance for critical care education.

Impact of super-resolution and image acquisition on the detection of calcifications in digital breast tomosynthesis

OBJECTIVES

A virtual clinical trial (VCT) method is proposed to determine the limit of calcification detection in tomosynthesis.

METHODS

Breast anatomy, focal findings, image acquisition, and interpretation (n = 14 readers) were simulated using screening data (n = 660 patients). Calcifications (0.2-0.4 mm3) were inserted into virtual breast phantoms. Digital breast tomosynthesis (DBT) acquisitions were simulated assuming various acquisition geometries: source motion (continuous and step-and-shoot), detector element size (140 and 70 µm), and reconstructed voxel size (35-140 µm). VCT results were estimated using multiple-reader multiple-case analyses and d' statistics. Signal-to-noise (SNR) analyses were also performed using BR3D phantoms.

RESULTS

Source motion and reconstructed voxel size demonstrated significant changes in the performance of imaging systems. Acquisition geometries that use 70 µm reconstruction voxel size and step-and-shoot motion significantly improved calcification detection. Comparing 70 with 100 µm reconstructed voxel size for step-and-shoot, the ΔAUC was 0.0558 (0.0647) and d' ratio was 1.27 (1.29) for 140 µm (70 µm) detector element size. Comparing step-and-shoot with a continuous motion for a 70 µm reconstructed voxel size, the ΔAUC was 0.0863 (0.0434) and the d' ratio was 1.40 (1.19) for 140 µm (70 µm) detector element. Small detector element sizes (e.g., 70 µm) did not significantly improve detection. The SNR results with the BR3D phantom show that calcification detection is dependent upon reconstructed voxel size and detector element size, supporting VCT results with comparable agreement (ratios: d' = 1.16 ± 0.11, SNR = 1.34 ± 0.13).

CONCLUSION

DBT acquisition geometries that use super-resolution (smaller reconstructed voxels than the detector element size) combined with step-and-shoot motion have the potential to improve the detection of calcifications.

CLINICAL RELEVANCE

Calcifications may not always be discernable in tomosynthesis because of differences in acquisition and reconstruction methods. VCTs can identify strategies to optimize acquisition and reconstruction parameters for calcification detection in tomosynthesis, most notably through super-resolution in the reconstruction.

KEY POINTS

• Super-resolution improves calcification detection and SNR in tomosynthesis; specifically, with the use of smaller reconstruction voxels. • Calcification detection using step-and-shoot motion is superior to that using continuous tube motion. • A detector element size of 70 µm does not provide better detection than 140 µm for small calcifications at the threshold of detectability.

Respiratory pressure and split flow data collection device with rapid occlusion attachment

Respiratory model-based methods require datasets containing enough dynamics to ensure model identifiability for development and validation. Rapid expiratory occlusion has been used to identify elastance and resistance within a single breath. Currently accepted practice for rapid expiratory occlusion involves a 100 ms occlusion of the expiratory pathway. This article presents a low-cost modular rapid shutter attachment to enable identification of passive respiratory mechanics. Shuttering faster than 100 ms creates rapid expiratory occlusion without the added dynamics of muscular response to shutter closure, by eliminating perceived expiratory blockage via high shutter speed. The shutter attachment fits onto a non-invasive venturi-based flow meter with separated inspiratory and expiratory pathways, established using one-way valves. Overall, these elements allow comprehensive collection of respiratory pressure and flow datasets with relatively very rapid expiratory occlusion.

Virtual patient with temporal evolution for mechanical ventilation trial studies: A stochastic model approach

BACKGROUND AND OBJECTIVE

Healthcare datasets are plagued by issues of data scarcity and class imbalance. Clinically validated virtual patient (VP) models can provide accurate in-silico representations of real patients and thus a means for synthetic data generation in hospital critical care settings. This research presents a realistic, time-varying mechanically ventilated respiratory failure VP profile synthesised using a stochastic model.

METHODS

A stochastic model was developed using respiratory elastance (Ers) data from two clinical cohorts and averaged over 30-minute time intervals. The stochastic model was used to generate future Ers data based on current Ers values with added normally distributed random noise. Self-validation of the VPs was performed via Monte Carlo simulation and retrospective Ers profile fitting. A stochastic VP cohort of temporal Ers evolution was synthesised and then compared to an independent retrospective patient cohort data in a virtual trial across several measured patient responses, where similarity of profiles validates the realism of stochastic model generated VP profiles.

RESULTS

A total of 120,000 3-hour VPs for pressure control (PC) and volume control (VC) ventilation modes are generated using stochastic simulation. Optimisation of the stochastic simulation process yields an ideal noise percentage of 5-10% and simulation iteration of 200,000 iterations, allowing the simulation of a realistic and diverse set of Ers profiles. Results of self-validation show the retrospective Ers profiles were able to be recreated accurately with a mean squared error of only 0.099 [0.009-0.790]% for the PC cohort and 0.051 [0.030-0.126]% for the VC cohort. A virtual trial demonstrates the ability of the stochastic VP cohort to capture Ers trends within and beyond the retrospective patient cohort providing cohort-level validation.

CONCLUSION

VPs capable of temporal evolution demonstrate feasibility for use in designing, developing, and optimising bedside MV guidance protocols through in-silico simulation and validation. Overall, the temporal VPs developed using stochastic simulation alleviate the need for lengthy, resource intensive, high cost clinical trials, while facilitating statistically robust virtual trials, ultimately leading to improved patient care and outcomes in mechanical ventilation.

Classification-based deep neural network vs mixture density network models for insulin sensitivity prediction problem

Model-based glycemic control (GC) protocols are used to treat stress-induced hyperglycaemia in intensive care units (ICUs). The STAR (Stochastic-TARgeted) glycemic control protocol - used in clinical practice in several ICUs in New Zealand, Hungary, Belgium, and Malaysia - is a model-based GC protocol using a patient-specific, model-based insulin sensitivity to describe the patient's actual state. Two neural network based methods are defined in this study to predict the patient's insulin sensitivity parameter: a classification deep neural network and a Mixture Density Network based method. Treatment data from three different patient cohorts are used to train the network models. Accuracy of neural network predictions are compared with the current model- based predictions used to guide care. The prediction accuracy was found to be the same or better than the reference. The authors suggest that these methods may be a promising alternative in model-based clinical treatment for patient state prediction. Still, more research is needed to validate these findings, including in-silico simulations and clinical validation trials.

Continuous time recurrent neural networks: Overview and benchmarking at forecasting blood glucose in the intensive care unit

OBJECTIVE

Blood glucose measurements in the intensive care unit (ICU) are typically made at irregular intervals. This presents a challenge in choice of forecasting model. This article gives an overview of continuous time autoregressive recurrent neural networks (CTRNNs) and evaluates how they compare to autoregressive gradient boosted trees (GBT) in forecasting blood glucose in the ICU.

METHODS

Continuous time autoregressive recurrent neural networks (CTRNNs) are a deep learning model that account for irregular observations through incorporating continuous evolution of the hidden states between observations. This is achieved using a neural ordinary differential equation (ODE) or neural flow layer. In this manuscript, we give an overview of these models, including the varying architectures that have been proposed to account for issues such as ongoing medical interventions. Further, we demonstrate the application of these models to probabilistic forecasting of blood glucose in a critical care setting using electronic medical record and simulated data and compare with GBT and linear models.

RESULTS

The experiments confirm that addition of a neural ODE or neural flow layer generally improves the performance of autoregressive recurrent neural networks in the irregular measurement setting. However, several CTRNN architecture are outperformed by a GBT model (Catboost), with only a long short-term memory (LSTM) and neural ODE based architecture (ODE-LSTM) achieving comparable performance on probabilistic forecasting metrics such as the continuous ranked probability score (ODE-LSTM: 0.118 ± 0.001; Catboost: 0.118 ± 0.001), ignorance score (0.152 ± 0.008; 0.149 ± 0.002) and interval score (175 ± 1; 176 ± 1).

CONCLUSION

The application of deep learning methods for forecasting in situations with irregularly measured time series such as blood glucose shows promise. However, appropriate benchmarking by methods such as GBT approaches (plus feature transformation) are key in highlighting whether novel methodologies are truly state of the art in tabular data settings.

How Electronic Medical Record Integration Can Support More Efficient Critical Care Clinical Trials

Large volumes of data are collected on critically ill patients, and using data science to extract information from the electronic medical record (EMR) and to inform the design of clinical trials represents a new opportunity in critical care research. Using improved methods of phenotyping critical illnesses, subject identification and enrollment, and targeted treatment group assignment alongside newer trial designs such as adaptive platform trials can increase efficiency while lowering costs. Some tools such as the EMR to automate data collection are already in use. Refinement of data science approaches in critical illness research will allow for better clinical trials and, ultimately, improved patient outcomes.

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.

An in vitro model to study suction events by a ventricular assist device: validation with clinical data

Introduction: Ventricular assist devices (LVADs) are a valuable therapy for end-stage heart failure patients. However, some adverse events still persist, such as suction that can trigger thrombus formation and cardiac rhythm disorders. The aim of this study is to validate a suction module (SM) as a test bench for LVAD suction detection and speed control algorithms. Methods: The SM consists of a latex tube, mimicking the ventricular apex, connected to a LVAD. The SM was implemented into a hybrid in vitro-in silico cardiovascular simulator. Suction was induced simulating hypovolemia in a profile of a dilated cardiomyopathy and of a restrictive cardiomyopathy for pump speeds ranging between 2,500 and 3,200 rpm. Clinical data collected in 38 LVAD patients were used for the validation. Clinical and simulated LVAD flow waveforms were visually compared. For a more quantitative validation, a binary classifier was used to classify simulated suction and non-suction beats. The obtained classification was then compared to that generated by the simulator to evaluate the specificity and sensitivity of the simulator. Finally, a statistical analysis was run on specific suction features (e.g., minimum impeller speed pulsatility, minimum slope of the estimated flow, and timing of the maximum slope of the estimated flow). Results: The simulator could reproduce most of the pump waveforms observed in vivo. The simulator showed a sensitivity and specificity and of 90.0% and 97.5%, respectively. Simulated suction features were in the interquartile range of clinical ones. Conclusions: The SM can be used to investigate suction in different pathophysiological conditions and to support the development of LVAD physiological controllers.

Airflow and dynamic circumference of abdomen and thorax for adults at varied continuous positive airway pressure ventilation settings and breath rates

Continuous positive airway pressure (CPAP) ventilation is a commonly prescribed respiratory therapy providing positive end-expiratory pressure (PEEP) to assist breathing and prevent airway collapse. Setting PEEP is highly debated and it is thus primarily titrated based on symptoms of excessive or insufficient support. However, titration periods are clinician intensive and can result in barotrauma or under-oxygenation during the process. Developing model-based methods to more efficiently personalise CPAP therapy based on patient-specific response requires clinical data of lung/CPAP interactions. To this end, a trial was conducted to establish a dataset of healthy subjects lung/CPAP interaction. Pressure, flow, and tidal volume were recorded alongside secondary measures of dynamic chest and abdominal circumference, to better validate model outcomes and assess breathing modes, muscular recruitment, and effort. N = 30 subjects (15 male; 15 female) were included. Self-reported asthmatics and smokers/vapers were included, offering a preliminary assessment of any potential differences in response to CPAP from lung stiffness changes in these scenarios. Additional demographics associated with lung function (sex, age, height, and weight) were also recorded.

Physiological Modeling of Hemodynamic Responses to Sodium Nitroprusside

BACKGROUND

Computational modeling of physiology has become a routine element in the development, evaluation, and safety testing of many types of medical devices. Members of the Food and Drug Administration have recently published a manuscript detailing the development, validation, and sensitivity testing of a computational model for blood volume, cardiac stroke volume, and blood pressure, noting that such a model might be useful in the development of closed-loop fluid administration systems. In the present study, we have expanded on this model to include the pharmacologic effect of sodium nitroprusside and calibrated the model against our previous experimental animal model data.

METHODS

Beginning with the model elements in the original publication, we added six new parameters to control the effect of sodium nitroprusside: two for the onset time and clearance rates, two for the stroke volume effect (which includes venodilation as a "hidden" element), and two for the direct effect on arterial blood pressure. Using this new model, we then calibrated the predictive performance against previously collected animal study data using nitroprusside infusions to simulate shock with the primary emphasis on MAP. Root-mean-squared error (RMSE) was calculated, and the performance was compared to the performance of the model in the original study.

RESULTS

RMSE of model-predicted MAP to actual MAP was lower than that reported in the original model, but higher for SV and CO. The individually fit models showed lower RMSE than using the population average values for parameters, suggesting the fitting process was effective in identifying improved parameters. Use of partially fit models after removal of the lowest variance population parameters showed a very minor decrement in improvement over the fully fit models.

CONCLUSION

The new model added the clinical effects of SNP and was successfully calibrated against experimental data with an RMSE of <10% for mean arterial pressure. Model-predicted MAP showed an error similar to that seen in the original base model when using fluid shifts, heart rate, and drug dose as model inputs.

Computational physiological models for individualised mechanical ventilation: a systematic literature review focussing on quality, availability, and clinical readiness

BACKGROUND

Individualised optimisation of mechanical ventilation (MV) remains cumbersome in modern intensive care medicine. Computerised, model-based support systems could help in tailoring MV settings to the complex interactions between MV and the individual patient's pathophysiology. Therefore, we critically appraised the current literature on computational physiological models (CPMs) for individualised MV in the ICU with a focus on quality, availability, and clinical readiness.

METHODS

A systematic literature search was conducted on 13 February 2023 in MEDLINE ALL, Embase, Scopus and Web of Science to identify original research articles describing CPMs for individualised MV in the ICU. The modelled physiological phenomena, clinical applications, and level of readiness were extracted. The quality of model design reporting and validation was assessed based on American Society of Mechanical Engineers (ASME) standards.

RESULTS

Out of 6,333 unique publications, 149 publications were included. CPMs emerged since the 1970s with increasing levels of readiness. A total of 131 articles (88%) modelled lung mechanics, mainly for lung-protective ventilation. Gas exchange (n = 38, 26%) and gas homeostasis (n = 36, 24%) models had mainly applications in controlling oxygenation and ventilation. Respiratory muscle function models for diaphragm-protective ventilation emerged recently (n = 3, 2%). Three randomised controlled trials were initiated, applying the Beacon and CURE Soft models for gas exchange and PEEP optimisation. Overall, model design and quality were reported unsatisfactory in 93% and 21% of the articles, respectively.

CONCLUSION

CPMs are advancing towards clinical application as an explainable tool to optimise individualised MV. To promote clinical application, dedicated standards for quality assessment and model reporting are essential. Trial registration number PROSPERO- CRD42022301715 . Registered 05 February, 2022.

Modelling patient specific cardiopulmonary interactions

Mechanical ventilation is well known for having detrimental effects on the cardiovascular system, particularly when using high positive end-expiratory pressure. High positive end-expiratory pressure levels cause a decrease in stroke volume, which, under normal conditions, usually bring about a decrease in stressed blood volume. Stressed blood volume, defined as the total pressure generating volume of the cardiovascular system, has been shown to be a potential index of fluid responsiveness, making it a potentially important diagnostic tool. Generally, respiratory and haemodynamic care are provided independently of one another. However, that positive end-expiratory pressure alters both stroke volume and stressed blood volume suggests both the pulmonary and cardiovascular state should be conjointly optimised and used to guide positive end-expiratory pressure. However, the complex and patient-specific nature of cardiopulmonary interactions which occur during mechanical ventilation presents a challenge for accurate modelling of respiratory and cardiovascular interactions required to better optimise care. Previous models attempting to incorporate cardiopulmonary interactions have suffered from poor reliability at higher PEEP levels, largely due to an exaggerated effect of intrathoracic pressure on the cardiovascular system. A new parameter, alpha, is added to a previously validated cardiopulmonary model, to modulate the percentage of intrathoracic pressure applied to the vena cava and left ventricle. The new parameter aims to increase reliability under high PEEP conditions as well as provide a patient specific solution to modelling cardiopulmonary interactions. The results from the identified optimal alpha are compared to the original model to investigate how this new parameter may be used to create a more patient-specific cardiopulmonary model, which would be better suited for guidance of care in the ICU.

Low-cost structured light imaging of regional volume changes for use in assessing mechanical ventilation

BACKGROUND

Optimal setting of mechanical ventilators is critical for improving outcomes. Accurate, predictive lung mechanics models are effective in optimizing MV settings, but only at a global level as they cannot estimate regional lung volume ventilation to assess the potential of local distension or under-ventilation. This study presents a low-cost structured light system for non-contact high resolution chest motion measurement to estimate regional lung volume changes.

METHODS

The system consists of a structured light projector and two cameras. A new pattern is designed to extract motion from sub-regions of the chest surface, and an efficient feature is proposed to provide a fast and accurate correspondence matching between two views. Reconstruction of 3D surface points is based on the matched points and stereo method. Asymmetric distribution of tidal volume into left and right lungs is estimated based on reconstructed regional chest expansion. A proof-of-concept experiment using a dummy model and two test lungs connected to a ventilator to provide differential chest expansion is conducted under tidal volumes of 400 ml, 500 ml and 600 ml, with results compared to the widely-used SURF and ORB methods.

RESULTS

Compared to the SURF and ORB methods, the proposed method is more computationally efficient with ∼40% less computational time cost, and higher accuracy for dense point correspondence. Finally, the proposed method estimated the region lung volumes with the maximum error of 8 ml under 600 ml tidal volume, indicating a good accuracy.

CONCLUSIONS

Surface reconstruction results in a proof-of-concept experiment with differential chest expansion show good performance for the proposed pattern and method in extracting the key information for regional chest expansion. The proposed method is generalizable, with potential for use in other applications.

Development of scalable lymphatic system in the 4D XCAT phantom: Application to quantitative evaluation of lymphoma PET segmentations

BACKGROUND

Digital anthropomorphic phantoms, such as the 4D extended cardiac-torso (XCAT) phantom, are actively used to develop, optimize, and evaluate a variety of imaging applications, allowing for realistic patient modeling and knowledge of ground truth. The XCAT phantom defines the activity and attenuation for a simulated patient, which includes a complete set of organs, muscle, bone, and soft tissue, while also accounting for cardiac and respiratory motion. However, the XCAT phantom does not currently include the lymphatic system, critical for evaluating medical imaging tasks such as sentinel node detection, node density measurement, and radiation dosimetry.

PURPOSE

In this study, we aimed to develop a scalable lymphatic system in the XCAT phantom, to facilitate improved research of the lymphatic system in medical imaging. Using this scalable lymphatic system, we modeled the lymph node conglomerate pathology that is characteristically observed in primary mediastinal B-cell lymphoma (PMBCL). As an extended application, we evaluated positron emission tomography (PET) image quantification of metabolic tumor volume (MTV) and total lesion glycolysis (TLG) of these simulated lymphomas, though the phantoms may be applied to other imaging modalities and study design paradigms (e.g., image quality, detection).

METHODS

A template model for the lymphatic system was developed based on anatomical data from the Visible Human Project of the National Library of Medicine. The segmented nodes and vessels were fit with non-uniform rational basis spline surfaces, and multichannel large deformation diffeomorphic metric mapping was used to propagate the template to different XCAT anatomies. To model conglomerates observed in PMBCL, lymph nodes were enlarged, converged within the mediastinum, and tracer concentration was increased. We used the phantoms as inputs to a PET simulation tool, which generated images using ordered subsets expectation maximization reconstruction with 2-8 mm Gaussian filters. Fixed thresholding (FT) and gradient segmentation were used to determine MTV and TLG. Percent bias (%Bias) and coefficient of variation (COV) were computed as measures of accuracy and precision, respectively, for each MTV and TLG measurement.

RESULTS

Using the methodology described above, we introduced a scalable lymphatic system in the XCAT phantom, which allows for the radioactivity and attenuation ground truth to be generated in 116 ± 2.5 s using a 2.3 GHz processor. Within the Rhinoceros interface, lymph node anatomy and function were modified to create a cohort of 10 phantoms with lymph node conglomerates. Using the lymphoma phantoms to evaluate PET quantification of MTV, mean %Bias values were -9.3%, -41.3%, and 20.9%, while COV values were 4.08%, 7.6%, and 3.4% using 25% FT, 40% FT, and gradient segmentations, respectively. Comparatively for TLG, mean %Bias values were -27.4%, -45.8%, and -16.0%, while COV values were 1.9%, 5.7%, and 1.4%, for the 25% FT, 40% FT, and gradient segmentations, respectively.

CONCLUSIONS

In this work, we upgraded the XCAT phantom to include a lymphatic system, comprised of a network of 276 scalable lymph nodes and corresponding vessels. As an application, we created a cohort of phantoms with lymph node conglomerates to evaluate lymphoma quantification in PET imaging, which highlights an important application of this work.

Virtual patient framework for the testing of mechanical ventilation airway pressure and flow settings protocol

BACKGROUND AND OBJECTIVE

Model-based and personalised decision support systems are emerging to guide mechanical ventilation (MV) treatment for respiratory failure patients. However, model-based treatments require resource-intensive clinical trials prior to implementation. This research presents a framework for generating virtual patients for testing model-based decision support, and direct use in MV treatment.

METHODS

The virtual MV patient framework consists of 3 stages: 1) Virtual patient generation, 2) Patient-level validation, and 3) Virtual clinical trials. The virtual patients are generated from retrospective MV patient data using a clinically validated respiratory mechanics model whose respiratory parameters (respiratory elastance and resistance) capture patient-specific pulmonary conditions and responses to MV care over time. Patient-level validation compares the predicted responses from the virtual patient to their retrospective results for clinically implemented MV settings and changes to care. Patient-level validated virtual patients create a platform to conduct virtual trials, where the safety of closed-loop model-based protocols can be evaluated.

RESULTS

This research creates and presents a virtual patient platform of 100 virtual patients generated from retrospective data. Patient-level validation reported median errors of 3.26% for volume-control and 6.80% for pressure-control ventilation mode. A virtual trial on a model-based protocol demonstrates the potential efficacy of using virtual patients for prospective evaluation and testing of the protocol.

CONCLUSION

The virtual patient framework shows the potential to safely and rapidly design, develop, and optimise new model-based MV decision support systems and protocols using clinically validated models and computer simulation, which could ultimately improve patient care and outcomes in MV.

Whole-lung finite-element models for mechanical ventilation and respiratory research applications

Mechanical ventilation has been a vital treatment for Covid-19 patients with respiratory failure. Lungs assisted with mechanical ventilators present a wide variability in their response that strongly depends on air-tissue interactions, which motivates the creation of simulation tools to enhance the design of ventilatory protocols. In this work, we aim to create anatomical computational models of the lungs that predict clinically-relevant respiratory variables. To this end, we formulate a continuum poromechanical framework that seamlessly accounts for the air-tissue interaction in the lung parenchyma. Based on this formulation, we construct anatomical finite-element models of the human lungs from computed-tomography images. We simulate the 3D response of lungs connected to mechanical ventilation, from which we recover physiological parameters of high clinical relevance. In particular, we provide a framework to estimate respiratory-system compliance and resistance from continuum lung dynamic simulations. We further study our computational framework in the simulation of the supersyringe method to construct pressure-volume curves. In addition, we run these simulations using several state-of-the-art lung tissue models to understand how the choice of constitutive models impacts the whole-organ mechanical response. We show that the proposed lung model predicts physiological variables, such as airway pressure, flow and volume, that capture many distinctive features observed in mechanical ventilation and the supersyringe method. We further conclude that some constitutive lung tissue models may not adequately capture the physiological behavior of lungs, as measured in terms of lung respiratory-system compliance. Our findings constitute a proof of concept that finite-element poromechanical models of the lungs can be predictive of clinically-relevant variables in respiratory medicine.

Computational cardiac physiology for new modelers: Origins, foundations, and future

Mathematical models of the cardiovascular system have come a long way since they were first introduced in the early 19th century. Driven by a rapid development of experimental techniques, numerical methods, and computer hardware, detailed models that describe physical scales from the molecular level up to organs and organ systems have been derived and used for physiological research. Mathematical and computational models can be seen as condensed and quantitative formulations of extensive physiological knowledge and are used for formulating and testing hypotheses, interpreting and directing experimental research, and have contributed substantially to our understanding of cardiovascular physiology. However, in spite of the strengths of mathematics to precisely describe complex relationships and the obvious need for the mathematical and computational models to be informed by experimental data, there still exist considerable barriers between experimental and computational physiological research. In this review, we present a historical overview of the development of mathematical and computational models in cardiovascular physiology, including the current state of the art. We further argue why a tighter integration is needed between experimental and computational scientists in physiology, and point out important obstacles and challenges that must be overcome in order to fully realize the synergy of experimental and computational physiological research.

Estimating Enhanced Endogenous Glucose Production in Intensive Care Unit Patients with Severe Insulin Resistance

BACKGROUND

Critically ill ICU patients frequently experience acute insulin resistance and increased endogenous glucose production, manifesting as stress-induced hyperglycemia and hyperinsulinemia. STAR (Stochastic TARgeted) is a glycemic control protocol, which directly manages inter- and intra- patient variability using model-based insulin sensitivity (SI). The model behind STAR assumes a population constant for endogenous glucose production (EGP), which is not otherwise identifiable.

OBJECTIVE

This study analyses the effect of estimating EGP for ICU patients with very low SI (severe insulin resistance) and its impact on identified, model-based insulin sensitivity identification, modeling accuracy, and model-based glycemic clinical control.

METHODS

Using clinical data from 717 STAR patients in 3 independent cohorts (Hungary, New Zealand, and Malaysia), insulin sensitivity, time of insulin resistance, and EGP values are analyzed. A method is presented to estimate EGP in the presence of non-physiologically low SI. Performance is assessed via model accuracy.

RESULTS

Results show 22%-62% of patients experience 1+ episodes of severe insulin resistance, representing 0.87%-9.00% of hours. Episodes primarily occur in the first 24 h, matching clinical expectations. The Malaysian cohort is most affected. In this subset of hours, constant model-based EGP values can bias identified SI and increase blood glucose (BG) fitting error. Using the EGP estimation method presented in these constrained hours significantly reduced BG fitting errors.

CONCLUSIONS

Patients early in ICU stay may have significantly increased EGP. Increasing modeled EGP in model-based glycemic control can improve control accuracy in these hours. The results provide new insight into the frequency and level of significantly increased EGP in critical illness.

Classification of clinically relevant intravascular volume status using point of care ultrasound and machine learning

Purpose

This is a foundational study in which multiorgan system point of care ultrasound (POCUS) and machine learning (ML) are used to mimic physician management decisions regarding the functional intravascular volume status (IVS) and need for diuretic therapy. We present this as an impactful use case of an application of ML in aided decision making for clinical practice. IVS represents complex physiologic interactions of the cardiac, renal, pulmonary, and other organ systems. In particular, we focus on vascular congestion and overload as an evolving concept in POCUS diagnosis and clinical relevance. It is critical for physicians to be able to evaluate IVS without disrupting workflow or exposing patients to unnecessary testing, radiation, or cost. This work utilized a small retrospective dataset as a feasibility test for ML binary classification of diuretic administration validated with clinical decision data. Future work will be directed toward artificial intelligence (AI) delivery at the bedside and assessment of the impact on patient-centered outcomes and physician workflow improvement.

Approach

We retrospectively reviewed and processed 1039 POCUS video clips, including cardiac, thoracic, and inferior vena cava (IVC) views. Multiorgan POCUS clips were correlated with clinical data extracted from the electronic health record and deidentified for algorithm training and validation. We implemented a two-stream three-dimensional (3D) deep learning approach that fuses heart and IVC data to perform binary classification of the need for diuretic use.

Results

Our proposed approach achieves high classification accuracy (84%) for the determination of diuretic use with 0.84 area under the receiver operating characteristic curve.

Conclusions

Our two-stream 3D deep neural network is able to classify POCUS video clips that match physicians' classification for or against diuretic use with high accuracy. This serves as a foundational step in the progress toward AI-aided diagnosis and AI implementation in the field of IVS evaluation by POCUS.

Evaluating Translational Methods for Personalized Medicine—A Scoping Review

The introduction of personalized medicine, through the increasing multi-omics characterization of disease, brings new challenges to disease modeling. The scope of this review was a broad evaluation of the relevance, validity, and predictive value of the current preclinical methodologies applied in stratified medicine approaches. Two case models were chosen: oncology and brain disorders. We conducted a scoping review, following the Joanna Briggs Institute guidelines, and searched PubMed, EMBASE, and relevant databases for reports describing preclinical models applied in personalized medicine approaches. A total of 1292 and 1516 records were identified from the oncology and brain disorders search, respectively. Quantitative and qualitative synthesis was performed on a final total of 63 oncology and 94 brain disorder studies. The complexity of personalized approaches highlights the need for more sophisticated biological systems to assess the integrated mechanisms of response. Despite the progress in developing innovative and complex preclinical model systems, the currently available methods need to be further developed and validated before their potential in personalized medicine endeavors can be realized. More importantly, we identified underlying gaps in preclinical research relating to the relevance of experimental models, quality assessment practices, reporting, regulation, and a gap between preclinical and clinical research. To achieve a broad implementation of predictive translational models in personalized medicine, these fundamental deficits must be addressed.

Gene expression profiling and protein-protein interaction analysis reveals the dynamic role of MCM7 in Alzheimer's disorder and breast cancer

The interrelation of cancer and Alzheimer's disorder (AD)-associated molecular mechanisms, reported last decade, paved the path for drug discoveries. In this direction, while chemotherapy is well established for breast cancer (BC), the detection and targeted therapy for AD is not advanced due to a lack of recognized peripheral biomarkers. The present study aimed to find diagnostic and prognostic molecular signature markers common to both BC and AD for possible drug targeting and repurposing. For these disorders, two corresponding microarray datasets (GSE42568, GSE33000) were used for identifying the differentially expressed genes (DEGs), resulting in recognition of CD209 and MCM7 as the two common players. While the CD209 gene was upregulated in both disorders and has been studied vastly, the MCM7 gene showed a strikingly reverse pattern of expression level, downregulated in the case of BC while upregulated in the case of AD. Thus, the MCM7 gene was further analyzed for expression, predictions, and validations of its structure and protein-protein interaction (PPI) for the possible development of new treatment methods for AD. The study concluded with indicative drug repurposing studies to check the effect of existing clinically approved drugs for BC for rectifying the expression levels of the mutated MCM7 gene in AD.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13205-022-03207-1.

Physiological trend analysis of a novel cardio-pulmonary model during a preload reduction manoeuvre

BACKGROUND AND OBJECTIVE

Mechanical ventilation causes adverse effects on the cardiovascular system. However, the exact nature of the effects on haemodynamic parameters is not fully understood. A recently developed cardio-vascular system model which incorporates cardio-pulmonary interactions is compared to the original 3-chamber cardiovascular model to investigate the exact effects of mechanical ventilation on haemodynamic parameters and to assess the trade-off of model complexity and model reliability between the 2 models.

METHODS

Both the cardio-pulmonary and three chamber models are used to identify cardiovascular system parameters from aortic pressure, left ventricular volume, airway flow and airway pressure measurements from 4 pigs during a preload reduction manoeuvre. Outputs and parameter estimations from both models are contrasted to assess the relative performance of each model and to further investigate the effects of mechanical ventilation on haemodynamic parameters.

RESULTS

Both models tracked measurements accurately as expected. There was no identifiable increase in error from the added complexity of the cardio-pulmonary model, with both models having a mean average error below 0.5% for all pigs. Identified left ventricle and vena cava elastances of the 3-chamber model was found to diverge exponentially with PEEP from identified left ventricle and vena cava elastances of the cardio-pulmonary model. The r² of the fit for each pig ranged from 0.888 to 0.998 for left ventricle elastance divergence and from 0.905 to 0.999 for vena cava elastance divergence. All other identified parameters showed no significant difference between models.

CONCLUSIONS

Despite the increase in model complexity, there was no loss in the cardio-pulmonary model's ability to accurately estimate haemodynamic parameters and reproduce system dynamics. Furthermore, the cardio-pulmonary model was able to demonstrate how mechanical ventilation affected parameter estimations as PEEP was increased. The 3-chamber model was shown to produce parameter estimations which diverged exponentially with PEEP, while the cardiopulmonary model estimations remained more stable, suggesting its ability to produce more physiologically accurate parameter estimations under higher PEEP conditions.

Estimating the incidence of spontaneous breathing effort of mechanically ventilated patients using a non-linear auto regressive (NARX) model

BACKGROUND AND OBJECTIVE

Mechanical ventilation (MV) provides breathing support for acute respiratory distress syndrome (ARDS) patients in the intensive care unit, but is difficult to optimize. Too much, or too little of pressure or volume support can cause further ventilator-induced lung injury, increasing length of MV, cost and mortality. Patient-specific respiratory mechanics can help optimize MV settings. However, model-based estimation of respiratory mechanics is less accurate when patient exhibit un-modeled spontaneous breathing (SB) efforts on top of ventilator support. This study aims to estimate and quantify SB efforts by reconstructing the unaltered passive mechanics airway pressure using NARX model.

METHODS

Non-linear autoregressive (NARX) model is used to reconstruct missing airway pressure due to the presence of spontaneous breathing effort in mv patients. Then, the incidence of SB patients is estimated. The study uses a total of 10,000 breathing cycles collected from 10 ARDS patients from IIUM Hospital in Kuantan, Malaysia. In this study, there are 2 different ratios of training and validating methods. Firstly, the initial ratio used is 60:40 which indicates 600 breath cycles for training and remaining 400 breath cycles used for testing. Then, the ratio is varied using 70:30 ratio for training and testing data.

RESULTS AND DISCUSSION

The mean residual error between original airway pressure and reconstructed airway pressure is denoted as the magnitude of effort. The median and interquartile range of mean residual error for both ratio are 0.0557 [0.0230 - 0.0874] and 0.0534 [0.0219 - 0.0870] respectively for all patients. The results also show that Patient 2 has the highest percentage of SB incidence and Patient 10 with the lowest percentage of SB incidence which proved that NARX model is able to perform for both higher incidence of SB effort or when there is a lack of SB effort.

CONCLUSION

This model is able to produce the SB incidence rate based on 10% threshold. Hence, the proposed NARX model is potentially useful to estimate and identify patient-specific SB effort, which has the potential to further assist clinical decisions and optimize MV settings.

Virtual and Artificial Cardiorespiratory Patients in Medicine and Biomedical Engineering

Recently, ‘medicine in silico’ has been strongly encouraged due to ethical and legal limitations related to animal experiments and investigations conducted on patients. Computer models, particularly the very complex ones (virtual patients—VP), can be used in medical education and biomedical research as well as in clinical applications. Simpler patient-specific models may aid medical procedures. However, computer models are unfit for medical devices testing. Hybrid (i.e., numerical–physical) models do not have this disadvantage. In this review, the chosen approach to the cardiovascular system and/or respiratory system modeling was discussed with particular emphasis given to the hybrid cardiopulmonary simulator (the artificial patient), that was elaborated by the authors. The VP is useful in the education of forced spirometry, investigations of cardiopulmonary interactions (including gas exchange) and its influence on pulmonary resistance during artificial ventilation, and explanation of phenomena observed during thoracentesis. The artificial patient is useful, inter alia, in staff training and education, investigations of cardiorespiratory support and the testing of several medical devices, such as ventricular assist devices and a membrane-based artificial heart.

Low cost circulatory pressure acquisition and fluid infusion rate measurement system for clinical research

Acquiring patient physiological waveforms is useful for studying hemodynamic management and developing medical monitoring systems. A low cost, Arduino controlled data acquisition system acquires arterial pressure waveforms (Edwards Lifesciences TruWave compatible) and measures fluid infusion rate using hanging scales. This system can be used at the same time as a clinical monitor, enabling recording of patient arterial pressure and fluid delivery for clinical research. The system is powered via a USB connection, which additionally provides serial output, aiding compatibility and customisation. A simple software user interface, developed in Python, shows outputs. Each data acquisition system, including all necessary connection cables costs ~US$90 and is multiple-use.

SASC: A simple approach to synthetic cohorts for generating longitudinal observational patient cohorts from COVID-19 clinical data

One of the impacts of the coronavirus disease 2019 (COVID-19) pandemic has been a push for researchers to better exploit synthetic data and accelerate the design, analysis, and modeling of clinical trials. The unprecedented clinical efforts caused by COVID-19’s emergence will certainly boost future robust and innovative approaches of statistical sciences applied to clinical fields. Here, we report the development of SASC, a simple but efficient approach to generate COVID-19-related synthetic clinical data through a web application. SASC takes basic summary statistics for each group of patients and attempts to generate single variables according to internal correlations. To assess the “reliability” of the results, statistical comparisons with Synthea, a known synthetic patient generator tool, and, more importantly, with clinical data of real COVID-19 patients are provided. The source code and web application are available on GitHub, Zenodo, and Mendeley Data.

•

From summaries of grouped variables, it is possible to generate synthetic data

•

Similarity between real and virtual cohorts visualized with Kaplan-Meyer plots

•

Shiny app for tuning and downloading the synthetic cohort provided to users

•

SASC performances and result accuracies are similar to or better than Synthea results

Virtual cohorts built on synthetic data have received attention lately as they can change the way we work with clinical data. Generating synthetic data very close to "real-world" data without having ethical restrictions can considerably help data analysts in several aspects such as didactic efforts, statistical modeling, and assessments of interventions. We developed SASC, a tool that uses summary statistics of real clinical data to produce synthetic data having the same or a very close summary to the reference variables. SASC displays a performance comparable with, or superior to, that of other virtual cohort tools like Synthea and allows users to optimize the synthetic dataset produced through a Shiny visualizer application interactively.

The unmet potential of artificial intelligence in veterinary medicine

Veterinary medicine is a broad and growing discipline that includes topics such as companion animal health, population medicine and zoonotic diseases, and agriculture. In this article, we provide insight on how artificial intelligence works and how it is currently applied in veterinary medicine. We also discuss its potential in veterinary medicine. Given the rapid pace of research and commercial product developments in this area, the next several years will pose challenges to understanding, interpreting, and adopting this powerful and evolving technology. Artificial intelligence has the potential to enable veterinarians to perform tasks more efficiently while providing new insights for the management and treatment of disorders. It is our hope that this will translate to better quality of life for animals and those who care for them.

Reconstructing asynchrony for mechanical ventilation using a hysteresis loop virtual patient model

BACKGROUND

Patient-specific lung mechanics during mechanical ventilation (MV) can be identified from measured waveforms of fully ventilated, sedated patients. However, asynchrony due to spontaneous breathing (SB) effort can be common, altering these waveforms and reducing the accuracy of identified, model-based, and patient-specific lung mechanics.

METHODS

Changes in patient-specific lung elastance over a pressure-volume (PV) loop, identified using hysteresis loop analysis (HLA), are used to detect the occurrence of asynchrony and identify its type and pattern. The identified HLA parameters are then combined with a nonlinear mechanics hysteresis loop model (HLM) to extract and reconstruct ventilated waveforms unaffected by asynchronous breaths. Asynchrony magnitude can then be quantified using an energy-dissipation metric, Easyn, comparing PV loop area between model-reconstructed and original, altered asynchronous breathing cycles. Performance is evaluated using both test-lung experimental data with a known ground truth and clinical data from four patients with varying levels of asynchrony.

RESULTS

Root mean square errors for reconstructed PV loops are within 5% for test-lung experimental data, and 10% for over 90% of clinical data. Easyn clearly matches known asynchrony magnitude for experimental data with RMS errors < 4.1%. Clinical data performance shows 57% breaths having Easyn > 50% for Patient 1 and 13% for Patient 2. Patient 3 only presents 20% breaths with Easyn > 10%. Patient 4 has Easyn = 0 for 96% breaths showing accuracy in a case without asynchrony.

CONCLUSIONS

Experimental test-lung validation demonstrates the method's reconstruction accuracy and generality in controlled scenarios. Clinical validation matches direct observations of asynchrony in incidence and quantifies magnitude, including cases without asynchrony, validating its robustness and potential efficacy as a clinical real-time asynchrony monitoring tool.

Nutritional therapy in critically ill patients with diabetes

PURPOSE OF REVIEW

There has been a significant increase in nutrition therapy related studies within the critical care cohort in recent years. Management of patients with both diabetes and stress hyperglycaemia through targeted nutrition interventions is no exception. The aim of this review is to outline current available diabetes specific nutrition formula, its impact on gastric emptying and subsequently glycaemic control as well as explore recent literature on the efficacy of utilizing nutrition support to optimize glycaemic control in critically ill patients.

RECENT FINDINGS

Studies explored within this review were similar in terms of outcomes measures, focusing primarily on insulin use and glycaemic control. Although there were promising results in terms of the impact of diabetes-specific nutrition formula on these outcome measures, there were no significant associations with clinical outcomes.

SUMMARY

The use of diabetes-specific formulae in critically ill patients with pre-existing diabetes and stress hyperglycaemia can be considered a logical approach to minimize the risks associated with high doses of insulin. Additional research is required to address the effects of these formulae on the dysglycaemia, nursing workload, safety of glycaemic control and cost-effectiveness.

Stochastic integrated model-based protocol for volume-controlled ventilation setting

BACKGROUND AND OBJECTIVE

Mechanical ventilation (MV) is the primary form of care for respiratory failure patients. MV settings are based on general clinical guidelines, intuition, and experience. This approach is not patient-specific and patients may thus experience suboptimal, potentially harmful MV care. This study presents the Stochastic integrated VENT (SiVENT) protocol which combines model-based approaches of the VENT protocol from previous works, with stochastic modelling to take the variation of patient respiratory elastance over time into consideration.

METHODS

A stochastic model of Ers is integrated into the VENT protocol from previous works to develop the SiVENT protocol, to account for both intra- and inter-patient variability. A cohort of 20 virtual MV patients based on retrospective patient data are used to validate the performance of this method for volume-controlled (VC) ventilation. A performance evaluation was conducted where the SiVENT and VENT protocols were implemented in 1080 instances each to compare the two protocols and evaluate the difference in reduction of possible MV settings achieved by each.

RESULTS

From an initial number of 189,000 possible MV setting combinations, the VENT protocol reduced this number to a median of 10,612, achieving a reduction of 94.4% across the cohort. With the integration of the stochastic model component, the SiVENT protocol reduced this number from 189,000 to a median of 9329, achieving a reduction of 95.1% across the cohort. The SiVENT protocol reduces the number of possible combinations provided to the user by more than 1000 combinations as compared to the VENT protocol.

CONCLUSIONS

Adding a stochastic model component into a model-based approach to selecting MV settings improves the ability of a decision support system to recommend patient-specific MV settings. It specifically considers inter- and intra-patient variability in respiratory elastance and eliminates potentially harmful settings based on clinically recommended pressure thresholds. Clinical input and local protocols can further reduce the number of safe setting combinations. The results for the SiVENT protocol justify further investigation of its prediction accuracy and clinical validation trials.

Model-based patient matching for in-parallel pressure-controlled ventilation

Background

Surges of COVID-19 infections have led to insufficient supply of mechanical ventilators (MV), resulting in rationing of MV care. In-parallel, co-mechanical ventilation (Co-MV) of multiple patients is a potential solution. However, due to lack of testing, there is currently no means to match ventilation requirements or patients, with no guidelines to date. In this research, we have developed a model-based method for patient matching for pressure control mode MV.

Methods

The model-based method uses a single-compartment lung model (SCM) to simulate the resultant tidal volume of patient pairs at a set ventilation setting. If both patients meet specified safe ventilation criteria under similar ventilation settings, the actual mechanical ventilator settings for Co-MV are determined via simulation using a double-compartment lung model (DCM). This method allows clinicians to analyse Co-MV in silico, before clinical implementation.

Results

The proposed method demonstrates successful patient matching and MV setting in a model-based simulation as well as good discrimination to avoid mismatched patient pairs. The pairing process is based on model-based, patient-specific respiratory mechanics identified from measured data to provide useful information for guiding care. Specifically, the matching is performed via estimation of MV delivered tidal volume (mL/kg) based on patient-specific respiratory mechanics. This information can provide insights for the clinicians to evaluate the subsequent effects of Co-MV. In addition, it was also found that Co-MV patients with highly restrictive respiratory mechanics and obese patients must be performed with extra care.

Conclusion

This approach allows clinicians to analyse patient matching in a virtual environment without patient risk. The approach is tested in simulation, but the results justify the necessary clinical validation in human trials.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12938-022-00983-y.

Protocol conception for safe selection of mechanical ventilation settings for respiratory failure Patients

BACKGROUND AND OBJECTIVE

Mechanical ventilation is the primary form of care provided to respiratory failure patients. Limited guidelines and conflicting results from major clinical trials means selection of mechanical ventilation settings relies heavily on clinician experience and intuition. Determining optimal mechanical ventilation settings is therefore difficult, where non-optimal mechanical ventilation can be deleterious. To overcome these difficulties, this research proposes a model-based method to manage the wide range of possible mechanical ventilation settings, while also considering patient-specific conditions and responses.

METHODS

This study shows the design and development of the "VENT" protocol, which integrates the single compartment linear lung model with clinical recommendations from landmark studies, to aid clinical decision-making in selecting mechanical ventilation settings. Using retrospective breath data from a cohort of 24 patients, 3,566 and 2,447 clinically implemented VC and PC settings were extracted respectively. Using this data, a VENT protocol application case study and clinical comparison is performed, and the prediction accuracy of the VENT protocol is validated against actual measured outcomes of pressure and volume.

RESULTS

The study shows the VENT protocols' potential use in narrowing an overwhelming number of possible mechanical ventilation setting combinations by up to 99.9%. The comparison with retrospective clinical data showed that only 33% and 45% of clinician settings were approved by the VENT protocol. The unapproved settings were mainly due to exceeding clinical recommended settings. When utilising the single compartment model in the VENT protocol for forecasting peak pressures and tidal volumes, median [IQR] prediction error values of 0.75 [0.31 - 1.83] cmH₂O and 0.55 [0.19 - 1.20] mL/kg were obtained.

CONCLUSIONS

Comparing the proposed protocol with retrospective clinically implemented settings shows the protocol can prevent harmful mechanical ventilation setting combinations for which clinicians would be otherwise unaware. The VENT protocol warrants a more detailed clinical study to validate its potential usefulness in a clinical setting.

Quantifying neonatal patient effort using non-invasive model-based methods

Patient-specific spontaneous breathing effort (SB) is common in invasively mechanically ventilated (MV) adult patients, and especially common in preterm neonates who are not typically sedated. However, there is no proven, ethically feasible and non-invasive method to quantify SB effort in neonates, creating the potential for model-based measures. Lung mechanics and SB effort are segregated using a basis function model to identify passive lung mechanics, and an additional time-varying elastance model to identify patient-specific SB effort and asynchrony as negative and positive added elastances, respectively. Data from ten preterm neonates on standard MV care in the neonatal intensive care unit (NICU) are used to assess this model-based approach, using area under the curve (AUC) for positive (asynchrony) and negative (SB effort) time-varying elastance. Median [interquartile-range (IQR)] of passive pulmonary lung elastance was 3.82 [2.09-5.80] cmH₂O/ml. Median [IQR] AUC quantified SB effort was -0.32 [-0.43--0.12]cmH₂O/ml. AUC quantified asynchrony was 0.00 [0.00-0.01]cmH₂O/ml, and affected 28% of the 25,287 total breaths. This proof of concept model-based approach provides a non-invasive, computationally straightforward, and thus clinically feasible means to quantify patient-specific spontaneous breathing effort and asynchrony.

Quantitative evaluation of PSMA PET imaging using a realistic anthropomorphic phantom and shell-less radioactive epoxy lesions

BACKGROUND

Positron emission tomography (PET) with prostate specific membrane antigen (PSMA) have shown superior performance in detecting metastatic prostate cancers. Relative to [18F]fluorodeoxyglucose ([18F]FDG) PET images, PSMA PET images tend to visualize significantly higher-contrast focal lesions. We aim to evaluate segmentation and reconstruction algorithms in this emerging context. Specifically, Bayesian or maximum a posteriori (MAP) image reconstruction, compared to standard ordered subsets expectation maximization (OSEM) reconstruction, has received significant interest for its potential to reach convergence with minimal noise amplifications. However, few phantom studies have evaluated the quantitative accuracy of such reconstructions for high contrast, small lesions (sub-10 mm) that are typically observed in PSMA images. In this study, we cast 3 mm-16-mm spheres using epoxy resin infused with a long half-life positron emitter (sodium-22; 22Na) to simulate prostate cancer metastasis. The anthropomorphic Probe-IQ phantom, which features a liver, bladder, lungs, and ureters, was used to model relevant anatomy. Dynamic PET acquisitions were acquired and images were reconstructed with OSEM (varying subsets and iterations) and BSREM (varying β parameters), and the effects on lesion quantitation were evaluated.

RESULTS

The 22Na lesions were scanned against an aqueous solution containing fluorine-18 (18F) as the background. Regions-of-interest were drawn with MIM Software using 40% fixed threshold (40% FT) and a gradient segmentation algorithm (MIM's PET Edge+). Recovery coefficients (RCs) (max, mean, peak, and newly defined "apex"), metabolic tumour volume (MTV), and total tumour uptake (TTU) were calculated for each sphere. SUVpeak and SUVapex had the most consistent RCs for different lesion-to-background ratios and reconstruction parameters. The gradient-based segmentation algorithm was more accurate than 40% FT for determining MTV and TTU, particularly for lesions [Formula: see text] 6 mm in diameter (R2 = 0.979-0.996 vs. R2 = 0.115-0.527, respectively).

CONCLUSION

An anthropomorphic phantom was used to evaluate quantitation for PSMA PET imaging of metastatic prostate cancer lesions. BSREM with β = 200-400 and OSEM with 2-5 iterations resulted in the most accurate and robust measurements of SUVmean, MTV, and TTU for imaging conditions in 18F-PSMA PET/CT images. SUVapex, a hybrid metric of SUVmax and SUVpeak, was proposed for robust, accurate, and segmentation-free quantitation of lesions for PSMA PET.

Quantifying ventilator unloading in CPAP ventilation

BACKGROUND

The intrinsic (muscular) patient effort driving inspiration in non-invasive ventilation modes, such as continuous positive airway pressure (CPAP) therapy, has not been identified from non-invasive data. Current CPAP settings are based on clinical judgment and assessment of symptoms of respiratory distress. Non-optimal settings, including too much positive end expiratory pressure (PEEP) can cause unintended lung injury and ventilator unloading, where patient effort drops and the CPAP device enables too much work being imposed on the injured lung. Currently, there is no non-invasive means of quantifying or identifying these effects.

METHODS

A novel model-based method of ascertaining intrinsic patient work of breathing (WOB) in CPAP is developed based on linear single compartment and 2nd order b-spline models previously used in invasive ventilation modes. Results are compared to current clinical indications, such as total Imposed WOB from the CPAP device and beak length, the latter of which is the clinical metric used to indicate alveolar overdistension. Intrinsic and Imposed WOB are compared. The hypothesis is that ventilator unloading can be assessed as a decrease in Intrinsic WOB relative to Imposed WOB, as PEEP and associated ventilator unloading rise. This hypothesis is tested using 14 subjects from a CPAP trial of several breathing rates at two PEEP levels.

RESULTS

The ratio of Intrinsic to Imposed WOB, normalised per unit tidal volume, decreased with increasing PEEP (4-7 cm H₂O), capturing the expected trend of ventilator unloading. Ventilator unloading was observed across all breathing rates. Beak length measurements showed no conclusive evidence of capturing overdistension at higher PEEP or ventilator unloading.

CONCLUSIONS

Patient Intrinsic WOB in CPAP was non-invasively quantified using model-based methods, based on pressure and flow measurements. The ratio of Intrinsic to Imposed WOB per unit tidal volume clearly and consistently showed ventilator unloading across all patients and breathing rates, with Intrinsic WOB decreasing with increasing PEEP. This trend was not observed in the current clinical metric of beak length. Non-invasively quantifying Intrinsic WOB and ventilator unloading is the critical first step to objectively optimising clinical CPAP settings, patient care, and outcomes.

Over-distension prediction via hysteresis loop analysis and patient-specific basis functions in a virtual patient model

BACKGROUND AND OBJECTIVE

Recruitment maneuvers (RMs) with subsequent positive-end-expiratory-pressure (PEEP) have proven effective in recruiting lung volume and preventing alveolar collapse. However, a suboptimal PEEP could induce undesired injury in lungs by insufficient or excessive breath support. Thus, a predictive model for patient response under PEEP changes could improve clinical care and lower risks.

METHODS

This research adds novel elements to a virtual patient model to identify and predict patient-specific lung distension to optimise and personalise care. Model validity and accuracy are validated using data from 18 volume-controlled ventilation (VCV) patients at 7 different baseline PEEP levels (0-12cmH₂O), yielding 623 prediction cases. Predictions were made up to ΔPEEP = 12cmH₂O ahead covering 6x2cmH₂O PEEP steps.

RESULTS

Using the proposed lung distension model, 90% of absolute peak inspiratory pressure (PIP) prediction errors compared to clinical measurement are within 3.95cmH₂O, compared with 4.76cmH₂O without this distension term. Comparing model-predicted and clinically measured distension had high correlation increasing to R² = 0.93-0.95 if maximum ΔPEEP ≤ 6cmH₂O. Predicted dynamic functional residual capacity (V_frc) changes as PEEP rises yield 0.013L median prediction error for both prediction groups and overall R² of 0.84.

CONCLUSIONS

Overall results demonstrate nonlinear distension mechanics are accurately captured in virtual lung mechanics patients for mechanical ventilation, for the first time. This result can minimise the risk of lung injury by predicting its potential occurrence of distension before changing ventilator settings. The overall outcomes significantly extend and more fully validate this virtual mechanical ventilation patient model.

Model-based Patient Matching for in-parallel Multiplexing Mechanical Ventilation Support

Surges of COVID-19 infections could lead to insufficient supply of mechanical ventilators, and rationing of needed care. Multiplexing mechanical ventilators (co-MV) to serve multiple patients is a potential temporary solution. However, if patients are ventilated in parallel ventilation, there is currently no means to match ventilation requirements or patients, with no guidelines to date for co-MV. This research uses patient-specific clinically validated respiratory mechanics models to propose a method for patient matching and mechanical ventilator settings for two-patient co-MV under pressure control mode. The proposed method can simulate and estimate the resultant tidal volume of different combinations of co-ventilated patients. With both patients fulfilling the specified constraint under similar ventilation settings, the actual mechanical ventilator settings for co-MV are determined. This method allows clinicians to analyze in silico co-MV before clinical implementation.

Model-based estimation of negative inspiratory driving pressure in patients receiving invasive NAVA mechanical ventilation

BACKGROUND

Optimisation of mechanical ventilation (MV) and weaning requires insight into underlying patient breathing effort. Current identifiable models effectively describe lung mechanics, such as elastance (E) and resistance (R) at the bedside in sedated patients, but are less effective when spontaneous breathing is present. This research derives and regularises a single compartment model to identify patient-specific inspiratory effort.

METHODS

Constrained second-order b-spline basis functions (knot width 0.05 s) are used to describe negative inspiratory drive (Pp, cmH₂O) as a function of time. Breath-breath Pp are identified with single E and R values over inspiration and expiration from n = 20 breaths for N = 22 patients on NAVA ventilation. Pp is compared to measured electrical activity of the diaphragm (Eadi) and published results.

RESULTS

Average per-patient root-mean-squared model fit error was (median [interquartile range, IQR]) 0.9 [0.6-1.3] cmH₂O, and average per-patient median Pp was -3.9 [-4.5- -3.0] cmH₂O, with range -7.9 - -1.9 cmH₂O. Per-patient E and R were 16.4 [13.6-21.8] cmH₂O/L and 9.2 [6.4-13.1] cmH₂O.s/L, respectively. Most patients showed an inspiratory volume threshold beyond which Pp started to return to baseline, and Pp at peak Eadi (end-inspiration) was often strongly correlated with peak Eadi (R²=0.25-0.86). Similarly, average transpulmonary pressure was consistent breath-breath in most patients, despite differences in peak Eadi and thus peak airway pressure.

CONCLUSIONS

The model-based inspiratory effort aligns with electrical muscle activity and published studies showing neuro-muscular decoupling as a function of pressure and/or volume. Consistency in coupling/dynamics were patient-specific. Quantification of patient and ventilator work of breathing contributions may aid optimisation of MV modes and weaning.

Tube-load model: A clinically applicable pulse contour analysis method for estimation of cardiac stroke volume

BACKGROUND AND OBJECTIVES

Accurate, reproducible, and reliable real-time clinical measurement of stroke volume (SV) is challenging. To accurately estimate arterial mechanics and SV by pulse contour analysis, accounting for wave reflection, such as by a tube-load model, is potentially important. This study tests for the first time whether a dynamically identified tube-load model, given a single peripheral arterial input signal and pulse transit time (PTT), provides accurate SV estimates during hemodynamic instability.

METHODS

The model is tested for 5 pigs during hemodynamic interventions, using either an aortic flow probe or admittance catheter for a validation SV measure. Performance is assessed using Bland-Altman and polar plot analysis for a series of long-term state-change and short-term dynamic events.

RESULTS

The overall median bias and limits of agreement (2.5th, 97.5th percentile) from Bland-Altman analysis were -10% [-49, 36], and -1% [-28,20] for state-change and dynamic events, respectively. The angular limit of agreement (maximum of 2.5th, 97.5th percentile) from polar-plot analysis for state-change and dynamic interventions was 35.6^∘, and 35.2^∘, respectively.

CONCLUSION

SV estimation agreement and trending performance was reasonable given the severity of the interventions. This simple yet robust method has potential to track SV within acceptable limits during hemodynamic instability in critically ill patients, provided a sufficiently accurate PTT measure.

The goldilocks problem: Nutrition and its impact on glycaemic control

BACKGROUND

Glucose intolerance and insulin resistance manifest as hyperglycaemia in intensive care, which is associated with mortality and morbidities. Glycaemic control (GC) may improve outcomes, though safe and effective control has proven elusive. Nutritional glucose intake affects blood glucose (BG) outcomes, but few protocols actively control it. This study aims to examine BG outcomes in the context of nutritional management during GC.

METHODS

Retrospective cohort analysis of 5 glycaemic control cohorts spanning 4 years (n = 273) from Christchurch Hospital Intensive Care Unit (ICU). GC is delivered using a single model-based protocol (STAR), with default 4.4-8.0 mmol/L target range via. modulation of insulin and nutrition. Clinical adaptations/cohorts include variations on upper target (UL-9 with 9.0 mmol/L, reducing workload and nutrition responsiveness), and insulin only (IO) with clinically set nutrition at 3 glucose concentrations (71 g/L vs. 120 and 180 g/L in the TARGET study).

RESULTS

Percent of BG hours in the 4.4-8.0 mmol/L range highest under standard STAR conditions (78%), and was lower at 64% under UL-9, likely due to reduced time-responsiveness of nutrition-insulin changes. By comparison, IO only resulted in 64-69% BG in range across different nutrition types. A subset of patients receiving high glucose nutrition under IO were persistently hyperglycaemic, indicating patient-specific glucose tolerance.

CONCLUSION

STAR GC is most effective when nutrition and insulin are modulated together with timely responsiveness to persistent hyperglycaemia. Results imply modulation of nutrition alongside insulin improves GC, particularly in patients with persistent hyperglycaemia/low glucose tolerance.

High Inter-Patient Variability in Sepsis Evolution: A Hidden Markov Model Analysis

BACKGROUND

Severe sepsis and septic shock are common in the intensive care unit (ICU) and contribute significantly to cost and mortality. Early treatment is critical but is confounded by the difficulty of real-time diagnosis. This study uses hidden Markov models (HMMs) to examine whether the time evolution of sepsis can add diagnostic accuracy or value using a proven set of bio-signals.

METHODS

Clinical data (N=36 patients; 6071 hours), including an hourly personalised insulin sensitivity metric. A two hidden state HMM is created to discriminate diagnosed cases (Severe Sepsis, Septic Shock) from controls (SIRS, Sepsis) states. Diagnostic performance is measured by ROC curves, likelihood ratios (LHRs), sensitivity/specificity, and diagnostic odds-ratios (DOR), for a best-case resubstitution estimate and a worst-case 80/20% repeated holdout analysis.

RESULTS

The HMM delivered near perfect results (95% Sensitivity; 96% Specificity) for best-case resubstitution estimates, but was comparatively poor (59% Sensitivity; 61% Specificity) for worst-case repeated holdout estimations. Adding the time evolution of sepsis did not add to the accuracy of diagnosis from using the signals alone without time history.

CONCLUSIONS

These potentially surprising results indicate significant inter-patient variability in the time evolution of sepsis, preventing effective diagnosis in the context of the bio-signals, data, and HMM topology used. Efforts for improved real-time, early sepsis diagnosis should concentrate on the robustness and efficacy of the bio-signals and data used, as well as the level of model complexity, to create more effective real-time classifiers.

Variational Autoencoder Modular Bayesian Networks for Simulation of Heterogeneous Clinical Study Data

In the area of Big Data, one of the major obstacles for the progress of biomedical research is the existence of data “silos” because legal and ethical constraints often do not allow for sharing sensitive patient data from clinical studies across institutions. While federated machine learning now allows for building models from scattered data of the same format, there is still the need to investigate, mine, and understand data of separate and very differently designed clinical studies that can only be accessed within each of the data-hosting organizations. Simulation of sufficiently realistic virtual patients based on the data within each individual organization could be a way to fill this gap. In this work, we propose a new machine learning approach [Variational Autoencoder Modular Bayesian Network (VAMBN)] to learn a generative model of longitudinal clinical study data. VAMBN considers typical key aspects of such data, namely limited sample size coupled with comparable many variables of different numerical scales and statistical properties, and many missing values. We show that with VAMBN, we can simulate virtual patients in a sufficiently realistic manner while making theoretical guarantees on data privacy. In addition, VAMBN allows for simulating counterfactual scenarios. Hence, VAMBN could facilitate data sharing as well as design of clinical trials.