Brain Gray Matter MRI Morphometry for Neuroprognostication After Cardiac Arrest

Neuroprotective Effects of Inhaled Xenon Gas on Brain Structural Gray Matter Changes After Out-of-Hospital Cardiac Arrest Evaluated by Morphometric Analysis: A Substudy of the Randomized Xe-Hypotheca Trial

BACKGROUND

We have earlier reported that inhaled xenon combined with hypothermia attenuates brain white matter injury in comatose survivors of out-of-hospital cardiac arrest (OHCA). A predefined secondary objective was to assess the effect of inhaled xenon on the structural changes in gray matter in comatose survivors after OHCA.

METHODS

Patients were randomly assigned to receive either inhaled xenon combined with target temperature management (33 °C) for 24 h (n = 55, xenon group) or target temperature management alone (n = 55, control group). A change of brain gray matter volume was assessed with a voxel-based morphometry evaluation of high-resolution structural brain magnetic resonance imaging (MRI) data with Statistical Parametric Mapping. Patients were scheduled to undergo the first MRI between 36 and 52 h and a second MRI 10 days after OHCA.

RESULTS

Of the 110 randomly assigned patients in the Xe-Hypotheca trial, 66 patients completed both MRI scans. After all imaging-based exclusions, 21 patients in the control group and 24 patients in the xenon group had both scan 1 and scan 2 available for analyses with scans that fulfilled the quality criteria. Compared with the xenon group, the control group had a significant decrease in brain gray matter volume in several clusters in the second scan compared with the first. In a between-group analysis, significant reductions were found in the right amygdala/entorhinal cortex (p = 0.025), left amygdala (p = 0.043), left middle temporal gyrus (p = 0.042), left inferior temporal gyrus (p = 0.008), left parahippocampal gyrus (p = 0.042), left temporal pole (p = 0.042), and left cerebellar cortex (p = 0.005). In the remaining gray matter areas, there were no significant changes between the groups.

CONCLUSIONS

In comatose survivors of OHCA, inhaled xenon combined with targeted temperature management preserved gray matter better than hypothermia alone.

CLINICAL TRIAL REGISTRATION

ClinicalTrials.gov: NCT00879892.

Application of multi-feature-based machine learning models to predict neurological outcomes of cardiac arrest

Cardiac arrest (CA) is a major disease burden worldwide and has a poor prognosis. Early prediction of CA outcomes helps optimize the therapeutic regimen and improve patients' neurological function. As the current guidelines recommend, many factors can be used to evaluate the neurological outcomes of CA patients. Machine learning (ML) has strong analytical abilities and fast computing speed; thus, it plays an irreplaceable role in prediction model development. An increasing number of researchers are using ML algorithms to incorporate demographics, arrest characteristics, clinical variables, biomarkers, physical examination findings, electroencephalograms, imaging, and other factors with predictive value to construct multi-feature prediction models for neurological outcomes of CA survivors. In this review, we explore the current application of ML models using multiple features to predict the neurological outcomes of CA patients. Although the outcome prediction model is still in development, it has strong potential to become a powerful tool in clinical practice.

Multiomic biomarkers after cardiac arrest

Cardiac arrest is a sudden cessation of heart function, leading to an abrupt loss of blood flow and oxygen to vital organs. This life-threatening emergency requires immediate medical intervention and can lead to severe neurological injury or death. Methods and biomarkers to predict neurological outcome are available but lack accuracy. Such methods would allow personalizing healthcare and help clinical decisions. Extensive research has been conducted to identify prognostic omic biomarkers of cardiac arrest. With the emergence of technologies allowing to combine different levels of omics data, and with the help of artificial intelligence and machine learning, there is a potential to use multiomic signatures as prognostic biomarkers after cardiac arrest. This review article delves into the current knowledge of cardiac arrest biomarkers across various omic fields and suggests directions for future research aiming to integrate multiple omics data layers to improve outcome prediction and cardiac arrest patient's care.

Artificial Intelligence and Machine Learning Applications in Critically Ill Brain Injured Patients

The utilization of Artificial Intelligence (AI) and Machine Learning (ML) is paving the way for significant strides in patient diagnosis, treatment, and prognostication in neurocritical care. These technologies offer the potential to unravel complex patterns within vast datasets ranging from vast clinical data and EEG (electroencephalogram) readings to advanced cerebral imaging facilitating a more nuanced understanding of patient conditions. Despite their promise, the implementation of AI and ML faces substantial hurdles. Historical biases within training data, the challenge of interpreting multifaceted data streams, and the "black box" nature of ML algorithms present barriers to widespread clinical adoption. Moreover, ethical considerations around data privacy and the need for transparent, explainable models remain paramount to ensure trust and efficacy in clinical decision-making.This article reflects on the emergence of AI and ML as integral tools in neurocritical care, discussing their roles from the perspective of both their scientific promise and the associated challenges. We underscore the importance of extensive validation in diverse clinical settings to ensure the generalizability of ML models, particularly considering their potential to inform critical medical decisions such as withdrawal of life-sustaining therapies. Advancement in computational capabilities is essential for implementing ML in clinical settings, allowing for real-time analysis and decision support at the point of care. As AI and ML are poised to become commonplace in clinical practice, it is incumbent upon health care professionals to understand and oversee these technologies, ensuring they adhere to the highest safety standards and contribute to the realization of personalized medicine. This engagement will be pivotal in integrating AI and ML into patient care, optimizing outcomes in neurocritical care through informed and data-driven decision-making.

Neuroimmune activation is associated with neurological outcome in anoxic and traumatic coma

The pathophysiological underpinnings of critically disrupted brain connectomes resulting in coma are poorly understood. Inflammation is potentially an important but still undervalued factor. Here, we present a first-in-human prospective study using the 18-kDa translocator protein (TSPO) radioligand 18F-DPA714 for PET imaging to allow in vivo neuroimmune activation quantification in patients with coma (n = 17) following either anoxia or traumatic brain injuries in comparison with age- and sex-matched controls. Our findings yielded novel evidence of an early inflammatory component predominantly located within key cortical and subcortical brain structures that are putatively implicated in consciousness emergence and maintenance after severe brain injury (i.e. mesocircuit and frontoparietal networks). We observed that traumatic and anoxic patients with coma have distinct neuroimmune activation profiles, both in terms of intensity and spatial distribution. Finally, we demonstrated that both the total amount and specific distribution of PET-measurable neuroinflammation within the brain mesocircuit were associated with the patient's recovery potential. We suggest that our results can be developed for use both as a new neuroprognostication tool and as a promising biometric to guide future clinical trials targeting glial activity very early after severe brain injury.

Mesocircuit mechanisms in the diagnosis and treatment of disorders of consciousness

The 'mesocircuit hypothesis' proposes mechanisms underlying the recovery of consciousness following severe brain injuries. The model builds up from a single premise that multifocal brain injuries resulting in coma and subsequent disorders of consciousness produce widespread neuronal death and dysfunction. Considering the general properties of cortical, thalamic, and striatal neurons, a lawful and specific circuit-level mechanism is constructed based on these known anatomical and physiological specializations of neuronal subtypes. The mesocircuit model generates many testable predictions at the mesocircuit, local circuit, and cellular level across multiple cerebral structures to correlate diagnostic measurements and interpret therapeutic interventions. The anterior forebrain mesocircuit is integrally related to the frontal-parietal network, another network demonstrated to show strong correlation with levels of recovery in disorders of consciousness. A further extension known as the "ABCD" model has been used to examine interaction of these models in recovery of consciousness using electrophysiological data types. Many studies have examined predictions of the mesocircuit model; here we first present the model and review the accumulated evidence for several predictions of model across multiple stages of recovery function in human subjects. Recent studies linking the mesocircuit model, the ABCD model, and interactions with the frontoparietal network are reviewed. Finally, theoretical implications of the mesocircuit model at the neuronal level are considered to interpret recent studies of deep brain stimulation in the central lateral thalamus in patients recovering from coma and in new experimental models in the context of emerging understanding of neuronal and local circuit mechanisms underlying conscious brain states.

Multimodal MRI-Based Whole-Brain Assessment in Patients In Anoxoischemic Coma by Using 3D Convolutional Neural Networks

BACKGROUND

There is an unfulfilled need to find the best way to automatically capture, analyze, organize, and merge structural and functional brain magnetic resonance imaging (MRI) data to ultimately extract relevant signals that can assist the medical decision process at the bedside of patients in postanoxic coma. We aimed to develop and validate a deep learning model to leverage multimodal 3D MRI whole-brain times series for an early evaluation of brain damages related to anoxoischemic coma.

METHODS

This proof-of-concept, prospective, cohort study was undertaken at the intensive care unit affiliated with the University Hospital (Toulouse, France), between March 2018 and May 2020. All patients were scanned in coma state at least 2 days (4 ± 2 days) after cardiac arrest. Over the same period, age-matched healthy volunteers were recruited and included. Brain MRI quantification encompassed both "functional data" from regions of interest (precuneus and posterior cingulate cortex) with whole-brain functional connectivity analysis and "structural data" (gray matter volume, T1-weighted, fractional anisotropy, and mean diffusivity). A specifically designed 3D convolutional neuronal network (CNN) was created to allow conscious state discrimination (coma vs. controls) by using raw MRI indices as the input. A voxel-wise visualization method based on the study of convolutional filters was applied to support CNN outcome. The Ethics Committee of the University Teaching Hospital of Toulouse, France (2018-A31) approved the study and informed consent was obtained from all participants.

RESULTS

The final cohort consisted of 29 patients in postanoxic coma and 34 healthy volunteers. Coma patients were successfully discerned from controls by using 3D CNN in combination with different MR indices. The best accuracy was achieved by functional MRI data, in particular with resting-state functional MRI of the posterior cingulate cortex, with an accuracy of 0.96 (range 0.94-0.98) on the test set from 10-time repeated tenfold cross-validation. Even more satisfactory performances were achieved through the majority voting strategy, which was able to compensate for mistakes from single MR indices. Visualization maps allowed us to identify the most relevant regions for each MRI index, notably regions previously described as possibly being involved in consciousness emergence. Interestingly, a posteriori analysis of misclassified patients indicated that they may present some common functional MRI traits with controls, which suggests further favorable outcomes.

CONCLUSIONS

A fully automated identification of clinically relevant signals from complex multimodal neuroimaging data is a major research topic that may bring a radical paradigm shift in the neuroprognostication of patients with severe brain injury. We report for the first time a successful discrimination between patients in postanoxic coma patients from people serving as controls by using 3D CNN whole-brain structural and functional MRI data. Clinical Trial Number http://ClinicalTrials.gov (No. NCT03482115).

Mapping the Unconscious Brain: Insights From Advanced Neuroimaging

SUMMARY

Recent advances in neuroimaging have been a preeminent factor in the scientific effort to unravel mechanisms of conscious awareness and the pathophysiology of disorders of consciousness. In the first part of this review, we selectively discuss operational models of consciousness, the biophysical signal that is measured using different imaging modalities, and knowledge on disorders of consciousness that has been gleaned with each neuroimaging modality. Techniques considered include diffusion-weighted imaging, diffusion tensor imaging, different types of nuclear medicine imaging, functional MRI, magnetoencephalography, and the combined transcranial magnetic stimulation-electroencephalography approach. In the second part of this article, we provide an overview of how advanced neuroimaging can be leveraged to support neurological prognostication, the use of machine learning to process high-dimensional imaging data, potential applications in clinical practice, and future directions.

Toward a coherent structuration of disorders of consciousness expertise at a country scale: A proposal for France

Probing consciousness and cognitive abilities in non-communicating patients is one of the most challenging diagnostic issues. A fast growing medical and scientific literature explores the various facets of this challenge, often coined under the generic expression of 'Disorders of Consciousness' (DoC). Crucially, a set of independent converging results demonstrated both (1) the diagnostic and prognostic importance of this expertise, and (2) the need to combine behavioural measures with brain structure and activity data to improve diagnostic and prognostication accuracy as well as potential therapeutic intervention. Thus, probing consciousness in DoC patients appears as a crucial activity rich of human, medical, economic and ethical consequences, but this activity needs to be organized in order to offer this expertise to each concerned patient. More precisely, diagnosis of consciousness differs in difficulty across patients: while a minimal set of data can be sufficient to reach a confident result, some patients need a higher level of expertise that relies on additional behavioural and brain activity and brain structure measures. In order to enable this service on a systematic mode, we present two complementary proposals in the present article. First, we sketch a structuration of DoC expertise at a country-scale, namely France. More precisely, we suggest that a 2-tiers network composed of local (Tier-1) and regional (Tier-2) centers backed by distant electronic databases and algorithmic centers could optimally enable the systematic implementation of DoC expertise in France. Second, we propose to create a national common register of DoC patients in order to better monitor this activity, to improve its performance on the basis of nation-wide collected evidence, and to promote rational decision-making.

Neuroprognostication after Cardiac Arrest: Who Recovers? Who Progresses to Brain Death?

Approximately 15% of deaths in developed nations are due to sudden cardiac arrest, making it the most common cause of death worldwide. Though high-quality cardiopulmonary resuscitation has improved overall survival rates, the majority of survivors remain comatose after return of spontaneous circulation secondary to hypoxic ischemic injury. Since the advent of targeted temperature management, neurologic recovery has improved substantially, but the majority of patients are left with neurologic deficits ranging from minor cognitive impairment to persistent coma. Of those who survive cardiac arrest, but die during their hospitalization, some progress to brain death and others die after withdrawal of life-sustaining treatment due to anticipated poor neurologic prognosis. Here, we discuss considerations neurologists must make when asked, "Given their recent cardiac arrest, how much neurologic improvement do we expect for this patient?"

Therapies to Restore Consciousness in Patients with Severe Brain Injuries: A Gap Analysis and Future Directions

Background/Objective

For patients with disorders of consciousness (DoC) and their families, the search for new therapies has been a source of hope and frustration. Almost all clinical trials in patients with DoC have been limited by small sample sizes, lack of placebo groups, and use of heterogeneous outcome measures. As a result, few therapies have strong evidence to support their use; amantadine is the only therapy recommended by current clinical guidelines, specifically for patients with DoC caused by severe traumatic brain injury. To foster and advance development of consciousness-promoting therapies for patients with DoC, the Curing Coma Campaign convened a Coma Science Work Group to perform a gap analysis.

Methods

We consider five classes of therapies: (1) pharmacologic; (2) electromagnetic; (3) mechanical; (4) sensory; and (5) regenerative. For each class of therapy, we summarize the state of the science, identify gaps in knowledge, and suggest future directions for therapy development.

Results

Knowledge gaps in all five therapeutic classes can be attributed to the lack of: (1) a unifying conceptual framework for evaluating therapeutic mechanisms of action; (2) large-scale randomized controlled trials; and (3) pharmacodynamic biomarkers that measure subclinical therapeutic effects in early-phase trials. To address these gaps, we propose a precision medicine approach in which clinical trials selectively enroll patients based upon their physiological receptivity to targeted therapies, and therapeutic effects are measured by complementary behavioral, neuroimaging, and electrophysiologic endpoints.

Conclusions

This personalized approach can be realized through rigorous clinical trial design and international collaboration, both of which will be essential for advancing the development of new therapies and ultimately improving the lives of patients with DoC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12028-021-01227-y.

Machine learning methods to improve bedside fluid responsiveness prediction in severe sepsis or septic shock: an observational study

BACKGROUND

Passive leg raising (PLR) predicts fluid responsiveness in critical illness, although restrictions in mobilising patients often preclude this haemodynamic challenge being used. We investigated whether machine learning applied on transthoracic echocardiography (TTE) data might be used as a tool for predicting fluid responsiveness in critically ill patients.

METHODS

We studied, 100 critically ill patients (mean age: 62 yr [standard deviation: 14]) with severe sepsis or septic shock prospectively over 24 months. Transthoracic echocardiography measurements were performed at baseline, after PLR, and before and after a standardised fluid challenge in learning and test populations (n=50 patients each). A 15% increase in stroke volume defined fluid responsiveness. The machine learning methods used were classification and regression tree (CART), partial least-squares regression (PLS), neural network (NNET), and linear discriminant analysis (LDA). Each method was applied offline to determine whether fluid responsiveness may be predicted from left and right cardiac ventricular physiological changes detected by cardiac ultrasound. Predictive values for fluid responsiveness were compared by receiver operating characteristics (area under the curve [AUC]; mean [95% confidence intervals]).

RESULTS

In the learning sample, the AUC values were PLR 0.76 (0.62-0.89), CART 0.83 (0.73-0.94), PLS 0.97 (0.93-1), NNET 0.93 (0.85-1), and LDA 0.90 (0.81-0.98). In the test sample, the AUC values were PLR 0.77 (0.64-0.91), CART 0.68 (0.54-0.81), PLS 0.83 (0.71-0.96), NNET 0.83 (0.71-0.94), and LDA 0.85 (0.74-0.96) respectively. The PLS model identified inferior vena cava collapsibility, velocity-time integral, S-wave, E/Ea ratio, and E-wave as key echocardiographic parameters.

CONCLUSIONS

Machine learning generated several models for predicting fluid responsiveness that were comparable with the haemodynamic response to PLR.

Functional and Structural Integrity of Frontoparietal Connectivity in Traumatic and Anoxic Coma

OBJECTIVES

Recovery from coma might critically depend on the structural and functional integrity of frontoparietal networks. We aimed to measure this integrity in traumatic brain injury and anoxo-ischemic (cardiac arrest) coma patients by using an original multimodal MRI protocol.

DESIGN

Prospective cohort study.

SETTING

Three Intensive Critical Care Units affiliated to the University in Toulouse (France).

PATIENTS

We longitudinally recruited 43 coma patients (Glasgow Coma Scale at the admission < 8; 29 cardiac arrest and 14 traumatic brain injury) and 34 age-matched healthy volunteers. Exclusion criteria were disorders of consciousness lasting more than 30 days and focal brain damage within the explored brain regions. Patient assessments were conducted at least 2 days (5 ± 2 d) after complete withdrawal of sedation. All patients were followed up (Coma Recovery Scale-Revised) 3 months after acute brain injury.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Functional and structural MRI data were recorded, and the analysis was targeted on the posteromedial cortex, the medial prefrontal cortex, and the cingulum. Univariate analyses and machine learning techniques were used to assess diagnostic and predictive values. Coma patients displayed significantly lower medial prefrontal cortex-posteromedial cortex functional connectivity (area under the curve, 0.94; 95% CI, 0.93-0.95). Cardiac arrest patients showed specific structural disturbances within posteromedial cortex. Significant cingulum architectural disturbances were observed in traumatic brain injury patients. The machine learning medial prefrontal cortex-posteromedial cortex multimodal classifier had a significant predictive value (area under the curve, 0.96; 95% CI, 0.95-0.97), best combination of subregions that discriminates a binary outcome based on Coma Recovery Scale-Revised).

CONCLUSIONS

This exploratory study suggests that frontoparietal functional disconnections are specifically observed in coma and their structural counterpart provides information about brain injury mechanisms. Multimodal MRI biomarkers of frontoparietal disconnection predict 3-month outcome in our sample. These findings suggest that fronto-parietal disconnection might be particularly relevant for coma outcome prediction and could inspire innovative precision medicine approaches.

Resuscitating the Globally Ischemic Brain: TTM and Beyond

Cardiac arrest (CA) afflicts ~ 550,000 people each year in the USA. A small fraction of CA sufferers survive with a majority of these survivors emerging in a comatose state. Many CA survivors suffer devastating global brain injury with some remaining indefinitely in a comatose state. The pathogenesis of global brain injury secondary to CA is complex. Mechanisms of CA-induced brain injury include ischemia, hypoxia, cytotoxicity, inflammation, and ultimately, irreversible neuronal damage. Due to this complexity, it is critical for clinicians to have access as early as possible to quantitative metrics for diagnosing injury severity, accurately predicting outcome, and informing patient care. Current recommendations involve using multiple modalities including clinical exam, electrophysiology, brain imaging, and molecular biomarkers. This multi-faceted approach is designed to improve prognostication to avoid "self-fulfilling" prophecy and early withdrawal of life-sustaining treatments. Incorporation of emerging dynamic monitoring tools such as diffuse optical technologies may provide improved diagnosis and early prognostication to better inform treatment. Currently, targeted temperature management (TTM) is the leading treatment, with the number of patients needed to treat being ~ 6 in order to improve outcome for one patient. Future avenues of treatment, which may potentially be combined with TTM, include pharmacotherapy, perfusion/oxygenation targets, and pre/postconditioning. In this review, we provide a bench to bedside approach to delineate the pathophysiology, prognostication methods, current targeted therapies, and future directions of research surrounding hypoxic-ischemic brain injury (HIBI) secondary to CA.

Out-of-hospital cardiac arrest: in-hospital intervention strategies

The prognosis after out-of-hospital cardiac arrest (OHCA) has improved in the past few decades because of advances in interventions used outside and in hospital. About half of patients who have OHCA with initial ventricular tachycardia or ventricular fibrillation and who are admitted to hospital in coma after return of spontaneous circulation will survive to discharge with a reasonable neurological status. In this Series paper we discuss in-hospital management of patients with post-cardiac-arrest syndrome. In most patients, the most important in-hospital interventions other than routine intensive care are continuous active treatment (in non-comatose and comatose patients and including circulatory support in selected patients), cooling of core temperature to 32-36°C by targeted temperature management for at least 24 h, immediate coronary angiography with or without percutaneous coronary intervention, and delay of final prognosis until at least 72 h after OHCA. Prognosis should be based on clinical observations and multimodal testing, with focus on no residual sedation.

A nonparametric Bayesian method of translating machine learning scores to probabilities in clinical decision support

BACKGROUND

Probabilistic assessments of clinical care are essential for quality care. Yet, machine learning, which supports this care process has been limited to categorical results. To maximize its usefulness, it is important to find novel approaches that calibrate the ML output with a likelihood scale. Current state-of-the-art calibration methods are generally accurate and applicable to many ML models, but improved granularity and accuracy of such methods would increase the information available for clinical decision making. This novel non-parametric Bayesian approach is demonstrated on a variety of data sets, including simulated classifier outputs, biomedical data sets from the University of California, Irvine (UCI) Machine Learning Repository, and a clinical data set built to determine suicide risk from the language of emergency department patients.

RESULTS

The method is first demonstrated on support-vector machine (SVM) models, which generally produce well-behaved, well understood scores. The method produces calibrations that are comparable to the state-of-the-art Bayesian Binning in Quantiles (BBQ) method when the SVM models are able to effectively separate cases and controls. However, as the SVM models' ability to discriminate classes decreases, our approach yields more granular and dynamic calibrated probabilities comparing to the BBQ method. Improvements in granularity and range are even more dramatic when the discrimination between the classes is artificially degraded by replacing the SVM model with an ad hoc k-means classifier.

CONCLUSIONS

The method allows both clinicians and patients to have a more nuanced view of the output of an ML model, allowing better decision making. The method is demonstrated on simulated data, various biomedical data sets and a clinical data set, to which diverse ML methods are applied. Trivially extending the method to (non-ML) clinical scores is also discussed.