Task-based EEG and fMRI paradigms in a multimodal clinical diagnostic framework for disorders of consciousness

1.1 Disorders of consciousness

From an operational definition perspective, consciousness is understood as a phenomenon classically described in terms of two main measurable components: (1) wakefulness, also termed arousal (i.e., sleep-wake cycles as evidenced by eye opening) (Giacino et al. 2002; Laureys et al. 2010), and (2) awareness (i.e., purposeful and non-reflexive movements and/or responsiveness to external stimuli) (Blumenfeld 2010; Giacino et al. 2002). Accordingly, the wide spectrum of conditions following sudden-onset acquired brain injury (ABI) in which these components are impaired are referred to as disorders of consciousness (DoC) (Giacino et al. 2014; Royal College of Physicians 2020).

The most commonly described DoC include coma, persistent or prolonged vegetative state (PVS; otherwise known as “unresponsive wakefulness syndrome” (UWS); PVS/UWS hereafter) (Laureys et al. 2004, 2010), minimally conscious state (MCS) and locked-in syndrome (LIS) (Edlow et al. 2021a). Coma refers to a state of complete unwakefulness and unawareness (i.e., with no eye opening and unresponsive); PVS/UWS refers to a state of wakefulness and unawareness (i.e., with eye opening but unresponsive) (Edlow et al. 2021a); MCS refers to a state of wakefulness with inconsistently reproducible responses indicating some degree of preserved awareness (i.e., with eye opening and severely limited responsiveness) (Laureys et al. 2004); LIS refers to a state of quadriplegia and anarthria but preserved higher cortical function as identified by eye-coded communication or by functional imaging (Bruno et al. 2011; Formisano et al. 2013). Below, alternative definitions, if used by cited studies, will be specified.

It is imperative that clinicians distinguish between MCS and PVS/UWS states because their prognoses and management are significantly different, ranging from the withdrawal of life support to active rehabilitation (Edlow et al. 2021b; Giacino et al. 2018b; Laureys et al. 2004, 2016; Royal College of Physicians 2020; Stender et al. 2014).

1.2 Clinical assessments of DoC

Currently, the gold standard of diagnosing DoC is by clinical assessment (Royal College of Physicians 2020). They are classified into conventional consensus assessments and standardized assessments. Consensus-based assessments involve making a diagnosis through agreement between members of a multi-disciplinary clinical team (Royal College of Physicians 2020; Schnakers et al. 2009). Ideally, this team should comprise physicians, nurses, physiotherapists, occupational and speech therapists (Giacino et al. 2018a; Royal College of Physicians 2020). Such assessments often consist of tests covering the domains of mental status, cranial nerve function, sensori-motor and brainstem function (Royal College of Physicians 2020).

Another approach to diagnosing DoC is by standardized assessments, which evidence shows can improve the accuracy of DoC diagnosis by up to 41 % when compared with consensus-based assessments (Giacino et al. 2018b; Schnakers et al. 2009; Wannez et al. 2017). The JFK Coma Recovery Scale-Revised (CRS-R), Wessex Head Injury Matrix (WHIM), Sensory Modality Assessment and Rehabilitation Technique (SMART), Western NeuroSensory Stimulation Protocol (WNSSP), and the DoC Scale (DOCS) have been recommended to varying extents by international guidelines, such as the American Congress of Rehabilitation Medicine (Giacino et al. 2018b; Seel et al. 2010). Of note, the CRS-R is a scale that is recommended by both the American Association of Neurology (Giacino et al. 2018a) and the Royal College of Physicians (Royal College of Physicians 2020). Assessment scales based on subjective reports, such as the DoC Feeling Score, are out of the scope of this text and will not be discussed.

The CRS-R is regarded as the gold standard for clinical diagnosis (Cruse et al. 2012; Giacino et al. 2004; La Porta et al. 2013). The total CRS-R score, ranging from 0 (minimum score) to 23 (maximum score), is a summation of several domains that describe a patient’s auditory, visual, motor and verbal functions, in addition to command-following, communication, and arousal functions (Giacino et al. 2004). The CRS-R also includes items that are tailored to the definition of MCS – such as visual object reaching/localization, intelligible verbalization and object manipulation contingent upon environmental stimuli (Calabro et al. 2016; Royal College of Physicians 2020) – items that are less or not emphasized in conventional consensus assessments. It is the only scale that meets all of the Aspen Workgroup criteria for MCS (Giacino et al. 2004; Zheng et al. 2023).

1.3 Misdiagnosis of DoC

Evidence shows that consensus-based assessments have a diagnostic error rate of up to 40 % (Andrews et al. 1996), even when performed by experienced clinicians (Giacino et al. 2018a). This is in spite of serial evaluations or involvement of a patient’s caregiver who may detect signs overlooked by the clinical team (Royal College of Physicians 2020).

For standardized assessments, a meta-analysis found that up to 14.4 % of PVS/UWS patients were able to wilfully modify brain activity on command (Kondziella et al. 2016), meaning that they were misdiagnosed. Another study found that 32 % of clinically unresponsive PVS/UWS patients, as identified by the CRS-R, demonstrated fluorodeoxyglucose positron emission tomography (¹⁸FDG-PET) neuroimaging characteristics of hypometabolism suggestive of MCS instead (Stender et al. 2014). Similarly, 8 % of PVS/UWS patients were found during fMRI tasks to be MCS instead (Bekinschtein et al. 2011).

Not only are clinical evaluations dependent on the experience and skills of the assessor (Andrews et al. 1996; Coleman et al. 2009a), but the tools themselves depend primarily on overt motor responses, e.g., eye opening or verbal response. It is widely accepted that consciousness does not just have an overt component (Edlow et al. 2021a,b; Kondziella et al. 2020); a covert component also exists, demonstrated by specific changes in neural activity patterns as the patient responds to commands or stimuli without exhibiting motor behavior (Provencio et al. 2020). Given the inadequacy of clinical assessments in reflecting both components of consciousness, misdiagnosis is likely to remain a problem (Andrews et al. 1996; Royal College of Physicians 2020; Schnakers et al. 2016; Stender et al. 2014).

In particular, the concept of PVS/UWS, first proposed by Jennett and Plum (1972), was originally coined to indicate the absence of cortical function based on observations of motor behaviors. This original definition makes no inference whatsoever with regard to the underlying pathology and/or the neuroanatomical location of the lesion (Coleman et al. 2009a; Giacino et al. 2014). Furthermore, Jennett and Plum emphasized the need for additional investigations after making a bedside determination of a PVS/UWS state and advised against viewing the condition as a diagnostic endpoint per se (Coleman et al. 2009b; Jennett and Plum 1972).

1.4 Ethical implications of misdiagnosis

The ethical implications of avoiding misdiagnosis of DoC and using advanced neuroimaging to distinguish MCS or higher states from PVS/UWS are profound and manifold. Three major areas of neuroethics in the context of DoC are discussed below.

One important ethical issue of misdiagnosis is the premature termination of life support. The work from Massachusetts General Hospital led by Edlow et al. (2017) confirmed results from prior studies that command-following and higher-order motor dissociation (HMD) can be detected in some PVS/UWS patients. Notably, all these patients showed better recovery to beyond a confusional state after 6 months (Edlow et al. 2017). Similarly, a recent study on cognitive motor dissociation (CMD) found that acute traumatic brain injury patients who showed EEG command-following had better functional outcome as early as 3 months after injury (Egbebike et al. 2022). On a more fundamental level, it has been long established that MCS has better functional outcome than the PVS/UWS (Bardin et al. 2011; Giacino and Kalmar 1997; Whyte et al. 2001). It is well known that withdrawal of life-sustaining treatment is a top cause of early death in acute DoC (Turgeon et al. 2011), and the decision is precisely made on prognostic estimations. Taken together, these results highlight the value of advanced neuro-investigations in establishing an early and accurate diagnosis, and thereby preventing the premature withdrawal of life-sustaining treatment.

Another important ethical issue is pain management. During clinical communication with relatives or carers, clinicians may feel the pressure to reassure that DoC patients are unable to experience pain and are therefore not suffering. Pain, like consciousness, is a subjective experience, but conventional physiological markers of pain, such as heart rate, have been shown to be unreliable markers (Laureys and Boly 2007). Remarkably, neuroimaging studies have shown evidence of pain perception capacity in some DoC patients (Chatelle et al. 2014; Claassen et al. 2024), suggesting that the possibility of false reassurance based on clinical gestalt only. As a result, some authors recommend that routine pain treatment should be universally given to DoC patients, to avoid unnoticed suffering as a consequence of misdiagnosis (Fins and Bernat 2018).

Neuroimaging studies, especially those demonstrating CF, also highlights the ethical issue of patient involvement and autonomy in care decision making. Despite logistic and resource limitations, it is important for clinicians to acknowledge the possibility of residual capacity to communicate. Other than diagnostic and prognostic implications (Monti et al. 2010), the potential of patients to communicate with their loved ones may hold personal and psychological significance, such as asking the patient about his/her subjective state of pain (Monti and Schnakers 2022) or wish regarding withdrawal of life-sustaining treatment (Chandler et al. 2017). This way, advanced neuroimaging may allow a regard for personhood that is otherwise unachievable based on a wrong behavioral diagnosis.

2.1 Concept of command following

CF is a phenomenon commonly assessed in neurobehavioral assessments and is also used in active paradigms such as task-based EEG. To be able to follow commands indicates that an individual must have a sufficiently intact network of consciousness, i.e., volitional capacity that allows the response to be made (Bruno et al. 2011; Claassen et al. 2024; Michael et al. 2021; Pan et al. 2020; Schiff 2006).

Covert CF is elusive to bedside examination and is detected by mental imagery imposed by active paradigms, such as task-based EEG or fMRI. The most commonly researched types of mental imagery are spatial and motor imagery (Monti et al. 2010). Spatial imagery involves requesting the patient to imagine navigating through a scene or space. Motor imagery involves requesting the patient to attempt a movement without its actual physical execution. To detect these mental responses, a patient is asked a series of closed yes/no questions that can be answered using specific mental imageries. The imagery responses are then retrieved as activations in pre-defined cortical areas (e.g., distinct activity in the supplementary motor area for motor imagery, or the parahippocampal gyrus for a spatial imagery) (Edlow et al. 2021a; Monti et al. 2010). For example, using spatial imagery, a patient can be trained to “imagine playing tennis” for a “yes” response, but to “imagine moving around a house” for a “no” response. Alternatively, using motor imagery, a patient can imagine “moving the right hand” “or moving the toes” to follow commands covertly (Fernandez-Espejo and Owen 2013). If the patient’s neural responses are consistent across multiple trials, he/she can be considered conscious.

Essentially, the detection of CF is unequivocal for preserved consciousness when positive (Cruse et al. 2011; Schiff and Fins 2016). A major limitation of investigations that involve CF paradigms is that it might be difficult for DoC patients to perform mental imagery tasks due to deficits in language comprehension, cognitive or executive function (Stender et al. 2014). Thus, the inability to perform a mental imagery task does not imply the absence of patient awareness (Gosseries et al. 2014a).

2.2 Electromyography

Electromyography is a method for identifying voluntary action in DoC, such as volitional micromovements which may be missed by clinical assessment (Habbal et al. 2014). Evidence has shown that EMG detected volitional responses to commands in 2 MCS patients who had no behavioral signs of CF (Lesenfants et al. 2016). In another study of 8 PVS/UWS patients, one patient demonstrated significant EMG activity suggestive of CF (Bekinschtein et al. 2008). Factors that limit the use of EMG to detect covert consciousness include its high false-negative rate, confounding from muscle spasticity, and inapplicability for patients with complete paralysis (Habbal et al. 2014; Lesenfants et al. 2016; Zheng et al. 2023).

2.3 EEG event-related potentials

Event-related potentials are specific changes in EEG signals in response to discrete events (Congedo and da Silva 2017). Generally, the longer the latency of EEG signals, the more suggestive of a complex and cognitive response to external stimuli (Davies et al. 2010; Jain and Ramakrishnan 2020; Lehembre et al. 2012). As such, late-latency ERPs involving fronto-temporo-parietal connectivity, such as P300 (including the P3a, P3b and nP3), N400, P600, have been proposed as potential EEG biomarkers for covert conscious perception. One important subset of ERP is the classic sensory (e.g., visual, auditory, somatosensory) evoked potentials (EPs), which can be administered using either passive or active paradigms. Evidence showed that novelty P3 parameters (nP3, part of the P300 complex) were able to detect CF utilizing an auditory subject-own-name (SON) active paradigm, where the patient was asked to count his/her own name. In the same study, a positive correlation of the amplitude of nP3 with CRS-R scores was also observed, suggesting that nP3 is able to distinguish between PVS/UWS and MCS (Risetti et al. 2013).

Overall, there remain limitations to the use of the ERP and EPs, and thus their utility to DoC diagnosis continues to be controversial (Giacino et al. 2018b; Zheng et al. 2023). This investigation technique might be limited by the heterogeneity of signal data and its dependency on signal interpretation (Jain and Ramakrishnan 2020; Risetti et al. 2013). Of note, the preference for late-latency ERPs has been challenged by evidence showing that the perceptual awareness negativity (PAN), an early to mid-latency awareness-related ERP signature, is a better marker than the late-latency P3b for awareness (Dembski et al. 2021; Forster et al. 2020). On the other hand, some have raised doubt on the PAN that it might not be a necessary marker and might be related to and influenced by attention (Bola and Doradzinska 2021; Koivisto et al. 2009).

2.4 Task-based EEG and fMRI

Task-based paradigms require a subject to perform a specific task on command, aiming to detect the willful, non-reflexive modulation of brain activities without explcit motor responses (Gosseries et al. 2014a). They have long been used to detect covert consciousness in DoC patients in research (Kondziella et al. 2016; Monti et al. 2010; Owen et al. 2006). Its unequivocal reflection of volitional capacity places it superior to resting-state or passive paradigms due to its capacity to circumvent confounding automatic brain activity. The results are also readily comprehended by clinicians and patient caregivers (Monti and Schnakers 2022). Of the different tasked-based investigations, EEG and fMRI are the most well-researched (Jain and Ramakrishnan 2020).

There has been strong and promising evidence for task-based EEG and fMRI. More than a decade ago, Cruse et al. (2012) found that task-based EEG detected CF in a patient who was diagnosed with PVS/UWS for 12 years, by using just 20 min of EEG recording, which is shorter than the time required to perform the CRS-R. For fMRI, the pivotal study by Owen et al. (2006) observed that task-based fMRI detected CF in a PVS/UWS patient, followed by a similar fMRI study that found that 17 % (four in 23) of PVS/UWS patients demonstrated CF (Monti et al. 2010). In another study, 15 % (four in 26) of patients with severe brain injury exhibited CF under fMRI despite having normal to mildly abnormal EEG findings, but with highly variable CRS-R scores (Forgacs et al. 2014). A meta-analysis revealed that 14 % of PVS/UWS patients showed CF using either task-based EEG/fMRI (Kondziella et al. 2016).

Although the CRS-R has been used as the gold standard to assess the multimodal investigations, its diagnostic sensitivity was only established through comparisons with another clinical scale – the disability rating scale (DRS) (Giacino et al. 2004). Interestingly, there is considerable overlap between their components, especially in terms of eye-opening (termed arousal in the original paper), communication and motor response (Rappaport 2005). The sensitivity/specificity of many multimodal investigations were in turn established with the CRS-R as a reference (Kondziella et al. 2016). Because of the lack of a true gold standard for consciousness, this degree of circular reasoning at the core of neurobehavioral scales is unavoidable. But for task-based paradigms as potential new reference standards, if they are unambiguously superior and have the potential to cause a definitional shift, comparing them against an imperfect current gold standard will result in false conclusions that limit their performance and utility (Gold et al. 2010). In this light, the question is not about whether to adopt the unambiguous task-based paradigms, but how.

2.5 TMS-EEG

Transcranial magnetic stimulation combined with high-density electroencephalography (TMS-EEG) is a non-invasive technique which measures the complexity of the brain’s response to perturbation. It serves as a marker of the level of consciousness in DoC, based on the theoretical idea that in disordered states of consciousness, there is disruption of local and global thalamo- and cortico-cortical information processing (Gosseries et al. 2014b; Julkunen et al. 2022). A TMS pulse is repeatedly administered to specific sites on the cerebral cortex, producing discharges at local and distant sites which are detected by high-density multi-channel EEG (Gosseries et al. 2014b; Massimini et al. 2005). Measures of cortical reactivity and connectivity are then derived from an analysis of the degree of specialization and integration of the EEG signals.

There has been promising evidence for TMS-EEG as a marker of consciousness. For instance, a study led by Massimini et al. (2012) found that wakeful subjects demonstrated long-lasting and long-range integrated EEG response to TMS; in contrast, PVS/UWS patients demonstrated a stereotypical strong local activation with poor propagation to adjacent areas and quick dissipation, whereas MCS patients demonstrated complex long-lasting widespread activation which propagates to distant areas (Gosseries et al. 2014b), similar to healthy awake subjects and REM sleep (Massimini et al. 2005, 2012). The technique has several advantages over resting-state EEG and ERPs (Ragazzoni et al. 2013), because it is perturbational instead of observational, which means that it bypasses the limitations imposed by lesions of motor, sensory, cognitive & language networks. It also has an advantage over task-based paradigms in that it requires no active participation from the subject being tested (Gosseries et al. 2014b).

Limitations of TMS-EEG include the need for specialized equipment, technical expertise and extensive preparations (Edlow et al. 2023; Julkunen et al. 2022). Further, false negatives cannot be ruled out, as cortical connectivity signals might relate to complex networks subserving other cortical activities, instead of the true circuitry underpinning consciousness (Ragazzoni et al. 2013). Thirdly, TMS may be unable to evoke measurable responses in brain-injured patients with a structurally damaged portion of the cortical surface (e.g. cortical lesions, skull breaches, sites of drain placement). Yet, this problem can be overcome by using image guidance (e.g. neuronavigation by structural MRI) to avoid the damaged cortical sites (Casali et al. 2013; Ragazzoni et al. 2013) and to identify the highly connected regions (Gosseries et al. 2014b).

2.6 PCI

The Perturbational Complexity Index (PCI) is an EEG-derived index of consciousness developed by Casali et al. (2013) that measures the amount of information contained in the integrated response of the thalamocortical system to TMS perturbation. It is produced by compressing the channel data from TMS-EEG using the lossless Lempel Ziv algorithm and quantifying them into a normalized index (Casali et al. 2013). Its value ranges from 0 (corresponding to the EEG response being minimally complex), to 1 (maximally complex).

There has been very promising evidence for PCI as a marker of consciousness. Firstly, it is a highly sensitive measure. In one study, it provides high sensitivity of 92 % in detecting MCS patients and identified a subset of 4 PVS/UWS patients with a potential for consciousness not expressed in behavior (Casali et al. 2013; Sinitsyn et al. 2020). In another study, an empirically obtained PCI cut-off (PCI*) of 0.31 discriminated between MCS and PVS/UWS with a sensitivity of 94.7 % (Casarotto et al. 2016). Secondly, like TMS-EEG, it completely avoids the false negatives of resting state, stimulation-based or task-based methods (Casali et al. 2013).

There are a few limitations to the use of PCI. Firstly, the lesion load might impose an upper bound to complexity, as evidenced by PCI < PCI* in some MCS patients (Casarotto et al. 2016). Also, complexity might also be affected by functional imbalances, e.g. thalamocortical bi-stability (a down-state after a period of activation) which transiently disables the brain from integrating information (similar to bistability during NREM sleep) (Massimini et al. 2005). Thirdly, there remains high operational difficulty and hence cost of obtaining high-quality EEG signals, neuronavigation for target localization, the associated nonparametric statistics and the use of the Lempel-Ziv algorithm (Edlow et al. 2023). In view of this, Comolatti et al. (2019) proposed a fast PCI (PCIst) algorithm which is not only faster than the Lempel- Ziv PCI, but also stable for TMS-EEG data with a small number of channels (Wang et al. 2022). Overall, it is thought that logistic and methodological limitations of PCI may be overcome by international efforts in refining the technique (Edlow et al. 2023).

3.1 Rationale of the framework

It is clear that a multimodal approach to improving the diagnostic accuracy of DoC is required (Coleman et al. 2009a; Kondziella et al. 2016; Owen 2019; Schiff 2006). The heterogeneity of study designs and the lack of an operationalized framework prohibit the adoption of auxiliary investigations (Farisco et al. 2022). Considering the strong evidence in support of the reliability of task-based EEG/fMRI (Owen et al. 2006; Schiff 2006), there has yet to be a clinical practice guideline that leverages and prioritizes the unique advantage of CF. In view of this, we propose a simple clinical decision framework focused on CF that streamlines the multimodal approach to DoC, with consideration that it can be supplemented by a highly sensitive technique such as the PCI.

3.2 Description of the framework

One of the key features of our framework is that we propose withholding the diagnosis of PVS/UWS made by the CRS-R for behaviorally unresponsive patients. The work-flow of our framework is summarized in Figure 1. When a patient with DoC is approached, he/she is first assessed clinically using the CRS-R (or equivalent) for any behavioral responses. If the patient is behaviorally responsive, a diagnosis of MCS can be made. If behaviorally unresponsive or uncertain, a diagnosis of PVS/UWS that would have been conventionally made by the CRS-R (Giacino et al. 2004) is withheld here. At this juncture, the patient will proceed to undergo task-based EEG followed by task-based fMRI, but other equivalent advanced imaging and electrophysiology techniques may be employed (Monti and Schnakers 2022). If CF is consistently detected (defined as in at least three out of four trials, taking reference from Giacino et al. (2004)) using either modality, a diagnosis of “at least MCS” is given (proposed by some to be labelled “non-behavioral MCS”, “MCS-star (MCS*)”, or “cognitive motor dissociation (CMD)”) (Bruno et al. 2011; Claassen et al. 2024; Schiff and Fins 2016; Thibaut et al. 2021). If CF is undetected, PCI may be used as a sensitive measure to overcome the issue of false negative of CF tasks, as it requires no active participation at all from the patient (Casarotto et al. 2016). This would form a diagnostic workflow which provides a logical and stepwise relief of functional demands from the patient, from motor to cognitive, at the expense of increasing technical difficulty.

Figure 1:

Conceptual framework for a multimodal approach to DoC with task-based EEG and fMRI. The lines and text in black illustrate a simplified version of the current practice using the CRS-R as the gold standard, which results in premature clinical diagnosis of PVS/UWS. The lines and text in red illustrate how multimodal investigations can prevent such a premature diagnosis. The inverted triangle to the left of the flowchart illustrates the relative demand of the test on the patient’s residual functions, with an emphasis on repeated assessments and the priority for task-based paradigms over other modalities. Importantly, the use of other unambiguous multimodal investigations is not ruled out. CF, PCI or other paraclinical evidence compatible with covert consciousness would yield a diagnosis of at least MCS before arriving at a final diagnosis of PVS/UWS. Abbreviations: CF, command following; CRS-R, coma recovery scale-Revised; DoC, disorders of consciousness; EEG, electroencephalography; fMRI, functional magnetic resonance imaging; MCS, minimally conscious state; PCI, perturbational complexity index; PVS/UWS, prolonged vegetative state/unresponsive wakefulness syndrome.

If the PCI is suggestive of unconsciousness, other multimodal investigations to detect signs compatible with covert consciousness should follow. If still no signs compatible with covert consciousness is detected, a diagnosis of PVS/UWS can finally be made. Ideally and where feasible, the test(s) at every step should be repeated (serial assessments) to increase their sensitivity (Owen 2019), similar to how the CRS-R can be repeated as appropriate (Monti and Schnakers 2022; Royal College of Physicians 2020).

Some points to note on interpreting results regarding CF are explained as follows. For CF in task-based EEG, it refers to above-chance classification of EEG signals (sensorimotor beta rhythms) after average spectral and single-trial analyses (Cruse et al. 2012); “no CF” refers to an absence of above-chance EEG classification. The reliability of the measurement is supported by the core principle of EEG that recordings determine altered states of consciousness by measuring changes in beta rhythms, detected through either event-related desynchronization (ERD) or event-related synchronization (ERS); ERD (a reduction in power of oscillations) is detected on the contralateral scalp and ERS (an increase in power of oscillations) on the ipsilateral scalp, with respect to the limb whose movement is imagined or planned by the patient (Cruse et al. 2012).

For CF in task-based fMRI, it refers to above-chance activation patterns in scan images that are comparable to patterns from healthy controls or compared to resting state, as assessed by suitable statistical analysis, e.g., image pre-processing with statistical parametric mapping coupled with random effect analysis (Stender et al. 2014); “no CF” refers to an absence of such above-chance activation patterns. The reliability of the measurement is supported by the core principle that when a DoC patient performs an imagery task, any task-related fMRI signals consistent with those observed in healthy individuals subject to the same task may serve as a proxy for unexecuted motor behaviors (Fernandez-Espejo and Owen 2013).

3.3 Advantages of the framework

We designed our simple framework with the following justifications. First, this framework identifies CF as a key biomarker for covert consciousness. CF, with its ability to reveal subclinical consciousness (Owen 2013), is especially helpful for patients of DoC with poor behavioral responses resulting from focally severe motor efferent or cortical insults, but with preserved consciousness (Forgacs et al. 2014). As for the choice of measurement tools, among the many candidates of non-behavioral markers, we have specified the use of task-based EEG and fMRI because they are the most well researched investigations associated with active paradigms in the literature (Jain and Ramakrishnan 2020). More importantly, many paradigms, such as resting-state and passive paradigms, suffer from uncertain interpretation of results related to different definitions of consciousness or the NCC, which is still open to much debate. By contrast, CF has the unique nature that a positive result unequivocally indicates preserved volitional capacity, which in turn reliably implies the presence of covert consciousness, as discussed previously. In short, CF completely frees the test of the problem of interpretation, which may confer the framework reasonably greater potential to be translated into routine clinical practice.

Second, our framework differs from other multimodal frameworks in that it highlights the priority for task-based paradigms over other paradigms. Although the need for a multimodal approach is overdue (Dale and Halgren 2001), there is limited effort in translating evidence from research into clinical practice through a feasible algorithm or protocol (Farisco et al. 2022; Schiff 2006). Multiple proposals of multimodal frameworks have indeed been previously made (Coleman et al. 2009a; Giacino et al. 2018b; Monti and Schnakers 2022; Zheng et al. 2023), but to the best of our knowledge, none has specifically emphasized and prioritized the use of task-based paradigms. Further, some protocols may recommend a wide array of multimodal investigations, ranging from structural MRI, to TMS-EEG or task-based EEG (Cavaliere et al. 2018; Willacker et al. 2022), or include tests whose diagnostic significance is uncertain (Coleman et al. 2009a). It is true that more tests should in principle improve overall reliability of the diagnoses, but to perform every test available, at the expense of the clinician’s valuable time, may not necessarily lead to diagnoses that are dramatically more accurate. Our simple framework attempts to balance the need to streamline the workup process for the clinician against the benefit of improved accuracy of diagnosis powered by multimodal investigations. In sum, this often-neglected perspective of prioritizing task-based paradigms might have the potential to develop into an effective and practical clinical algorithm, if corroborated by further clinical research.

Third, our framework specifies the relation of multimodal investigations to clinical assessments based on practical considerations. Our framework identifies the order of conducting the clinical/multimodal tests. Specifically, the CRS-R should be performed first as a screening tool, for its low cost and relative ease of administration (Coleman et al. 2009a; Giacino et al. 2014; Kondziella et al. 2016; Zheng et al. 2023), but the use of CRS-R does not preclude the use of alternative clinical scales of comparable reliability. Task-based paradigms (CF) should follow next, for reasons explained above. Of this, EEG-based CF should precede fMRI for its relative low cost, point-of-care access and excellent temporal resolution (Faugeras et al. 2011; Fernandez-Espejo and Owen 2013; Kondziella et al. 2016; Zheng et al. 2023). The use of a highly sensitive technique, such as the PCI, may be used next to overcome the issue of false negatives, if the above task-based paradigms yield no evidence of CF. It is well established that task-based paradigms and PCI should be used together rather than exclusively in a multimodal approach (Kondziella et al. 2020). Although the PCI has the unique advantage of relieving the patient of not just sensori-motor but also cognitive demands (Casali et al. 2013), here we propose to use it after task-based paradigms owing to its especially high technical and computational demands as discussed above, while noting that these limitations might be overcome in the future (Edlow et al. 2023). The diagnostic process is then followed by other valuable but potentially ambiguous multimodal investigations, such as ERP and EMG, the inclusion of which would help minimize the false negatives of active paradigms. However, for any paradigms, it may be impossible to rule out false negatives until a true NCC is confirmed, which can only be solved by further basic science research on the NCC.