Eye tracking as a novel approach to cognitive assessment in DOC patients based on evidence from a multicenter clinical trial

Abstract

The proper assessment of cognitive functioning is the subject of numerous scientific studies. Current behavioral methods essentially preclude this assessment in patients with disorders of consciousness (DOC). However, the development of eye-tracking technology (ETT) and the establishment of contact via this channel have created an opportunity to diagnose the cognitive function (CF) of DOC patients. The purpose of this study was to assess the level of CF of DOC patients for whom vision is the only channel of communication to introduce the CF assessment tool for DOC patients for whom vision remains the only channel for communication. The clinical multicenter study involved 31 DOC patients, whose attention, language functions, visual-spatial functions, personal orientation, memory, and abstract thinking were assessed three times (T1-T3) using the Cognitive Functions Assessment (CFA) scale, installed on the C-EYE X system. The collected data were compared with the CF assessment results obtained with the Coma Recovery Scale-Revised (CRS-R), and then statistically analyzed. There were no statistically significant differences between different time points. Patients scoring higher on the CRS-R receive higher values on the CFA. Statistically significant and moderate correlations were found between the total CFA and CRS-R scores. The results of the study indicate that the diagnosis of CF made with the use of the CFA allows for an accurate assessment of a wide spectrum of CF as well as makes their assessment independent of the experience of the examiner and the cooperation of the patient. The use of ETT, without reducing the quality of the examination, may reduce the workload of qualified personnel. The study was registered on the ClinicalTrials.gov clinical trials platform ID NCT05536921.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-27996-6.

Introduction

A reliable and detailed neuropsychological diagnosis is of vital importance in the process of patient recovery¹. At the core of the diagnostic assessment are preserved cognitive functions (CF), because they are essential for information processing. They include ability to recognize people and situations and assign the correct meaning to them². Memory also plays a key role, because is important for learning, performing daily tasks, and maintaining social relationships³. The ability to learn quickly and permanently is the basis of adaptation to changing conditions and situations in which people find themselves⁴. In turn, proper adaptation requires accurate assessment and analysis of the collected information and the ability to apply in both simple and very complex situations^5,6. The appropriate level of preserved CF is also important for maintaining mental health, as it helps to cope with stress, anxiety, and depression⁷.

The disorders of consciousness (DOC) is a group of clinical conditions that pose real challenge to accurate diagnosis of CF⁸. Core symptoms of DOC include severe alteration in awareness and arousal, and also serious reduction of communication abilities. There are three types of DOC: coma, unresponsive wakefulness syndrome/vegetative state (UWS/VS), and minimally conscious state (MCS). Briefly, criteria for distinguishing these are based on the presence of appropriate responses to sensory stimulation and commands, indicating awareness of self and surroundings (for a more comprehensive description of DOC, see Zhang et al., 2025) Recently, MCS has been subdivided into MCS- and MCS+, with the latter diagnosed when there is evidence of language abilities⁹. The wide range of deficits, affecting many cognitive domains, appear as a natural consequence of disturbances in the state of consciousness and the assessment of CF is particularly challenging when it must be done for DOC patients who are unable to communicate verbally, yet they seem to regain awareness, by fulfilling the Aspen Workgroup criteria for minimally conscious state¹⁰.

The clinical condition of a DOC patient is directly related to the location and extent of structural brain damage, and this fact determines the prognosis and the extent of potential cognitive recovery. Due to the high level of misdiagnosis of DOC patients, CF assessment should not be limited to observations of the overt spontaneous behaviors and responses to environmental stimuli¹¹. There is no doubt that assessing CF can serve as an indicator of the degree of brain dysfunction and it can predict possible recovery of consciousness¹². Utilization of the preserved CF offers a basis for cooperation in the rehabilitation process, while at the same time, they are also a conduit for the use of assistive communication technologies (e.g. eye-tracking technique)¹³.

For DOC patients, even a basic ability to process information is important both for the quality of the health care and for the assessment of their clinical condition¹⁴, and may affect decisions regarding care and therapy^15,16. Diagnosing the CF of DOC patients poses special challenges, as it requires going beyond assessment relying only on observing behavioral responses. Many solutions have been proposed so far, including fMRI, PET, and EEG¹⁷, interviews with family and caregivers¹⁸, the therapist’s subjective assessment¹⁹, as well as assessments of specialists (i.a. neurologists, psychologists)²⁰. A significant segment of assessment is the observation of responses to sensory stimulation²¹, that can be effective in assessing baseline CF²². Another useful indicator is the assessment of procedural memory²³, spatial working memory²⁴, motor abilities and visual perception²⁵, as well as observational scales for DOC patients²⁶. The advantage of the above-mentioned methods of assessing CF in patients with DOC is that they are widely known and used in clinical practice. However, these methods have their limitations, which mainly include either the exclusion of the patient from the process of assessing their cognitive functioning (the patient remains passive during such an assessment) or a high risk of subjectivity in the assessment of cognitive functions, as well as the high cost of technical equipment and serious technical challenges related to approaches based on monitoring of brain activity, such as EEG, PET or fMRI. In the group of patients who are unable to communicate verbally, eye-trackers have for some time been used to enable them to establish contact with their surroundings. Further development of this technology has allowed for new applications of eye-trackers in assessing various CF in neurological patients, including those with DOC²⁷. In multicenter clinical trials eye-trackers have been used to assess functions such as attention²⁷, visuospatial function^28,29, language function³⁰, and memory³¹. The results of this studies, conducted with patients in a UWS, MCS, and eMCS state, indicate that (1) oculomotor training can help restore visuospatial functions²⁸, (2) the level of language functions examined using an eye tracker suggests that their level of consciousness is higher than that assessed using traditional methods³⁰, (3) memory impairments in people in these states most affected working memory³¹. These latest reports are consistent with earlier studies, which show that the way images are visually processed, as well as discernible gaze patterns, can provide information on working memory processes^32,33, and help assess working memory in patients with aphasia and other communication problems^34,35.The purpose of this study was to evaluate the clinical utility of the tool based on eye-tracking technology in the assessment the cognitive profile of DOC patients. Verification of the tool’s diagnostic properties will provide the basis for its further development, with the aim of ensuring the most accurate assessment of cognitive functioning in patients for whom vision is the only channel of communication.

Material and methods

Participants and study design

Approval for the research was granted by the Senate Committee on Research Ethics at the Wroclaw University of Health and Sport Sciences (Decision 11/2022), as well as being aligned with the regulations of the 1975 Declaration of Helsinki. The study was registered on the ClinicalTrials.gov clinical trials platform ID NCT05536921.

This multicenter study was conducted between July 2022 and February 2023 across five Polish centers (located in Krakow, Sawice, Częstochowa, Wrocław, and Zawiercie) and one German clinical center (University Clinical Hospital Oldenburg). The study enrolled centers’ patients with severe brain injury who were undergoing neurorehabilitation. A positive medical assessment conducted by a neurologist based on an analysis of the patient’s medical records and the meeting of the inclusion criteria, was a prerequisite for participation: (1) age ≥ 18 years, (2) legal guardian’s consent to participate in the study and access to medical records, (3) a medical diagnosis indicating damage to the central nervous system (CNS), (4) possession of and access to an imaging examination reports (MR or CT) and, as an alternative, of the oculometric test, ophthalmic examination, and hearing assessment, (5) providing a list of medications used by the patient that may affect the results obtained in tests of cognitive function, (6) physician’s approval (e.g., neurologist, neurosurgeon, internist) for participation in the clinical trial, following the review of the study protocol, including the ability to communicate solely thorough eye-gaze interaction (no verbal, sign language, or other forms of communication possible), the absence of dementia and aphasic disorders prior to the event that caused the CNS damage and the patient’s current condition, preservation of at least one functioning eyeball (ability to establish cooperation with an eye tracker).

The exclusion criteria were: (1) a vision impairment (refractive error) diagnosed before the injury, requiring the use of glasses with lenses of more than ± 3 diopters, (2) the inclusion of pharmacological treatment during the study (observation), which may affect the patient’s cognitive functioning – both in terms of an increase in cognitive abilities, as well as their impairment/dementia.

A total of 68 patients were recruited for the study (66 in Poland and 2 in Germany). Due to the small number of patients, Oldenburg Hospital/Germany was excluded from the study. The remaining patients underwent medical screening, and 19 were excluded either for not meeting the inclusion criteria or for meeting the exclusion criteria (Fig. 1). Of the 47 patients who qualified for further procedure, 31 (66,0%) underwent successful calibration to assess their cognitive function. Time since injury is presented in Table 1.

Fig. 1.

Flow chart of study enrolment, allocation, and analysis.

Table 1.

Basic demographics and clinical data patients qualified for the clinical study.

Category					N	%
Sex		Men			16	51,6
		Women			15	48,4
Nature of the brain injury
Craniocerebral trauma					10	28,6
Stroke					11	31,4
Hypoxia due to cardiovascular causes					6	17,1
Hypoxia due to respiratory causes					4	11,4
Status epilepticus					2	5,7
Metabolic disease					1	2,9
Other					1	2,9
Variable	Mean	SD	Min	Q1	Median	Q3	Max
GCS	6,9	1,9	3,0	6,0	7,0	8,0	11,0
Time since injury [months]	15.7	28.0	0	3.0	6.0	14.0	150.0

The aetiology of brain damage in the patients whose results were analyzed varied; the most common causes were stroke (31,4%) and craniocerebral trauma (28,6%). Details of aetiology and demographics obtained from medical records are shown in Table 1.

To better illustrate the state of consciousness of the patients included in the study, the GCS score (Glasgow Coma Scale) was included in the description of their clinical condition. It is used in clinical centers in Poland and other European countries to diagnose patients with chronic DOC and those in the acute phase of the disease. The range of variability in the score that a patient can obtain ranges from 3 to 15 points; the lower the values, the more severe the disturbance of consciousness³⁶.

Study procedure and data collection methods

The level of cognitive functioning of participating patients was assessed using the proprietary Cognitive Functions Assessment (CFA) scale, installed on the C-EYE X device.

The scale development consisted of multiple stages. In the first stage a group of experts with experience in neuropsychological research (including the authors of this paper) prepared test proposals. During a series of meetings between experts and software developers, the software functionality was refined by adjusting appropriate parameters. External expert (neuropsychology, neurorehabilitation specialists) opinions were also consulted. During the second of test development involved conducting a pilot study with 15 patients during which software stability, hardware functionality were tested, and test content was validated, along with the method of calculating and collecting results. As a result of the pilot studies, significant revisions were made to the CFA test, removing some tests and introducing substantial changes regarding stimulus presentation methods and scoring rules.

The CFA scale consists of a set of clinical tests designed touse the meaningful eye movement responses to assess the level of cognitive function. Once it was developed, it was adapted into an eye tracker for patients who do not communicate verbally. The examples of the tasks (screenshots) and a comprehensive description of the CFA scale items, including task descriptions and scoring criteria, is provided in S the Supplementary materials (see, Fig. X1 and Table X1).The C-Eye X consists of a 19-inch moving screen mounted on a tripod, so it is possible to rotate the screen to place it 60 cm in front of the subject’s face. The device does not require the subject’s head to be stabilized, as it can compensate for small head movements. The infrared-emitting illuminators are located below the monitor screen and do not interfere with the patient’s use of the device. The technical parameters of the C-Eye X are as follows: sampling rate of 33 Hz, accuracy of 0.5 degrees of visual angle, speed threshold of 40 cm/s. Because the C-Eye X is tailored to patients who have one or both eyeballs functioning, either monocular or binocular eye-tracking mode can be used. Each test was preceded by a single-point calibration to determine the location of the subject’s eye fixation point (the 2D image from the IR camera is processed on the device’s screen). The eye tracker determined the direction of the user’s gaze - i.e., the eye fixation point the user was looking at based on the location of the center of the pupil and two infrared reflections (corneal reflections, called “glints”) on the cornea of the eye. Calibration consisted of the patient looking at and keeping his or her gaze on a flashing red dot, surrounded by a white border, which was presented in the center of the screen. The fixation of the patient’s gaze lasted no more than 10 s and was considered complete by the device when the system detected the correct fixation of the patient’s gaze on the calibration dot. If proper gaze fixation did not occur during this time, the system aborted the calibration process, informing the patient and the examiner that the calibration had failed and should be repeated.

The system recorded the patient’s response if the dwell time on the target area exceeded 1.2 s. This value was defined empirically during prior validation work and internal observations with neurological patients after brain injury. It represents a balance between (a) avoiding accidental selections from brief or unintentional glances and (b) enabling patients with visual or attentional deficits to reliably confirm their intended choice.

The initial visit (visit 1) was aimed at collecting basic sociodemographic and clinical information about the patient, including establishing information on the results of the state of consciousness upon admission to the rehabilitation centers and the diagnosis made. Visits 2–4 (designated hereafter as measurement points T1, T2, T3) were devoted to assessing the study patients cognitive functions using CFA; each visit was preceded by the administration of CRS-R. The cognitive function testing protocol consisted of administering CFA three times over 14 days, with intervals of at least 1 day, to allow recovery of patient after the assessment which might be demanding for this group of patients. Moreover, including multiple CFA assessment within the fourteen days period in the study protocol followed the recent recommendations of professional bodies related to diagnosis of DOC^37,38. The examinations at each center were conducted by the neurorehabilitation specialists trained in working with an eye tracker and certified in the administration of CRS-R. Standardization of data collection by researchers from different centers was ensured through joint training sessions and compliance checks, which were organized and supervised by an international expert in CRS-R data collection. The duration of one patient testing session did not exceed 60 min and depended on the level of cooperation between the patient and examiner.

Each CFA administration was preceded by an assessment of the patient’s level of neurocognitive functioning done with CRS-R]. CRS-R is a recommended tool for evaluating patients with chronic DOC^38–40. CRS-R enables differential DOC diagnosis, prognostic evaluation, and treatment planning], and its Polish version was validated⁴¹. The scale is organized into 6 functional subscales—auditory, visual, motor, oromotor, communication, and arousal—containing 23 hierarchically structured items. Each subscale ranges from basic reflexive responses at the lower end to complex cognitively mediated behaviors at the upper end. The possible total score ranges 0–23 points reflecting general neurocognitive abilities of the DOC patient, and the diagnostic classification is based on individual subscale performance.

Clinicians administering the CFA, and CRS-R assessments were authorized to input results to the online data repository following each assessment session. Typically, however, the principal investigator handled data submission after each visit. All clinicians received training and were knowledgeable about the study hypotheses. The study design and required execution standards minimize clinician-observer bias risk, with the principal investigator maintaining oversight of all procedures.

The last visit (no 5) was dedicated to clinical diagnosis performed by the attending physician. It included clinical interview using the provided patient evaluation form, and required examining the patient, assessing general condition, documenting treatment changes, noting medical problems, verifying adverse events, and concluding observation.

The collected patient data was processed using the GoInsights™ platform, which meets all FDA 21 CFR Part 11 and GCP 5.5.3 requirements for electronic data. The safeguards used in GoInsights™ as well as the quality procedures for data collection minimize the risk of incomplete data entry. All data has been recorded in the electronic patient record (eCRF).

Data analysis

The main aim of the data analysis was to examine concurrent validity of the CFA scale examined with CRS-R results as the reference standard. The secondary aim was to evaluate the essential psychometric indicators of the scale. The actual study was preceded by the preparation of a Statistical Analysis Plan. The distribution of variables was evaluated using a quantile-quantile (Q-Q) chart and the results of the Shapiro-Wilk test. Depending on the distribution of the data, the mean and standard deviation, median and lower and upper quartiles, as well as maximum and minimum values were calculated. Homogeneity of variance was assessed using Levene’s test, assuming a statistical significance level of α < 0.05.

The raw CFA and CRS-R scores were normalized to the percentage of the maximum possible total score for each tool (denoted as total CFA score % and total CRS-R score %, respectively).

For each subsequent visit (T1-T3), the percentage changes in total and subscale scores compared to T1 are presented. Measurement equivalence of the methods used was assessed using Passing-Bablok regression. This regression compares the slope of the linear regression with the value of 1 and the intercept of the linear regression with the value of 0. If the 95% confidence interval does not include 1 or 0, respectively, this means that there is a significant bias between compared tests. Passing–Bablok orthogonal regressions were performed for each time point and for the combined T1–T3 score, defined as the percentage of the maximum possible result across the three visits. All analyses were performed by the experienced biostatistician using the R statistical package, ver. 3.6.3⁴².

In order to examine the essential psychometric indicators of the CFA scale, the internal consistency of the subscales was evaluated using Cronbach’s alpha, and the test-retest reliability of the subscales was assessed using intraclass correlation coefficients (ICC). The internal consistency and reliability analyses were performed using custom MATLAB scripts⁴³ and the ICC add-on⁴⁴.

Results

Table 2. shows the sums of the patients’ scores at each time point, the raw change between T3 and T1, and the sums of the totals for all tests. There were no statistically significant differences in mean values between different time points (T1 vs. T2 vs. T3) for each test. There was no difference between CFA and CRS-R in the raw change between time points T3 and T1, nor in the overall percentage change across all time points (2.42% ± 10.08% vs. 0.28% ± 5.38%; p = 0.27), in the percentage change from T1 to T3 (T1 - T3; 41.67% ± 22.03% vs. 54.32% ± 29.77%; p = 0.18).

Table 2.

Comparison of the CFA with the CRS-R and the ANOVA-type results of repeated measurements comparison between the CFA and the CRS-R.

Time	CFA Total % points	CRS-R Total % points
	Mean ± SD Median (Q1 – Q3)
T1	40.86% ± 23.56% 37.50% (25.00% − 47.92%)	54.14% ± 29.82% 52.17% (26.09% − 78.26%)
T2	40.86% ± 23.11% 33.33% (25.00% − 45.83%)	54.42% ± 30.72% 56.52% (23.91% − 80.43%)
T3	43.28% ± 21.29% 37.50% (31.25% − 47.92%)	54.42% ± 29.68% 52.17% (21.74% − 80.43%)
Relative % change T3 vs. T1	2.42% ± 10.08% 4.17% (−4.17% − 12.50%)	0.28% ± 5.38% 0.00% (−4.35% − 0.00%)
Sum (T1-T3) of total % points	41.67% ± 22.03% 31.94% (30.56% − 45.83%)	54.32% ± 29.77% 52.17% (22.46% − 79.71%)
Factor	F ratio	p
Test type	3.576	0.0640
Time	1.312	0.2793
Interaction Test x Time	1.064	0.3535

The next step in the analysis was a two-way ANOVA to see if there were statistical differences between the CRS-R total % results and theCFA Total % score (Test type factor) concerning time points (Time factor). We observed a difference between CFA and CRS-R that did not reach statistical significance (main effect of Test type: p = 0.06), while there were no significant effects of Time or interaction between Time and Test type (Table 2).

Pearson correlation coefficients between the time points both for CRS-R and CFA scales are shown in the Table X2 (Supplementary material).

To estimate the co-variance level between two measurement diagnostic methods (CFA and CRS-R) used in the study, Passing-Bablok regression was used, which, without taking into account the causal relationship between them, compares the slope of the linear regression with a value of 1 and the intersection of the linear regression with a value of 0. The results of the statistical method used for each time point are shown in Table 3.

Table 3.

Results of Passing-Bablok regressions between the CFA and the CRS-R at each time point and for the sum T1-T3.

Time-point	β0	± 95% CI for β0	β1	± 95% CI for β1
T1	4.77	−14.66–16.67	0.68	0.38–0.96
T2	12.50	−4.17–45.88	0.48	−0.37–0.85
T3	21.49	3.33–52.10	0.44	−0.32–0.800
Sum T1-T3	16.22	−6.03–27.23	0.47	0.14–0.92

β₀ – regression coefficient (intercept), β₁ – regression coefficient (slope), CI – confidence interval, significance level p < 0.05.

For measurement points T1 and T2, there was a significant bias between both tests, increasing with increasing test values (β1 = 0.68 and β1 = 0.48, respectively). In addition, for the sum T1-T3, there was a significant bias between both tests, increasing with increasing test values (β1 = 0.47) (Figs. 2, 3, 4 and 5). This means that patients scoring higher on the CRS-R scale, despite positive correlation between both scales receive progressively lower values on the CFA scale. Statistically significant and moderate correlations were found between the CFA scale and the CRS-R scale at time point T1 and for the sum of scores from time point T1 to time point T3.

Fig. 2.

The Passing-Bablok regression between the CFA and the CRS-R scores at the T1 time point (p_corr<0.05).

Fig. 3.

The Passing-Bablok regression between the CFA and the CRS-R at the T2 time point (p_corr=0.09).

Fig. 4.

The Passing-Bablok regression between the CFA and the CRS-R at the T3 time point (p_corr=0.09).

Fig. 5.

The Passing-Bablok regression between the CFA and the CRS-R for the sum from T1 – T3 (p_corr<0.05).

In order to evaluate the internal consistency of the CFA subscales, Cronbach’s alpha was calculated for each subscale after averaging subscale items across the three time points. The detailed results are presented in Table X3 (Supplementary materials). The internal consistency of the subscales varied, with Attention, Language Skills, Memory, and Abstract Thinking showing good to acceptable reliability. For the remaining subscales—Visuospatial Functions and Personal Orientation—we obtained questionable and poor estimates of reliability, respectively.

In order to evaluate test-retest reliability of the subscales, intraclass correlation coefficients (ICC 3,1), calculated across three time points, were used. The detailed results are shown in Table X4 (Supplementary materials). The test-retest reliability also varied across the subscales, with Attention showing good reliability, while Language Skills, Memory, and Abstract Thinking exhibited moderate reliability. The lowest reliability scores were obtained for Visuospatial Functions and Personal Orientation.

Discussion

The dynamic development of modern technologies, including those used in the medical field, creates entirely new opportunities for diagnosing and collaborating with patients. This is especially important in emergency medicine and those areas where previously used methods of diagnosis and therapy of patients, especially those who until recently were not given a chance to survive and live, are slowly becoming irreplaceable because they accelerate the establishment of accurate diagnosis and decision-making^45,46. The prerequisite for proper therapeutic management is the establishment of an accurate diagnosis, which for the personnel caring for the patient then becomes the basis for treatment plan, including further therapeutic program⁴⁷. Making an accurate diagnosis assumes particular importance in the case of patients (regardless of their age) with whom there is no verbal contact, whether they are children, patients of various ages who remain in a DOC state, or those who have suffered permanent brain damage that prevents them from verbal contact. Hence, among other things, comes the constant search for such methods of diagnosis that will be as objective as possible, but will also make the diagnosis independent of the examiner. Procedures used in this way significantly increase the chances of the accuracy of the decisions made, which are influenced, among other things, by such factors as the appropriate high level of staff training and minimizing subjectivity in the assessment of the patient’s condition⁴⁸.

Problems with measurement accuracy include the traditionally used GCS scale, which is being replaced in an increasing number of countries around the world by the much more accurate CRS-R. This offers opportunities to reduce misdiagnosis rate, the percentage of which for brain-injured patients who do not communicate verbally reaches 40%^11,49. However, it should be noted that both the GCS and CRS-R are performed by medical personnel, so it is up to the experience of these personnel to determine whether the limitations of the GCS and CRS-R, which are known from the literature^50–53, will affect the finalscore that the patient as a result of the diagnostic process.

The solution proposed in this paper is to assess the patient’s cognitive function status with the CFA scale using eye tracking. The results of assessing these functions in patients who do not communicate verbally were compared with another scale that has a similar purpose, but different diagnostic sensitivity and measurement technique, the CRS-R tool.

The first point to note is that a comparison of the two methods of assessing CF (cognitive functions) indicates that the differences between them are not significant. This is important information from the point of view of further inference, however, a deeper analysis of the obtained results provides cognitively interesting information. Outliers are more common; this may suggest either the greater sensitivity of the CFA scale, or that a certain level of cognitive function (as indicated by total CRS-R % scores at level 50%) is required to obtain a meaningful CFA result. Below it, the CFA appears to manifest a floor effect. However, there may be another reason, and it may be related to the level of training of the testing staff. When the same patient is evaluated by the same examiner several times, there may be a routine in evaluating the patient with the CRS-R. Assuming, however, that the examiners made every effort to make the results of both tests as accurate as possible, it should be assumed that the assessment of CF with the CFA seems to be more reliable, also taking into account the fact that it is an assessment independent of the examiner. In the case of patients with severe brain damage who do not communicate verbally, for whom vision is the only channel of communication, the high variability of CF scores has already been shown in earlier studies. Some authors suggested that it could be related to both the emotional state of the patient and, for example, meteorological factors^31,54. In addition, it is evident that the level of consciousness in DOC patients is not constant but may fluctuate, thus affecting the results of a CF test on a given day of diagnosis, or therapy⁵⁵.

In comparing the two methods used, it can also be noted that only at T1 are the results quite similar, while at subsequent iterations of the CF diagnosis (time points T2 and T3), the regression equation curves show that the two methods differ to a greater extent, with the level of CF determined using CFA being at a lower level compared to CRS-R as the results obtained at this diagnosis increase, yet the linear relationship between both tools remains significant. The variability in the slope of the regression curves is also an important issue, this may be an effect of measurement instability; on the one hand, the CRS-R results are more variable (higher SD T1, T2 and T3), but on the other hand, there may be a floor effect in the CFA, as the test was more difficult to perform correctly for a large group of patients (especially those who obtained scores below 50%). This gives rise to the assumption that CFA is a method that more cautiously assesses the level of CF. This is especially true when considering it as a “mechanical” and therefore somewhat depersonalized way of assessing a patient’s condition. Shifting the focus from the examiner to the mechanical device poses some challenges, not least in terms of the accuracy of the diagnosis for further use. Evaluating the above-described fact from this perspective, it is possible to take the diagnosis made with the eye tracker as safe from the point of view of patient care, especially in such a severe condition. Until recently, patients with severe brain damage were treated as individuals for whom no specific diagnostic and therapeutic measures were taken. However, successive scientific reports, gradually revealing the world of their inner experience, even despite a low or very low assessment of the state of consciousness made with the GCS or other behavioral scales, show that this is a group of patients to whom much more attention should be paid in clinical measures. Such an approach also meets the expectations directed by medical personnel and patients’ caregivers to include the widest possible variety of methods of assessing the patient’s clinical condition, which will allow the most objective assessment of the patient’s state of health^56,57. It is worth mentioning here that a separate challenge is to implement into clinical practice an objective method of diagnosing the state of consciousness and cognitive profile of a non-verbal patient in the shortest possible time. This is particularly important because the multidisciplinary medical personnel help to make the best decisions in intensive care units (ICUs). The use of high-tech tools based on, among other things, eye tracking, such as the C-EYE X system used in this study, is a good predictor in this context.

The obtained psychometric indicators reveal considerable variability in the internal consistency and test-retest reliability of the test subscales. The profile of these indicators suggests, on the one hand, that multiple testing is be necessary to yield diagnostically valuable results, and on the other hand, highlights the importance of qualitative analysis, as individual items may provide critical insights into the patient’s neurocognitive abilities.A particularly important issue, both from the perspective of the patient himself, as well as from the perspective of relatives and staff who care for him, is also the contact that is established with the patient using the eye tracker. This contact facilitates not only communication and diagnosis of the patient’s state of consciousness⁵⁸, but also, as it turns out, diagnosis of the patient’s level of cognitive functioning. This opens up a whole new field of discussion about the patient, who ceases to be an object of action and becomes a subject for the multidisciplinary team dealing with him. Thus, he significantly enters the role of a full participant in the treatment/clinical procedure, which is enshrined in the patient’s Bill of Rights.

Limitations

The comparison of methods for diagnosing cognitive function made in this study, in addition to promising predictions for further diagnostic development, however, carries certain limitations. We realize that the size of the study group, although collected in a multicenter study, dictates caution in interpreting the results obtained. We therefore view the present study as a significant starting point for further clinical research in this area. The question of the patients’ ability to work with the eye tracker, which may have been relevant to the results obtained in the study, also remains open. Moreover, as evidenced in the inclusion and exclusion criteria, this technique is suitable for diagnosing only a subset of eligible patients, specifically those with a reasonably high level of residual visuomotor abilities. To this end, in future studies, the patient will be familiarized with the eye tracker beforehand, which will help in adapting the muscles of his eyeballs to work with the device, and thus may help in a more adequate assessment of the CF condition.

Conclusions

1. The eye tracking technology used for the study may be stricter method of CF diagnosis at higher consciousness level, and its results are comparable to the methods used so far. Despite being based only on visual interaction it seems to be a suitable screening tool giving an overall picture of the cognitive functioning of a patient who does not communicate verbally.

2. The diagnosis of CF by means of CFA seems to take more account of the variation of these functions in different patients, as well as to a much greater extent than previously used methods, making their assessment independent of the experience of the examiner and the cooperation of the patient.

3. The use of eye movement tracking technology for CF assessment has the potential to reduce the cost of working with patients by reducing the workload of skilled personnel while maintaining the high accuracy of this assessment.

4. Accurate diagnosis of CF patients who do not communicate verbally, for whom the only channel of communication is eye-gaze interaction, leads to a change in the understanding of their role in the process of diagnosis and therapy. In this approach, the patient ceases to be only an object of the measures taken and becomes a full participant in clinical practice.

5. CFA tool seems to work better for patients with higher CRS-R scores, for those who obtain lower results, the floor effect was observed.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conception and design of the study (GZ, KKN, BK, MB), acquisition and analysis of data (GZ, KKN, BK, MB), drafting of the manuscript or figures (GZ, KKN, BK, MB), study supervision (GZ, BK, MB).

Funding

The research was co-funded by the Polish National Center for Research and Development, grant number POIR.01.01.01-00-2125/20.

Data availability

The datasets used and/or analysed during the current study available to qualified researchers from the corresponding author on reasonable request.

Declarations

Competing interests

All authors received the compensation provided by the NCBiR grant according to their involvement in the implementation of this study. Each author exercised the utmost care and diligence to ensure that the course of the study was of the highest scientific value.

Ethics approval and consent to participate

Approval for the research was granted by the Senate Committee on Research Ethics at the Wroclaw University of Health and Sport Sciences (Decision 11/2022). The study was registered on the ClinicalTrials.gov clinical trials platform ID NCT05536921. Informed consent was obtained from the legal guardians of all subjects involved in the study.

Consent for publication

All the authors have approved the manuscript and agree to submit it to a scientific journal.