Spontaneous Eye Blinking Evaluation for Cognitive Assessment of Individuals With Severe Acquired Brain Injury

Overall aim: to provide preliminary results on the value of spontaneous eye blinking features (i.e., rate, amplitude, duration, and variability in intervals between blinks), which do not rely on voluntary behavioural responses, for improving the assessment of attentional abilities and consciousness in patients with sABI.

Specific aims:

1. to confirm the diagnostic value of EBR in pDoC patients, as in Magliacano et al.'s (2021) study;
2. to improve diagnostic accuracy of patients with prolonged Disorders of Consciousness (pDoC) by evaluating additional blinking features (i.e., amplitude, duration, variability in intervals between blinks) for discriminating patients in UWS from those in MCS or conscious.
3. to explore the usefulness of eye blinking features analysis for assessing residual attentional abilities in full-conscious patients recovering from pDoC.
4. prototyping a machine-learning-enabled pipeline for supporting consciousness diagnosis based on Eye Blink Rate (EBR) and EOG-derived features. The interpretability and reliability aspects will be carefully evaluated to obtain a clinically usable solution for supporting diagnostic procedures and planning tailored cognitive rehabilitation.

Overall, this study is proposing a preliminary investigation, with pilot samples of both healthy and test population, for the analysis of EBR and EOG-derived biomarkers. These results will lay the foundations for the development and validation of multifactorial decisional algorithms for patients' diagnostic and prognostic stratification, based on easily collectable clinical markers.

This project has an observational, cross-sectional design. All patients with sABI consecutively admitted will be screened. We plan to enrol a convenience test sample of 35 patients with sABI including 10 patients with pDoC (5 in UWS, 5 in MCS) and 25 patients who recovered full consciousness after sABI.

A benchmark population of 20 healthy individuals, balanced for age and sex with the patient sample, will be also enrolled, by means of word-of-mouth according to a snowball sampling. This population will undergo anamnestic interview, and neurological and primary cognitive examination for excluding history or presence of neurological disorders. At study entry, patients' demographic (e.g., age, sex), anamnestic (e.g., aetiology, time post-injury, brain lesion location), and clinical data will be collected. Patients' clinical diagnosis and cognitive functioning will be classified according to their best score on the repeated Coma Recovery Scale-Revised (CRS-R) and on Levels of Cognitive Functioning (LCF), respectively.

Within 2 weeks from study entry, each patient will attend an electroencephalogram-electrooculogram (EEG-EOG) recording session. Eye blinking will be examined during a rest condition (duration: 8 min) and during an active auditory oddball task (duration: 8 min), consisting of randomly intermixed tones (non-target: 500 Hz, overall probability: 80%, n=312; target: 1000 Hz, overall probability: 20%, n=78) presented with a 1-s inter-stimulus interval through headphones. To complement clinical assessment, the background activity on resting EEG and the presence of the P300 component on event-related potentials following the auditory oddball paradigm will be evaluated. Patients' consciousness (CRS-R) and cognitive functioning (LCF) levels will be gathered on the day of the EEG-EOG recording session, and considered for statistical analysis.

Control participants will undergo the same EEG-EOG recording as the patients. For both participant groups, blinks will be defined as a sharp positive peak followed by a shallow negative deflection in a time window of <400 ms on the EOG. Moreover, to prevent awareness of blink recording affects blink features, throughout recording sessions participants will be never informed of the blinking recording, but they will be encouraged to stay relaxed with their eyes open. This procedure will be conducted regardless of the participant (patient or control) group.

Biostatistical analyses will be exploited for the investigation of associations between EOG-derived biomarkers and residual attentional abilities in conscious sABI patients under different acquisition paradigms, in response to Objective 3. Objective 4 will be addressed through the development of Machine Learning-based prototypes for the diagnosis of patients with pDoC using information derived from EOG signals, complemented by EEG and clinical data. The interpretability and error analyses will pose the premises for the identification of eventual covert patterns contained in EOG data.