Long-Term EEG Monitoring in Patients with Alzheimer’s Disease Using Ear-EEG: A Feasibility Study

Abstract

Background:

Previous studies have reported that epileptiform activity may be detectible in nearly half of patients with Alzheimer’s disease (AD) on long-term electroencephalographic (EEG) recordings. However, such recordings can be uncomfortable, expensive, and difficult. Ear-EEG has shown promising results for long-term EEG monitoring, but it has not been used in patients with AD.

Objective:

To investigate if ear-EEG is a feasible method for long-term EEG monitoring in patients with AD.

Methods:

In this longitudinal, single-group feasibility study, ten patients with mild to moderate AD were recruited. A total of three ear-EEG recordings of up to 48 hours three months apart for six months were planned.

Results:

All patients managed to wear the ear-EEG for at least 24 hours and at least one full night. A total of 19 ear-EEG recordings were performed (self-reported recording, mean: 37.15 hours (SD: 8.96 hours)). After automatic pre-processing, a mean of 27.37 hours (SD: 7.19 hours) of data with acceptable quality in at least one electrode in each ear was found. Seven out of ten participants experienced mild adverse events. Six of the patients did not complete the study with three patients not wanting to wear the ear-EEG anymore due to adverse events.

Conclusion:

It is feasible and safe to use ear-EEG for long-term EEG monitoring in patients with AD. Minor adjustments to the equipment may improve the comfort for the participants.

Keywords

Alzheimer’s disease ear-EEG electroencephalography long-term EEG wearable

INTRODUCTION

Patients with Alzheimer’s disease (AD) are at an increased risk of developing epileptic seizures [1 –6], which may increase with the severity of AD [7, 8] and may be detrimental for their functional abilities. An indicator of seizures is epileptiform activity in the electroencephalography (EEG). Epileptiform activity may not be accompanied by clinical manifestations. Furthermore, the most prevalent type of seizure in patients with AD is focal impaired awareness seizures [5, 9], which may mimic symptoms of AD and may go unnoticed by patient and caregiver. Therefore, seizures may be much more common in patients with AD than previously reported. This has been underscored in a study showing that epileptiform activity without clinical seizure phenomena is present in 42.4% of patients with AD [10]. However, although in-patient EEG monitoring over longer periods may be more effective, such in-patient monitoring can be difficult for the patients with dementia, is expensive and not possible for extended periods. This implies the importance of both recognizing epileptiform activity in patients with AD and the need for easy and inexpensive continuous monitoring.

In recent years, methods for out-patient assessment of a variety of biometric data in AD have been investigated [11]. However, no devices have so far been applied for continuous EEG recordings in AD. In patients with epilepsy, several methods for long-term EEG [12] as well as other wearable devices [13] have been studied including ear-EEG [14] and subcutaneous electrodes [15]. Ear-EEG is a method where EEG signals are acquired from electrodes placed within a customized earpiece inserted into each ear canal. The ear-EEG has several advantages including that the patients can stay in a familiar environment with minimal interference of everyday life while undergoing long-term EEG monitoring and are able to easily remove and insert the device back into the ear canal [16 –18]. However, the EEG coverage of the brain using ear-EEG involves mainly the temporal lobes. This may be less of a disadvantage when examining patients with AD since the majority of epileptiform activity arise from the temporal lobes [10, 19]. Also, previous studies have shown a good correspondence between ear-EEG and scalp EEG [20, 21]. In addition, a recent study found that ear-EEG can be used to detect EEG patterns associated with focal temporal lobe seizure [14]. Other studies have shown that ear-EEG performs well in sleep scoring [21 –24]. Therefore, ear-EEG may be a suitable tool to investigate whether patients with AD develop subclinical epileptiform activity.

Since out-patient long-term EEG monitoring of patients with AD may pose other challenges than in cognitively perserved persons, it is relevant to examine the feasibility of such methods, and to determine the best tradeoffs between both data quality and compliance. Specifically, patients with AD suffer from memory problems and other cognitive deficits like executive functions [25], which may give rise to problems planning the recordings, setting up the equipment, or remembering the instructions. This may lead to lower compliance and quality of the recordings. Furthermore, patients with AD may have a different tolerance to pain [26, 27], or may not respond adequately to other safety issues.

In the present feasibility study, we therefore aimed to investigate if ear-EEG is a feasible and safe method for long-term EEG monitoring of patients with AD.

MATERIALS AND METHODS

Patients

Patients with mild to moderate AD were recruited from either the Memory Clinic at Rigshospitalet, Copenhagen or the Memory Clinic at Zealand University Hospital, Roskilde. All patients met the NIA-AA criteria for probable AD with amnestic presentation [28] and the diagnosis was determined based on a consensus conference. Inclusion criteria were 1) age between 50-90 years, 2) a Mini-Mental Status Examination (MMSE) [29] score between 16-28, 3) native Danish speaker, 4) at least 7 years of education, 5) hearing and vision sufficient for neuropsychological examination, 6) the general health conditions of the patient allowed participation in the study (as judged by the principal investigator), 7) did not suffer from any psychiatric (except mild depression) or neurological conditions that affects the brain except AD, 8) any treatment with either anti-dementia medication or selective serotonin reuptake inhibitors (if relevant) were given in a stable regimen within the last three months up to inclusion, 9) no alcohol or drug abuse within the last two years, 10) an MRI or CT scan that supported the diagnosis of AD, 11) no contraindications for MRI, and 12) lived with a caregiver who were able to assist the patient with the home recordings.

The following exclusion criteria were applied: 1) epilepsy prior to the diagnosis of AD, 2) focal pathology (except AD related atrophy) in the hippocampus, i.e., hippocampal sclerosis, 3) living with relative with serious illness or impaired activities of daily living, 4) living in a nursing home, 5) treated with anti-epileptic medication, tricyclic antidepressants or antipsychotics, 6) daily or almost daily administration of medication with known anticholinergic or adrenergic effect, which may affect cognitive abilities or EEG, 7) large cerebral infarctions or more than four lacunar infarctions on MRI, 8) suffering from facial tics/facial hyperkinetic disorders, or 9) daily use of hearing aids.

The study was approved by the Capital Region Ethics Committee (H-17035751) and by the Danish Medicines Agency (2017112288). All participants gave their written informed consent before participating in the study. The study is registered at clinicaltrials.gov (NCT04436341).

Study design

In this longitudinal single-group feasibility study, a total of 8 visits including an MRI scan was planned over a 6-month period (see Fig. 1). The three two-day ear-EEG recordings were performed three months apart, i.e., at the beginning of the study, at three months and at 6 months follow-up. At visit 1 (baseline), informed consent was obtained followed by assessment of medical history, a physical and neurological examination, the MMSE (to assess global cognition) and an imprint of the ears using Otoform a soft ear impression silicone. Subsequently, the patient underwent the following: visit 2) MRI scan, visit 3) included standard EEG recording together with ear-EEG, Functional Assessment Questionnaire IADL (FAQ IADL) [30] (to assess everyday function) and neuropsychiatric inventory (NPI) [31] (to assess behavioral and psychological symptoms). Further, the patient and the caregiver were briefly instructed about seizures and introduced to a seizure dairy to take home, and information and training on how to use the ear-EEG at home. This included a log-sheet for the patient or caregiver to take notes on when the patient wore the ear-EEG. After approximately 48 hours (visit 4) the patient and/or the caregiver returned with the equipment and a questionnaire was used, see below. For most of the visit, the outer ear was examined for any minor lesions. Visit 5 and 6 and visit 7 and 8 were the same as visit 3 and 4 (except MMSE).

Fig. 1

Study design. Firstly, the patients were included in the study after written informed consent. The figure shows the following visits with visit 5 and 6 being three months after the visit 4 and visit 7 and 8 being three months after visit 5.

Due to the COVID pandemic, the time between visits was prolonged including the time between visit 4 and 5, and visit 6 and 7.

A short questionnaire was administered by the investigator when the patient and/or the caregiver returned the ear-EEG (visit 4, 6, and 8). The questionnaire included the following questions: a) how did it feel to wear the ear-EEG, b) did you have any problems sleeping with the equipment, c) did the ear-EEG ever fall out of the ear, and d) do you have any suggestions for improvement.

Ear-EEG

The ear-EEG consists of six dry-electrodes mounted in earpieces which are custom-made using ear imprints and 3D-printing and therefore fitted individually to each participant. The ear-piece was placed inside the ear canal and in the concha (Fig. 2A) [20]. Earpieces were worn in both ears. The labeling convention of the ear-EEG electrodes has previously been described [32], and the labeling in the current study can be seen in Fig. 2A and B. The ground electrode was placed approximately 2 cm under the midline of the clavicle on either the right or left side. The ear-EEG recordings were performed using the TMSi Mobita EEG amplifier (TMSi systems), which was borne in a small bag around the neck and connected to the electrodes through individual wires. For the recordings, a sampling rate of 1000 Hz was chosen.

Fig. 2

Figure showing the two side of the ear-EEG (A and B). The electrodes are marked with the respective names.

Data quality

The EEG data were loaded into MATLAB (v2020b) using a custom script for loading poly5 files. Afterwards, the EEGs were bandpass filtered from 0.5-70 Hz and subsequently notch filtered from 49-51 Hz and 99-101 Hz using the pop_eegfiltnew function from EEGLAB [33]. Next, the following artefact rejection pipeline, which has previously been used in other studies [21], was used. Firstly, if a channel were outside the dynamical range of the amplifier, which is a typical indication of an electrode with poor contact to the body, the reading was replaced with a ‘NaN’-value, and hence automatically discarded upon loading. The discarded data was considered in the subsequent analysis of data quality. Afterwards, large amplitude samples isolated to a single channel were labelled as artifacts. Lastly, movement or muscle activation, which may also result in large amplitude deviations across multiple channels were labelled as artifacts. The Matlab scripts for preprocessing can be found in the Supplementary Material.

To assess the amount of data after the pre-processing, the amount of time when data from at least one electrode in each ear were present ((ELA or ALB or ELC or ELE or ELI or ELT) and (ERA or ARB or ERC or ERE or ERI or ERT)) was calculated. In addition, the amount of available data for each electrode separately were also examined. Lastly, the amount of time where data from at least one electrode on each ear were present during the night (10 pm-8am) and day (8am-10pm) was examined. Here, we only used the patients’ reported start time of recording due to the discrepancy between amount of data recorded on the EEG recorder and self-reported recording times. The code for the calculations can be found in Supplementary Material.

Assessment of feasibility

Feasibility was assessed based on the overall ability of conducting the study including safety, acceptability, improvements in physical capacity, and the participants’ ability to complete the physical tests. Safety was assessed based on the occurrence of adverse events (AE) and serious adverse events (SAE, i.e., AEs leading to, e.g., hospitalization and death). Acceptability was assessed on the basis of attendance rate, level of training intensity achieved (% of heart rate reserve and perceived exertion), number of drop-outs, and self-report. Furthermore, the ability of the clinical outcome measures to assess change over time (i.e., absence of floor or ceiling effects) wasevaluated.

Feasibility was assessed based on the overall ability of the patients to wear the ear-EEG including responses to questionnaire, data quality and the amount of data recorded, and safety. Safety was assessed based on the occurrence of adverse events (AE) including serious adverse events (SAE) (e.g., events leading to hospitalization). The duration of time that the patient wore the ear-EEG was based on the log-sheet that the patient and/or caregiver filled out. In addition, we evaluated the amount of data of acceptable quality after pre-processing. This included the useable data at night (10pm – 8am) as compared to daytime (8am – 10pm). Two of the 19 recordings were not included in the night/day calculations due to either only having recorded during the daytime (n = 1) or did not fill out the log-sheet as instructed(n = 1).

The predetermined criteria to test whether ear-EEG is a feasible tool for long-term EEG monitoring was that at least 80% of the participants wore the ear-EEG for at least 24 h and at least one night over each 2-day period completed. The predetermined threshold of 24 h was used as previous studies using conventional long-term EEG to examine epileptiform activity in patients with AD had recorded for up to 24 h.

A time-frequency representation, a so-called spectrogram, was used to visualize the data for a single subject over the whole recording. Here, the pre-processed signal was averaged to a single channel for the left side and all NaN values were substituted with zeros. The spectrogram was calculated using a 5000-point discrete-time Fourier transform with a Hamming window, 10% overlap (equal to 500) and 5000 frequency points. The mean frequency was calculated between 1 and 70 Hz. The displayed times are from the noted start times of recordings performed by the patient and caregiver. Due to average referencing of the signal, the amplitude of the signal on the left side will be equal to the right side, which means that we did not perform a spectrogram for both left and right side. The full code for creating and visualizing the spectrogram and mean frequency can be found in the supplementary material.

To visualize the data, we showed 10 s of data with artefacts that would be removed using the pre-processing script as well as 10 s of data without artefacts recorded at night. In addition, we showed the data in the frequency domain using the EEGLAB spectopo function between all available channels and visualized from 0-60 Hz.

Adverse events

Adverse events were recorded throughout the whole study period. In addition, the investigators were automatically notified of any hospitalization during the time the patients were enrolled in the study. An SAE was defined according to the international criteria (Good Clinical Practice).

RESULTS

A total of 10 patients with AD were eligible for inclusion and gave informed consent. Baseline characteristic including results from cognitive tests can be seen in Table 1.

Table 1

Baseline demographics and clinical scales

Gender, female/male	4/6
Age, mean (SD)	74.7 (6.0)
Education in years, mean (SD)	13.5 (2.99)
Symptom onset in years, mean (SD)	2.92 (1.6)
MMSE, mean (SD)	23.7 (4.0)
ADL, mean (SD)	14.2 (5.8)
NPI, mean (SD)	3.8 (2.7)

MMSE, Mini-Mental State Examination; CCMT, Category Cued Memory Test; ADL, Activities of Daily Living; NPI, Neuropsychiatric Inventory; SD, standard deviation.

Feasibility

All patients wore the ear-EEG for at least 24 h and at least one full night. A total of 19 ear-EEG recordings were performed out of a possible total of 30. Mean self-reported length of the first ear-EEG recording was 36.97 h (SD: 3.93 h, Range: 28.75 h – 42.58 h). The mean self-reported length for all 19 recordings were 37.15 h (SD: 8.96 h, range: 9.25 h – 58.5 h). Recording duration, MMSE, and days since last ear-EEG recording for the patients who participated for more than one ear-EEG session can be seen in Table 2. When interviewed, half the patients described difficulties putting on and wearing the equipment. All participants needed their caregiver’s help to either adjust the earpiece into the ear or start the recording. Four of the participants described that the EEG recorder was too heavy or uncomfortable for wearing around the neck during the day.

Table 2

Participants with more than one ear-EEG session

	First ear-EEG	Second ear-EEG	Third ear-EEG
	recording	recording	recording
	(n = 5)	(n = 5)	(n = 4)
Ear-EEG recording duration, mean	38.32 h	36.88 h	37.93 h
MMSE, mean (SD)	24.2 (4.0)	23.2 (4.8)	22.8 (3.7)
Days since last ear-EEG recording		248.4 (89.8)	179 (124.5)

MMSE, Mini-Mental State Examination.

A total of seven participants developed mild tenderness in or around the ear. This included three incidences of mild redness of the skin on the external ear and two incidences where participants developed small abrasions of the skin. One of the participants with abrasions developed pain and swelling of the left chin indicative of a mild skin infection. Three patients did not show any discomfort or side-effects to wearing the ear-EEG. Only one of the adverse events led to medical contact and none of them lasted more than a week. None were classified as SAEs.

Six participants dropped out after the first (n = 5) or second time (n = 1) that they wore the ear-EEG. Four participants completed all visits. The reasons for dropping out were: not wanting to wear the ear-EEG anymore due to adverse events (n = 3), not wanting to participate during the COVID-19 pandemic (n = 1), personal reasons (n = 1), and moved to a nursing home (n = 1).

Data quality

Plots of 10-s segments of artefactual and acceptable-quality data in both time and frequency domain can be found in Fig. 3. The mean number of recorded hours as recorded on the EEG recorder were 38.42 (SD: 9.44 h). When investigating the amount of time where at least one electrode had sufficient artifact free data on each side, an average of 11.05 h (SD: 5.61 h) were removed. This resulted in a mean number of 27.37 h (SD: 7.19 h) of at least one-electrode data corresponding to total 71.24% of the data being of acceptable quality.

Fig. 3

Figure showing a 10-s ear-EEG segment with A) artefacts or B) without artefacts in both time and frequency domain. Pulse artefacts are visible in B). In the frequency domain, it can be observed that much more high frequency activity (>30 Hz) is present in A as compared to B.

Data quality for each of the channels are reported in Table 3. Overall, the channels with the least discarded data were ELC, ELE, ERC, ERE, and ERI.

Table 3

Data for each electrode after pre-processing

	Percentage of data
ELA	44.33
ELB	41.84
ELC	64.46
ELT	49.17
ELE	65.96
ELI	55.15
ERA	53.54
ERB	49.01
ERC	64.52
ERT	50.41
ERE	68.89
ERI	61.63

The Ear-EEG electrodes are labelled by 3-letter acronyms: E stands for ear followed by letter L for left or R for right and the third letter corresponds to a specific electrode position, see Fig. 2.

Regarding the distribution of data quality across the day, we found that during the nighttime 84.64% of the data remained after preprocessing as compared to 60.77% of daytime data.

Spectrogram and mean frequency for the ear-EEG signal can be seen in Fig. 4. As can be seen on the figure, there is a decrease in high-frequency activity and a lower mean frequency at night, which is probably due to the patient sleeping.

Fig. 4

Spectrogram for the mean ear-EEG signal and mean frequency from the left ear-EEG. As can be seen on the figure, there is a decrease in high-frequency activity and a lower mean frequency as night, which could be due to the patient sleeping. The displayed times are from the noted start times of recordings performed by the patient and caregiver.

DISCUSSION

This is the first study to provide data on the feasibility of out-patient long-term EEG monitoring in patients with mild to moderate AD. All patients were able to wear the device for at least one recording (>24 h) and thus, according to our predefined threshold, long-term EEG recording is feasible in AD patients. Data quality was acceptable and nighttime recording yielded a markedly higher proportion of data with acceptable quality. Although 7 out of 10 participants developed mild discomfort while wearing the device, there were no SAEs, and therefore it may be considered safe to use ear-EEG for long-term out-patient monitoring in patients with AD.

To our knowledge, long-term out-patient EEG monitoring with a wearable device has never been performed in patients with AD but has been used in patients with epilepsy. Here, studies have shown that multiple factors affect long-term engagement including the patient being undisturbed by the device during daily activities and sleep [34, 35] as well as the appearance of the device [35, 36]. In the current prototype of the ear-EEG setup, the TMSi amplifier was heavy to carry around the neck in a bag and interfered with daily activities. A solution was to change to a belt bag strapped around the waist that could store the TMSi amplifier, which makes it more comfortable and easier for the participants to perform daily activities. In addition, the device was difficult to conceal, which may have prevented some of the patients from wearing it outside of the house or limited the activities patients engaged in during the recording. Furthermore, half of the patients had trouble wearing it at night, which led to lower quality of sleep. One contributing factor could be that if patients slept on the side, the earpieces may have caused pressure on the ear. Development of devices for long-term EEG recordings in patients with AD should have an increased focus on comfort during sleep in future prototypes of the ear-EEG device or a solution to support the head when sleeping, i.e., a C-shaped pillow to decrease the pressure on the ear. In a previous study investigating the ability of ear-EEG to detect the seizures in patients with suspected temporal lobe epilepsy (n = 15), the majority of patients experienced some degree of soreness but this only led to discontinuation of the use of the ear-EEG for two of the participants (13.3%) [14], which is less than in the current study with discontinuation of three participants (30%). However, the patients with epilepsy were for the most part younger and were hospitalized at the time of EEG recordings, which made it easier to seek help. Furthermore, less electrodes were used (three on each side) as compared to the present study. The higher number of electrodes and thus closer proximity of the electrodes to each other may have been the cause of pain. Specifically, ELE and ELI, and ERE and ERI electrodes, may have been placed too close within the ear canal, which led to increased pressure. The positioning of the two electrodes in the ear canal should be adjusted. An overview of the suggestions for improving the ear-EEG recordings can be found in Table 4.

Table 4

Table showing the challenges during the ear-EEG recordings as well as suggested improvement for long-term ear-EEG recordings in patients with AD

#	Patient reported discomfort	Mitigation strategies
1	Mild tenderness of the ear	•Breaks throughout the recording, i.e., during meals
2	Minor abrasions in the ears	•Less sharp electrode edges
3	Disturbed sleep due to ear-EEG	•Improved comfort when laying on the side for the ear-EEG
		•Support for the head, i.e., C-shaped pillow around the ear
4	Amplifier too heavy and	•Belt bag strapped around the waist
	interferes with daily activities	•Smaller amplifier
5	Unwanted attention to the device in public	•Developing a smaller amplifier and less wires from the ear-EEG

A study has previously looked at data quality using the same ear-EEG setup as presented here with dry electrodes and investigated out-patient sleep monitoring. The study included young healthy participants (mean age: 25.9 years) and showed very good data quality with 91% accepted data with overall good data quality across the electrodes [21]. We found that 71.24% of the data was acceptable data quality with recordings at night (10 pm–8 am) yielding the most acceptable data quality (84.18% compared to 61.15% during the day). Compared to the study in a younger population, nighttime recordings in terms of the amount of data of acceptable data quality was similar. As compared to the study investigating out-patient sleep monitoring [21], we found large differences in the data quality between electrodes, which may be due to participants/caregiver had to place them in the patients ears themselves. Specifically, it may be difficult to adjust the A and B electrodes (see Fig. 2) within the concha cymba. To increase the likelihood that the caregivers can use the equipment correctly, an instruction sheet could be helpful and has been implemented in future out-patient ear-EEG recordings.

When plotting a spectrogram from an ear-EEG recording, we found that more activity was present in 7-9 Hz during the day as compared to the night (11pm-10am) as well as a lower mean-frequency at night. We hypothesize that this is due to the patient sleeping since a previous study comparing changes in the frequency domain has found slowing of activity during sleep as compare to wakefulness [23]. Another explanation for lower mean frequency during sleep could be less artefacts due to muscle activity. To further explore the changes at night, it would be of interest to apply methods developed for automatic sleep staging using ear-EEG [21], but the methods have not been validated in older participants or patients with AD.

Most participants in the study (n = 7) reported mild tenderness in or around the ear after wearing the ear-EEG. In comparison, another long-term ear-EEG study [21] found occasional soreness and only one incident of skin irritation (n = 20). However, an important difference is that in this study the EEG was measured for a long continuous period, whereas in [21] it was only measured at night with breaks during the day. In a study investigating ear-EEG in patients with temporal lobe epilepsy the majority of participants experienced soreness, which increased as more time went on [14]. However, this study used another prototype of the ear-EEG and applied gel instead of dry electrodes, which makes the comparison between the current study more difficult. Therefore, future studies using ear-EEG for long-term out-patient recordings should consider breaks throughout the day especially in patients with dementia. These breaks could conveniently be placed during meals since chewing leads to considerable artefacts in the ear-EEG. Furthermore, two participants developed small abrasions, which we hypothesize could be due to either the edges of the electrodes, the caregiver inserting the ear-EEG or adjustment of the ear-EEG throughout the recording by the participant. In one of these cases, we found that the participant developed pain and swelling of the left cheek, which may have been caused by an infection. The patient was completely recovered after five days. This type of adverse event has not previously been reported using ear-EEG but should be considered when informing the participants of the possible risk. Overall, the adverse events were considered mild, and no SAEs were found, and it therefore seems safe to apply ear-EEG in this patient population. So far, no other wearable modalities for long-term EEG monitoring in patients with AD have been investigated. Methods like sheet-type EEG [37] or around the ear (cEEGrid) [38] may lead to less side effects in patients with AD but studies using other wearable EEG devices for long-term out-patient recordings in AD are needed.

The study has some limitations. First, the number of subjects is low due to the nature of a feasibility study but a more representative sample of patients with AD may have given rise to different considerations regarding the use of ear-EEG. Furthermore, the quality of the ear imprints may have varied due to either errors in the process or the patient moving while the imprint was made. This may have led to inaccuracies of the imprint, which in turn may have caused tenderness in the ear. We did not implement different recording lengths (i.e., one long recording compared to multiple shorter recordings) for the participants, which may have given a better understanding on how to use ear-EEG for long-term monitoring in patients with AD. Furthermore, the TMSi amplifiers does not record time during the EEG recordings, which made the subsequent analysis highly dependent on the log-sheets and was therefore not possible for one of the recordings. Lastly, a more comprehensive questionnaire to understand the comfortability of ear-EEG should be administered in future studies.

Conclusion

Ear-EEG is a feasible technique for long-term out-patient monitoring of patients with mild to moderate AD with only mild adverse events for the participants. The overall data quality was good with an excellent quality during the night, which is optimal since a study found that most of the reported epileptiform activity in patients with AD were present at night [19]. In addition, it was possible by visualizing the ear-EEG using a spectrogram to see changes between day and night. The electrodes located within the ear canal (ELE/ELI, and ERE/ERI) showed the best data quality. This may be due to these electrodes being easier to correctly place within the ear or that the electrodes within the ear have lower impedance due to a moister environment [16]. Minor adjustments may increase the comfort for the participants. Future studies using more participants will show whether ear-EEG is able to detect epileptiform activity in patients with AD.

Footnotes

ACKNOWLEDGMENTS

The study was funded by the Alzheimer Research Foundation (grant number: 181003), Toyota Foundation (KJ/BG-9171 F), Axel Juul Muusfeldts Foundation (2016-527), Ellen Mørchs foundation (J.nr. 32491419), Rigshospitalet Research Foundation and T&W Engineering. None of the funding parties had a role in the collection, analysis, and interpretation of data or in the writing of the manuscript or whether to publish the results of the study. Two authors (MLR, MCH) who are employees of T&W Engineering contributed to the interpretation and writing of the report.

We would like to thank Susanne Kristiansen and Oda Jakobsen for their help recruiting participants and helping during study visits.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Scarmeas

, Honig

, Choi

, Cantero

, Brandt

, Blacker

, Albert

, Amatniek

, Marder

, Bell

, Hauser

, Stern

(2009) Seizures in Alzheimer disease. Arch Neurol 66, 992–997.

Amatniek

, Hauser

, DelCastillo-Castaneda

, Jacobs

, Marder

, Bell

, Albert

, Brandt

, Stern

(2006) Incidence and predictors of seizures in patients with Alzheimer’s disease. Epilepsia 47, 867–872.

Lozsadi

, Larner

(2006) Prevalence and causes of seizures at the time of diagnosis of probable Alzheimer’s disease. Dement Geriatr Cogn Disord 22, 121–124.

Bernardi

, Scaldaferri

, Vanacore

, Trebbastoni

, Francia

, D’Amico

, Prencipe

(2010) Seizures in Alzheimer’s disease: A retrospective study of a cohort of outpatients. Epileptic Disord 12, 16–21.

Rao

, Dove

, Cascino

, Petersen

(2009) Recurrent seizures in patients with dementia: Frequency, seizure types, and treatment outcome. Epilepsy Behav 14, 118–120.

Risse

, Lampe

, Bird

, Nochlin

, Sumi

, Keenan

, Cubberley

, Peskind

, Raskind

(1990) Myoclonus, seizures, and paratonia in Alzheimer disease. Alzheimer Dis Assoc Disord 4, 217–225.

Romanelli

, Morris

, Ashkin

, Coben

(1990) Advanced Alzheimer’s disease is a risk factor for late-onset seizures. Arch Neurol 47, 847–850.

Förstl

(1992) Neurologic signs in Alzheimer’s disease. Arch Neurol 49, 1038.

Pandis

, Scarmeas

(2012) Seizures in Alzheimer disease: Clinical and epidemiological data. Epilepsy Curr 12, 184–187.

10.

Vossel

, Ranasinghe

, Beagle

, Mizuiri

, Honma

, Dowling

, Darwish

, Van Berlo

, Barnes

, Mantle

, Karydas

, Coppola

, Roberson

, Miller

, Garcia

, Kirsch

, Mucke

, Nagarajan

(2016) Incidence and impact of subclinical epileptiform activity in Alzheimer’s disease. Ann Neurol 80, 858–870.

11.

Kourtis

, Regele

, Wright

, Jones

(2019) Digital biomarkers for Alzheimer’s disease: The mobile/wearable devices opportunity. NPJ Digit Med 2, 9.

12.

Casson

, Yates

, Smith

, Duncan

, Rodriguez-Villegas

(2010) Wearable electroencephalography. IEEE Eng Med Biol Mag 29, 44–56.

13.

Brinkmann

, Karoly

, Nurse

, Dumanis

, Nasseri

, Viana

, Schulze-Bonhage

, Freestone

, Worrell

, Richardson

, Cook

(2021) Seizure diaries and forecasting with wearables: Epilepsy monitoring outside the clinic. Front Neurol 12, 690404.

14.

Zibrandtsen

, Kidmose

, Christensen

, Kjaer

(2017) Ear-EEGdetects ictal and interictal abnormalities in focal and generalizedepilepsy - A comparison with scalp EEG monitoring. ClinNeurophysiol 128, 2454–2461.

15.

Weisdorf

, Duun-Henriksen

, Kjeldsen

, Poulsen

, Gangstad

, Kjær

(2019) Ultra-long-term subcutaneous home monitoring of epilepsy-490 days of EEG from nine patients. Epilepsia 60, 2204–2214.

16.

Kappel

, Rank

, Toft

, Andersen

, Kidmose

(2019) Dry-contact electrode ear-EEG. IEEE Trans Biomed Eng 66, 150–158.

17.

Looney

, Kidmose

, Park

, Ungstrup

, Rank

, Rosenkranz

, Mandic

(2012) The in-the-ear recording concept: User-centered and wearable brain monitoring. IEEE Pulse 3, 32–42.

18.

Looney

, Park

, Kidmose

, Rank

, Ungstrup

, Rosenkranz

, Mandic

(2011) An in-the-ear platform for recording electroencephalogram. In IEEE, 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

19.

Lam

, Sarkis

, Pellerin

, Jing

, Dworetzky

, Hoch

, Jacobs

, Lee

, Weisholtz

, Zepeda

, Westover

, Cole

, Cash

(2020) Association of epileptiform abnormalities and seizures in Alzheimer disease, Neurology 95, e2259–e2270.

20.

Mikkelsen

, Kappel

, Mandic

, Kidmose

(2015) EEG recorded from the ear: Characterizing the ear-EEG method. Front Neurosci 9, 438.

21.

Mikkelsen

, Tabar

, Kappel

, Christensen

, Toft

, Hemmsen

, Rank

, Otto

, Kidmose

(2019) Accurate whole-night sleep monitoring with dry-contact ear-EEG. Sci Rep 9, 16824.

22.

Nakamura

, Alqurashi

, Morrell

, Mandic

(2020) Hearables: Automatic overnight sleep monitoring with standardized in-ear EEG sensor. IEEE Trans Biomed Eng 67, 203–212.

23.

Zibrandtsen

, Kidmose

, Otto

, Ibsen

, Kjaer

(2016) Case comparison of sleep features from ear-EEG and scalp-EEG. Sleep Sci 9, 69–72.

24.

Jorgensen

, Zibrandtsen

, Kjaer

(2019) Ear-EEG-based sleepscoring in epilepsy: A comparison with scalp-EEG, J Sleep Res 29, e12921.

25.

Guarino

, Favieri

, Boncompagni

, Agostini

, Cantone

, Casagrande

(2019) Executive functions in Alzheimer disease: A systematic review. Front Aging Neurosci 10, 437.

26.

Benedetti

, Vighetti

, Ricco

, Lagna

, Bergamasco

, Pinessi

, Rainero

(1999) Pain threshold and tolerance in Alzheimer’s disease. Pain 80, 377–382.

27.

Defrin

, Amanzio

, de Tommaso

, Dimova

, Filipovic

, Finn

, Gimenez-Llort

, Invitto

, Jensen-Dahm

, Lautenbacher

, Oosterman

, Petrini

, Pick

, Pickering

, Vase

, Kunz

(2015) Experimental pain processing in individuals with cognitive impairment: Current state of the science. Pain 156, 1396–1408.

28.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

, Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

29.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

30.

Pfeffer

, Kurosaki

, Harrah

, Chance

, Filos

(1982) Measurement of functional activities in older adults in the community. J Gerontol 37, 323–329.

31.

Cummings

, Mega

, Gray

, Rosenberg-Thompson

, Carusi

, Gornbein

(1994) The Neuropsychiatric Inventory: Comprehensive assessment of psychopathology in dementia. Neurology 44, 2308.

32.

Kidmose

, Looney

, Mandic

(2012) Auditory evoked responses from Ear-EEG recordings. In IEEE, 2012 Annual International Conference of the IEEE engineering in Medicine and Biology Society.

33.

Delorme

, Makeig

(2004) EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods 134, 9–21.

34.

Van de Vel

, Smets

, Wouters

, Ceulemans

(2016) Automated non-EEG based seizure detection: Do users have a say? Epilepsy Behav 62, 121–128.

35.

Hoppe

, Feldmann

, Blachut

, Surges

, Elger

, Helmstaedter

(2015) Novel techniques for automated seizure registration: Patients’ wants and needs. Epilepsy Behav 52, 1–7.

36.

Simblett

, Biondi

, Bruno

, Ballard

, Stoneman

, Lees

, Richardson

, Wykes

(2020) Patients’ experience of wearing multimodal sensor devices intended to detect epileptic seizures: A qualitative analysis. Epilepsy Behav 102, 106717.

37.

, Egawa

, Yoshimoto

, Mizutani

, Kobayashi

, Tachibana

, Takahashi

(2019) Potential clinical applications and future prospect of wireless and mobile electroencephalography on the assessment of cognitive impairment. Bioelectricity 1, 105–112.

38.

Bleichner

, Debener

(2017) Concealed, unobtrusive ear-centered EEG acquisition: cEEGrids for transparent EEG. Front Hum Neurosci 11, 163.