Development of a multimodal seizure detection instrument to improve safety and disease management of epilepsy patients at home

Epilepsy is a chronic brain disease characterized by the unpredictable
recurrence of seizures. We want to develop, validate and clinically evaluate a
new multimodal seizure detection instrument (MSDI, see Fig 1) to alert for
major epileptic seizures (generalized tonic-clonic, tonic, clonic, versive or
hypermotor) during sleep at home or a sheltered home environment. This will
provide a major step in patient safety, care, quality of life and disease
management. Epilepsy is one of the most common neurological conditions. Its
prevalence is 0.7% (120.000 people in the Netherlands), of whom 25% have
regular, intractable seizures, especially children with epilepsy syndromes and
patients with gross brain abnormality and cognitive impairments. About half of
seizures will be at night, posing problems in these vulnerable patient groups
who depend on caregivers not sleeping in the same bed: parents taking care of
their child, or nursing personnel in a sheltered home environment. In case of a
major seizure, the caregiver has to intervene with medication, provide
protection against injury or confusional wandering and give care for comfort.
Development of automated home seizure detection systems is in its infancy, both
scientifically and clinically. Presently, a reliable seizure detection and
alert system at night is lacking. Combined EEG-video is the gold standard for
in-hospital seizure detection. EEG, however, is unsuitable at home, as it is
liable to artefact, requires expert interpretation and is uncomfortable. Home
detection currently consists of unreliable audio-alarms. Thus, many nocturnal
epileptic seizures will go unnoticed, and false alarms occur due to snoring and
other noises. Caregivers do not like to depend on audio-alarms if the security
of the patient is not guaranteed, and will decide to sleep in the same room,
disrupting parental life and potentially hamper the development of autonomy of
the child, or rely on professional care which is costly and also without
instruments. In this project we have developed a new MSDI device using an
optimized combination of non EEG sensors. Preliminary studies led us to select
4 modalities: audio, automated video frame analysis, heart rate and 3D
accelerometry, yielding 10 output variables. We expect that, rather than
relying on one modality, combining modalities will reduce the number of false
positive and false negative alerts.

The overall main objective is to develop an optimal combination of sensors and
techniques to detect major nocturnal seizures with high sensitivity and
specificity in an extramural setting. The study we present here (which is phase
2 out of 3 of the development process) focuses on: a) optimizing existing
techniques and algorithms, especially related to motion sensing; b) optimizing
sensitivity and specificity of the combined technology of heart rate variation,
accelerometry, audiometry and video frame analysis in a large group of
in-hospital patients using the gold standard of combined video-EEG.

The whole project is divided into 3 phases; the current study relates to phase
2. In phase 1 we have perfected the technology and algorithms of the individual
modalities and built the MSDI. In this protocol, we describe phase 2 of the
development process, in which we will test the MSDI in the target population
(n=100) by in-hospital sampling its output simultaneous with the gold standard
of clinical EEG-video. We will then optimize the settings of output variables
to obtain a maximum ROC with sufficient sensitivity (e.g. 90%) and specificity
(e.g. 75% positive predictive value). In phase 3 we will check the performance
of the device at home (n=40) by comparison with the home gold standard of video
observation, and assess the technical feasibility, compliance, effects on
well-being and implementation of the device into novel care models. The phase 3
study will be presented to the METC in a later stage. During phase 2 and 3,
end-users are involved in development, implementation, practical and privacy
issues. We will address ethical implications of the use of the device and the
moral aspects of the development and introduction of this technology which will
influence the behaviour and experiences of users.

The multimodal seizure detection device will run along with clinical video-EEG
registration during the night. The system hardware consists of:
(a) video-audio: this will run independently from the professional video-audio
that is part of the clinical video-EEG recording system. Hardware consists of a
commercial digital camrecorder that is mounted on a standard in the corner of
the sleeping room and is plugged into the power net. Output will be saved on
the hard disk of the MSDI laptop.
(b) ECG and accelerometry: this is a 3-lead system with sensors on the chest
and on both upper arms. The upper arm units are attached by 2 elastic fasteners
with the units directed anteriorly, laterally or posteriorly according to the
wishes of the patient. One of the sensors is also a wireless transmittor that
sends output to:
(c) the MSDI laptop, which is stored in a cupboard in a corner of the room.
During phase 2 of the study all algorithms will be applied offline after data
acquisition. An online alarm is therefore not part of the system at this stage.
The main risk of the system is that the upper arm sensors somehow interfere
with normal sleep, or that the ECG electrodes cause an allergic skin reaction.
In these cases, the sensors will be removed.
The ECG electrodes are standard electrodes, and the accelerometry units are CE
certified. The MSDI will also be tested and approved before clinical use by the
technical department of the UMC Utrecht.
Risk is therefore negligible.
Benefits for the patient will be small if any, and might consist of an
MSDI-detected seizure during the night that was missed by the clinicians while
reviewing the clinical video-EEG registration.
Projected benefits may be large when the MSDI will prove successful at the end
of the study and the patient may be in the target population for the final
MSDI.