WO2015039744A1

WO2015039744A1 - Method and device for measuring cerebral perfusion

Info

Publication number: WO2015039744A1
Application number: PCT/EP2014/002511
Authority: WO
Inventors: Stephanie Gollwitzer; Teja Grömer; Hajo Hamer; Rüdiger Hopfengärtner
Original assignee: Universitätsklinikum Erlangen
Priority date: 2013-09-18
Filing date: 2014-09-17
Publication date: 2015-03-26
Also published as: DE102013015427A1

Abstract

A method for measuring the cerebral perfusion of a patient (8) is disclosed, comprising the following steps: continuously recording, over a long period of time, and storing an EEG signal (E) of the patient (8); cleaning the EEG signal (E) of artefacts; carrying out power spectral analysis of the artefact-cleaned EEG signal in certain frequency bands; cleaning the traces obtained by power spectral analysis around a common trend of all traces; detecting deviations (D) of the trend-cleaned power values of at least one trace from a baseline; and displaying significant deviations (D) and/or triggering an alarm (A) when a significant deviation (D) is determined. An associated device (1) comprises an EEG machine (2) and a control unit (3) which is connected thereto for data communication and which is designed to carry out the method automatically.

Description

description

Method and device for measuring cerebral perfusion

The invention relates to a method for measuring cerebral perfusion. The method is used in particular for the detection of

vasospasm-related imminent cortical ischemia in patients with subarachnoid hemorrhage. The invention further relates to an apparatus for carrying out the method.

Subarachnoid haemorrhages are among the most common acute neurological conditions. In addition to rebleeding and cerebrospinal fluid disturbances, vasospasm and resulting cortical ischaemia represent the greatest risk to the affected patients.

Gold standard for the detection of vasospasm is currently the transcranial Doppler and duplex sonography. However, this only allows a punctual examination of the patient, for example once a day. Events between the examinations are not detected late or can not be treated effectively.

The invention has for its object to provide a related improved method for measuring cerebral perfusion. In particular, the method is intended to enable the early detection of an imminent cortical ischemia as a result of vasospasm. A further object of the invention is to provide a device which is particularly suitable for carrying out the method.

According to the invention, this object is achieved by the features of claim 1. Advantageous embodiments and further developments of the invention are set forth in the subclaims and the description below. To effectively prevent vasospasm-related infarcts and thus to improve the functional outcome of patients, continuous monitoring of cortical perfusion seems desirable. This can

can be ensured by a continuous EEG recording. Compromising brain function due to disturbances of cortical perfusion is measured directly here. It is recognized that reversible neuronal changes that occur in an abnormal decrease in cerebral blood flow lead to measurable EEG changes. Delayed cerebral ischaemia (DCI for short) can thus be detected before irreversible cell death and at best by appropriate countermeasures - e.g. by the administration of nimodipine, by intra-arterial spasmolysis or by triple-H therapy.

Cortical ischemia is recognized in the EEG as a focal reduction of the power in the frequency spectrum of the EEG signal. This effect is automatically detected by the inventive method, this for an automated artifact elimination and power spectral analysis, the trend correction of raw data and the automated detection of significant deviations from includes a baseline.

According to the method, the quantitative EEG (qEEG) directly the

Compromise of brain function, and thus indirectly measured cerebral perfusion. The method comprises the following steps:

1) Continuous long-term derivation of the patient's EEG signal and storage of this EEG signal, especially on a hard disk. For the EEG recording, an electrode assembly according to the international 10-20 system is suitably used. In order to simplify the process or, if not otherwise possible (if, for example, not all electrodes of the 10-20 system can be attached due to existing bandages), in an alternative embodiment of the invention, advantageously only a subset of the electrodes (eg F7 / F8, T3 / T4, C3 / C4, P3 / P4, 02/01, Fz, Cz and Pz) were used for EEG recording. The sampling rate of the multi-channel EEG derivative preferably carries at least 200Hz. The raw EEG signal is preferably recorded in the frequency range of at least 0.1 Hz-70 Hz (more preferably, for example, between 0.05 Hz and 100 Hz). The recording of the EEG signal is preferably carried out for at least 12 hours, expediently over several days, in particular over at least 4 days. Preferably, the EEG signal is recorded over a period of between 5 and 12 days, a maximum of a few weeks. For the detection of vasospasm-related cortical ischemia following subarachnoid hemorrhage (SAB), recording is advantageously started as soon as possible, in particular within the first three days after detection of SAB.

2) Automated artifact cleansing (i.e., artifact detection and elimination) which filters out artifacts that typically overlay the EEG signal and that would therefore significantly distort the result of a quantitative analysis. Artifacts are generally all signal components of the measured EEG signal, which are not related to the electrical brain activity to be measured but to other triggering events, e.g. EMG, body and eye movements, etc. are due. In an advantageous embodiment of this artifact cleaning process, each channel of the EEG signal is first subdivided into (time) intervals (of, for example, 2 seconds each) within the framework of a so-called windowing. Each of these intervals then calculates an average (arithmetic mean, median, etc.) of the amplitude per channel, this average value being subtracted from each sample signal value. From this new signal, the maximum value and the minimum value are determined in a next step and the difference between these two extreme values is determined. If the amount of this difference exceeds a certain fixed threshold (e.g., 200pV), that interval for that channel will be disregarded for further analysis.

3) Spectral analysis (power) of the artifact-corrected EEG signal in certain frequency bands. The power spectral analysis is carried out in particular in the frequency bands delta (0.5 - 3.5Hz), theta (4-7.5Hz), alpha (8-12Hz) and beta (12.5-22Hz). In the context of the invention may also be higher Frequency bands are taken into account in the power spectral analysis. Preferably, at least two channels of the

aretfactured EEG signal, which are assigned to different brain regions or vascular flow areas, considered separately.

4) Trend correction of the traces resulting from the power spectral analysis. In order to identify power changes in the considered frequency bands, which occur not globally, but focal (only in a specific brain region or a specific vascular flow area) and thus are likely correlate of a DCI, the common trend of all considered tracks is calculated and each Track adjusted for this common trend. The track refers to the power values resulting from the power spectral analysis, which are assigned to a channel or channel group of the EEG signal and a frequency band under consideration. In order to achieve a corresponding correction, the common fluctuations of all derivatives are preferably calculated and the individual derivatives are then calculated with the common variation. This is achieved in particular by first calculating the median of the power history in each track; each track is then normalized with its median. Subsequently, the median track is formed from the averaged tracks. As a result, the trend track calculated in this way divides each of the original tracks. As a result of the trend adjustment, the individual derivatives are formed with greatly narrowed distribution widths, so that regional events such as a focal power decrease in a certain frequency range can be detected with fail-safety.

5) Detecting deviations of the trended power values of at least one track from a baseline. Preferably, all recorded tracks are examined for deviations from the respective baseline. The baseline is calculated in an expedient embodiment of the method for each track from a part of the EEG signal acquired at the beginning of the EEG recording, in particular during the first 6 to 24 hours after the start of the EEG recording. For the detection of vasospasm-related cortical ischemia following a subarachnoid hemorrhage (SAB), the basline is expediently made from one part of the EEG signal recorded within the first 24 hours after a subarachnoid haemorrhage has been detected, which is usually vasospasm-free.

6) Display of the detected deviations in real time and / or triggering of an alarm, if a significant focal deviation of the power from the baseline in at least one selected frequency band is determined. In this case, a deviation is recognized to be significant in particular if it exceeds a predetermined tolerance (in particular a threshold value or threshold value range). In the former case, in any case, those tracks are displayed in which a significant deviation from the baseline is found. A significant, in particular focal deviation is evaluated according to the method as an indication of a DCI. The display or triggering of an alarm in this case allows the treating medical team a fail-safe early detection of DCI, without the need for special special medical knowledge on the part of the practitioner.

The device according to the invention comprises an EEG device for detecting an EEG signal of a patient as well as a control unit connected therewith in terms of data communication technology. The control unit is - switching and / or program technology - set up for the automatic implementation of the method according to the invention described above. The control unit is thus in particular designed to

to control the EEG device for continuous long-term derivation of an EEG signal of the patient and to store this EEG signal,

- to clean up the EEG signal for artifacts,

- To subject the artifact-corrected EEG signal of a power spectral analysis in special frequency bands and

- To detect deviations of all or individual tracks from a baseline in real time and to capture

- indicate recorded deviation (at least significant deviations) and / or trigger an alarm if significant focal deviation The power of the baseline is determined in at least one selected frequency band.

In particular, the control unit is constituted by a control computer (for example, a personal computer) in which a control program (that is, control software) is installed, so that the above-described method is automatically performed when the control program is executed.

With regard to the design details and variants of the functional design of the device, reference is made mutatis mutandis to the above statements relating to the method according to the invention.

An embodiment of the invention will be described below with reference to a drawing. Show:

Fig. 1 is a schematic block diagram of an apparatus for

Measurement of cerebral perfusion (cerebral blood flow), with an EEG device and a control unit connected to it by data communication technology,

Fig. 2 is a flowchart performed by the device

Method, and

FIGS. 3-5 show results of a pilot study using the method.

1 shows, in a crude schematic simplification, a device 1 for measuring the compromise of brain function, and thus indirectly for measuring cerebral perfusion. The method is used in particular for the detection of a vasospasm-related imminent cortical ischemia in patients with Subarachnoidalblutungen means of quantitative EEG.

The device 1 comprises an EEG device 2 and a control computer 3 (eg a personal computer). The control computer 3 is via a data transmission line 4 (eg in the form of an Ethemet connection of an ordinary in itself LAN) connected to the EEG device 2 for data transmission. In the control computer 3, a control program 5 is installed executable, which - as described in more detail below - on the one hand implemented functions for controlling the EEG device 2 via control commands C and on the other hand functions for the evaluation of the detected by the EEG device 2 EEG signal E.

The EEG device 2 comprises a number of electrodes 6 for measuring the electrical activity of the brain, which are applied during operation of the device 1 to the scalp of a patient 8, shown by way of example on a patient couch 7 in FIG. The number and arrangement of the electrodes 6 on the head 9 of the patient 8 is preferably selected according to the 10-20 system. However, only a subset of the electrodes provided according to the 10-20 system can be used in the operation of the device 1, in particular - according to the nomenclature of the 10-20 system - the electrodes C3 / C4, P3 / P4, 02/01 , T3 / T4, F3 / F4 (with electrodes F3 / F4 alternatively replaced by electrodes F5 / F6).

To the control computer 3 means for data input and data output are connected in a conventional manner, in particular a screen 10th

The automatically performed during the course of the control program 5 by means of the device 1 method is shown schematically in Fig. 2.

Thereafter, in a first step 20 - by appropriate control of the EEG device 2 - automatically the EEG signal E of the patient 8 over a period of at least 4 days (preferably between 5 and 12 days) using a sampling rate of the multi-channel -EEG derivative of 256Hz recorded in a frequency range of 0.08 Hz to 86 Hz.

The recorded EEG signal E is stored in a step 21 by the control program 5 on a hard disk of the control computer 3. For automated artifact correction, in a step 22, each channel of the EEG signal E is divided into (time) intervals of 2 seconds each. In these intervals, in each case one (in particular arithmetic) average of the amplitude per channel is calculated, this mean value being subtracted from each sample signal value. From this new signal, the maximum value and the minimum value are determined and the difference between these two extreme values is determined. Exceeds the amount of this difference a certain specified

Threshold of 200pV, this interval will not be considered for this channel for further analysis.

At the thus artifact-adjusted raw EEG signals, a power spectral analysis is performed in a step 23. For this purpose, the

Artifact-adjusted raw EEG signals subjected to a short-time Fourier transform (STFT) to determine the corresponding spectral frequency components in a time window (preferably with a time extension of 2 seconds). For the windowing, a Welch window (essentially parabolic characteristic) is used, taking into account a 50% overlap of adjacent intervals. From the frequency distribution (in particular with a spectral resolution of 0.5 Hz) of the signal, the corresponding power values in the various frequency bands (delta, theta, alpha and beta band, and possibly higher frequencies, such as gamma) are now determined ,

From the thus obtained power values in the individual channels or a plurality of channel groups combined according to belonging to a common brain region, the respective median value of the power in the respective frequency band is automatically determined for every 60 second intervals, the individual brain activity of the patient in this time interval should represent.

The thus obtained absolute values of the power of the respective frequency bands in the individual channels or regionally grouped channel groups are averaged over a period of one minute. In order to differentiate focal changes from global changes, which may have a variety of causes (increase in intracranial pressure, iatrogenic change in sedation depth, metabolic processes, etc.), detrending is performed in step 24: in 60s EEG intervals The absolute power value of each channel is then divided by this median, thereby removing generalized EEG changes over time from the individual EEG channels. Signals are thus adjusted for the trend.

In order to detect signs of vasoplasmic cortical ischemia and the resulting compromise of brain function, in a step 25, a characteristic of the absence of vasoplasmic signal curve ("baseline") of the trend-corrected EEG signal is determined.

The baseline is determined from the EEG signal recorded during the first 6 to 24 hours after subarachnoid hemorrhage at the beginning of EEG monitoring. This is based on the finding that no cerebral vasospasms are regularly present during this period. Rather, the occurrence of vasospasm after clinical studies is expected from the third day after the subarachnoid hemorrhage.

Significant deviations D of individual traces (e.g., decrease in alpha power) from this baseline are displayed visually on screen 10 in real time numerically or graphically (trend curve, color bar, etc.). The attending physician will be shown in this way threatening events clearly. The treating physician can thereby recognize signs of a DCI fail-safe without the need for expert knowledge regarding EEG diagnostics.

Optionally, in addition, an alarm A is automatically triggered in a step 26 if one or more tracks deviate from the baseline by more than a predetermined tolerance value (eg 40% of the baseline) over a specific time (eg more than 4 hours). The alarm A can according to FIG. 1 as - in particular due - visual notice on the screen 10 are issued. Additionally or alternatively, the alarm A can also be generated in other ways, for example by switching on a warning light, generating an acoustic signal, automatically sending an e-mail, SMS, etc.

If the generation of the alarm A according to step 26 is provided, the constant display of the trend-corrected EEG signal and its deviation A from the baseline can also be omitted. Thus, in a variant of the method, it is provided that the trend-corrected EEG signal and its deviation A from the baseline is displayed on the screen 10 only on user interaction, in particular on request by the attending physician.

The steps 20 to 26 are performed in parallel to each other. The recorded EEG signal E is therefore already evaluated during the duration of the recording (steps 20, 21) (steps 22 to 26).

By the method described above, spasms can be detected earlier by several hours or days than by previous methods (in particular transcranial Doppler sonography, CCT). Thus, it is possible to dramatically advance the therapeutic window of this serious complication of SAB.

Description of a pilot study carried out by the method:

Aneurysmal subarachnoid hemorrhage (SAB) is still associated with high morbidity and mortality. (1) Delayed cerebral ischaemia (DCI) has a significant impact on mortality and functional outcome. (3) Vasospasm, defined as an arterial narrowing measurable by angiography or Trascranial Doppler sonography (TCD) has been strongly associated with DCI. (4-5) However, there is evidence to show that DCI is also the result of other mechanisms such as cortical spreading depression

microcirculatory dysfunction, which can not be detected by TCD. (5-6) Approximately 20% of DCI after SAB are not clinically often identified or identified too late, mostly due to limited clinical assessment in severely affected, comatose SAB patients. (7) Early detection of impending ischemia, which is a prerequisite for the prevention of infarcts, therefore remains a challenge.

The EEG sensitively reflects metabolic deterioration resulting from decreased cerebral blood flow. (8-9) It has been shown that EEG changes such as flattening and loss of higher frequencies occur rapidly when cerebral blood flow is below 0.16-0.17 ml ^" G ¹ * min ¹ drops. (10-1 1) Continuous EEG monitoring has been shown to be a promising approach for the early detection of impending DCI in SAB patients. (12-19) Powerspectral analysis of crude EEG data provides a rapid and robust screening method for long-term EEG recordings. (20) Earlier studies (15), (12) used their information from artifact-free, visually pre-selected EEG clips or used only a 2-channel EEG, (3), our goal was to develop a semiautomatic method to detect trends in the field

show and analyze unselected multichannel EEG over several days. In doing so, we wanted to provide a real-time analytics methodology that would enable non-EEG everyday experts to reliably detect and locate DCI-related threats at the earliest possible stage.

Participation in the study was offered to all patients enrolled in the Neurological Intensive CaseUnit-NICO at the University Hospital Erlangen between November 2011 and February 2013, who suffered from a non-traumatic SAB of any clinical severity. The diagnosis had to be made through a CT scan or, if negative, through

Xanthochromism of the liquor (SF) to be confirmed. The following inclusion criteria had to be met: age> 18 years, symptom onset 24 hours before admission. Patients were excluded from the study for angiographically or sonographically proven vasospasm prior to EEG monitoring. The study was endorsed by the local ethics committee. All patients or their relatives signed a written statement before beginning studies.

All patients received diagnostic angiography on the day of admission. In all patients, SAB was caused by aneurysmal bleeding. All aneurysms were treated by endovascular coiling. An extraventricular drainage was introduced in case of an imminent or existing CSF disturbance. All patients were treated with nimodipine. (22)

Our current SOP included monitoring of intracranial pressure, systolic, diastolic and mean arterial blood pressure, heart rate, temperature and oxygen saturation. Neurological status was assessed at least three times daily by ICU staff. All patients received transcranial Doppler (TCD) or duplex (TCCS) scans at least every other day.

CT examinations were performed with clinical deterioration or increase in Doppler frequencies.

Relevant vasospasm was diagnosed by TCD (defined as systolic peak frequencies above 4 kHz) or TCCS (defined as flow rates> 200 cm / s or an increase of> 50 cm / s in 24 hours in accordance with the guidelines of the German Neurological Society) or by angiography. The clinical criteria of DCI were a new focal neurological deficit that was not caused by other conditions such as an increase in intracranial pressure or rebleeding.

In the case of relevant vasospasms or DCI, a triple H therapy was used

(Hypervolemia, hypertension, hemodilution) started. (23) In persistence of neurological deficits, patients received interventional angiography and were treated by means of chemical vasospasmolysis or balloon angioplasty. (24-25) EEG monitoring was started within the first 48 hours of admission and maintained for a maximum of 12 days, but interrupted earlier if the patient was discharged from the intensive care unit or at the request of the patient or relative. We used surface cup electrodes, fixed with collodion. We limited the EEG to a limited number of 10 electrodes, attached to the international 10-20 system: C3 / C4, P3 / P4, O2 / O, T3 / T4, F3 / F4 (alternatively, if unavoidable because of head dressings F5 / 6 instead of F3 / F4). Simplified EEG montages are widely used and successfully used for EEG monitoring under intensive care conditions, although general recommendations regarding optimal assembly and number of electrodes are still lacking. (26), (13), (12) Our goal was a compromise between stably maintaining record and coverage of all vascular territories. The EEG data was digitized with a sampling rate of 256 Hz and a 0.08 Hz high pass filter and a low pass filter of 86 Hz. We used bipolar mounting

(F4-C4, T4-P4, P4-O2, F3-C3, T3-P3, P3-O1), under the hypothesis that

F4-C4 (F3-C3) and T4-P4 (T4-P3) represent the anterior and medial canal area, while P4-O2 (P3-O1) capture changes in the posterior circulation.

To perform a reliable semi-automatic qEEG analysis, automated artifact cleansing is essential. To achieve this, we used a standardized and efficient technique that has been described and evaluated in detail in previous work. (20, 27)

The remaining raw EEG signal went through a two second short-term Fourier transform with a 50% overlap of adjacent epochs. The frequency distribution in each epoch was identified, which made it possible to determine the power values in the following frequency bands: Delta 0.5 - 3.5Hz, Theta 4-7.5Hz, Alpha 8-12Hz, Beta 12.5-22Hz. The median of each minute of power values in these bands and the alpha-delta ratio were calculated separately to produce a linear time series for further analysis. All EEG channels are subject to common variations due to global effects such as changes in intracranial pressure, depth of sedation and analgesia or circadian rhythm. In order to unmask focal changes that we posit to be correlates of impending DCI, several procedures were needed.

Because all channels in their power are highly correlated with each other over time (R> 0.92, p <0.001 for each combination of channels) - even though they have different power levels - we established a trend cleanup procedure that eliminated common variations and uncovered regional power drops.

First, each median one-minute power value was divided by the median of all 1-minute power values of the same channel over 24-hour EEG recording for normalization. Second, the trend adjustment followed. For this purpose, the median of all normalized power values of the 6 channels was calculated at one point in time in order to generate the normalized common trend of all channels. Each channel was then shared by the common trend to eliminate common variations and emphasize focal changes.

To detect focal EEG changes over time from day two of monitoring, the median was 60 normalized and trend adjusted one minute power values per hour and in each channel compared to the median of the normalized and trend adjusted power values of the same channel of the baseline period. The baseline was defined as a minimum of 6 and a maximum of 12 hours of uninterrupted EEG recording within the first 24 hours of EEG monitoring, assuming that there were none at this time

Vasospasmen or DCI are present.

We individually correlated qEEG changes in each patient

Vasospasm or DCI detected by TDC / TCCS or CT / MRI were. To calculate the strength of the association between qEEG changes and DCI, we used receiver operator characteristics (ROC), looking at each hemisphere separately.

In order to explore the validity of the level and duration of EEG power depletion, we assessed the sensitivity and specificity for various power-draw scenarios, from 10 to 100%, and consistently a minimum of 1 to a maximum of 8 hours. To define the optimal threshold on the ROC curve, we calculated the Youden Index (J), a function of sensitivity and specificity (28). The Youden index provides the optimal cut-off for the differentiation strength of a marker, if sensitivity and specificity are weighted equally. It moves between 0 and 1, with values near 1 illustrating the effectiveness of the marker.

Of 22 consecutive SAB patients enrolled during the study phase, 12 met the inclusion criteria and were included. 5 patients were sedated and mechanically ventilated during the entire EEG monitoring, 7 patients were awake and spontaneously breathing.

Fifty percent (N = 6) of the patients developed vasospasms (Table 2). The media (MCA) and anterior stromal areas (ACA) were most commonly affected. Isolated vasospasm of the posterior cerebral artery (PCA) and basilar artery (BA) did not occur. In two patients, vasospasm led to infarcts, visible by CT / MRI. One of these patients died as a result of a complete media and anterior infarction.

Decreases in total alpha and theta capacity showed the strongest association with vasospasms / DCI (Youden Index J = 0.79 point, Table 3). In both frequency bands a power decrease of 40% marked the optimal threshold. The duration of the power decrease, which had the highest significance, differed slightly; it was five hours for alpha and six hours for theta.

The optimal combined cut-off for power loss and duration proved to be highly sensitive (100%) with a specificity of 78% for both frequency bands (Table 3). Changes in the beta band showed a lower correlation with Vasospasm / DCI, presumably due to contamination with muscle artifacts. The weakest correlation existed between DCI and delta power fluctuations.

Consequently, the alpha-delta ratio had only a moderate correlation

Vasospasms / DCI (sensitivity 62.5%, specificity 85.7%). In 5 out of 6 cases, EEG changes preceded vasospamen / DCI detection by TCD or CCT / MRI. In Patient 2, TCD vasospasm detection was 4 days ahead of EEG power loss. On average, EEG power decreases occurred 2.3 days earlier (SD 3.3, area 2.5) than detection of vasospasm / DCI in TCD / TCCS or CCT / MRI.

In patient 10, the Alpha-Power dropped significantly in the left PCA region for 31 hours without evidence of vasospasm / DCI in TCD / TCCS or CCT / MRI. Visual inspection of the original EEG at that time revealed focal slowing and amplitude flattening without artifacts obscuring the EEG. Patient 12 exhibited ubiquitously accelerated flow velocities

TCD / TCCS on day 4 (Table 2), which did not reflect the EEG. The patient, however, did not develop DCI and recovered without neurological deficits. In addition, the patient had hemolytic anemia with hemoglobin levels that dropped below 100 g / l from day 4 after symptom onset and did not recover until day 13, in parallel with TCCS changes. We therefore considered anemia to be the cause of TCCS flux acceleration, while we found the assumption of ubiquitous critical vasospasms, which did not lead to infarcts or neurological deficits, to be less plausible.

Patient 9 had a pronounced alpha decrease bilaterally, starting on day 3. TCD detected vasospasm in the right MCA on day 7. During the entire monitoring, no vasospasm was seen on the left side. While the Alphapower of the right hemisphere recovered in the following days, it continued to fall in all channels on the left side. CCT and MRI on day 10 showed complete infarction in the left anterior and medial canal area, resulting in death of the patient. The aim of the study was to develop an automated, observer-independent, EEG-based method to identify impending infarcts in SAB patients. To ensure clinical applicability, we applied a new EEG algorithm based on continuous EEG monitoring without visual preselection of specific EEG sections.

In this setting, a 40% decrease in alpha and theta power lasting longer than 5 and 6 hours, respectively, showed a strong association with vasospasms / DCI and was able to complete this complication 2.3 days before detection by other methods such as TCD in the Predict most cases.

However, the approach has been challenging in several ways. First, maintaining high quality EEG monitoring over several days and intensive care conditions is difficult and laborious. (29) Secondly, it is inevitable that EEG leads in the ICU are contaminated by various artifacts, and thirdly, the EEG sensitively but nonspecifically responds to various systemic changes that make the EEG difficult, DCI from other factors, such as metabolic changes, changes of intracranial pressure or sedation effects (30-31). We tried to overcome these difficulties by using a reduced yet multichannel assembly that could be sustained over several days at a reasonable cost and yet able to reliably detect focal changes. In addition, we implemented an efficient automatic artifact elimination that has been extensively evaluated in previous studies (20, 27).

This allowed the undistorted analysis of changes in the alpha, theta, and delta bands. However, it was impossible to automatically differentiate between beta activity and muscle artifacts, resulting in low usability of beta fluctuations. The key to distinguishing DCI and other induced EEG changes has been the observation of vasospasms as predominantly focal events, (32-33) while most other events affecting the EEG of ICU patients are global changes cause. Due to the trend adjustment, generalized fluctuations are excluded from the further analysis, whereby focal changes in the EEG are made visible. A theoretical disadvantage of this approach could be the lack of evidence of global changes that can not be predicted with certainty that they are not also caused by the SAB. In any case, in 5 of 6 patients in our cohort, DCI was clearly confined to one or two adjacent arterial flow areas, and could be so well captured by the algorithm.

One patient had ubiquitously increased flow rates in TCD / TCCS, which were more likely to be due to anemia (34-36) than to generalized vasospasm. In addition, the patient developed no infarction or neurological deficits. The EEG was not susceptible to this particular source of error and was therefore superior to the TCCS here.

To be a reliable marker for DCI, Alpha and Theta power had to stop for a certain amount of time. The optimal cut-off duration was slightly shorter for alpha than for theta. Accordingly, alpha appears to be the preferred parameter in both conscious and comatose patients. In our cohort, physiological variations such as sleep-wake patterns (including alpha rhythm) were symmetric in their appearance, at least partially below trend trimming, and thus remained below the power threshold associated with vasospasms / DCI. The method therefore seems to be as applicable for less severely affected awake patients as severely affected SAB patients.

One patient in our cohort without vasospasm / DCI detection in TCD and imaging had a significant left occipital alpha decrease, which according to the study protocol had to be interpreted as false positive. Since the therapy also decreased in this patient, an alternative explanation would be that the sensitive EEG correctly identified diminished cortical perfusion as a sign of impending DCI in this area, but therapeutic measures or the natural course of the disease prevented Perfusion deficit developed into a permanent and more severe DCI, which would have been reflected in other diagnostic procedures.

Unexpectedly, deltapower changes and the alpha-delta-power ratio showed only a weak to moderate correlation with vasospasms in our cohort. At first glance, this seems contradictory to other studies. Ciaassen and his colleagues saw a high probability of patients with an alpha-delta-power ratio falling by more than 10% in 6 consecutive measurements or by more than 50% in a single shot to develop DCI. (15) Stuart et al. also described the alpha-delta-power ratio as the most predictive marker of DCI in the intracortical EEG. (14) One possible explanation for the divergent findings could be a hypothesis proposed by Machado et al. (37), who used qEEG in patients with median infarctions and found that delta capacity increases occurred in the ischemic center, while endangered tissue, penumbra and environmental edema were characterized by alpha and theta changes. Similarly, alpha in our study marks an early point in time when perfusion drops below a critical level and produces endangered tissue but no permanent ischemia. This thesis is supported by patient 3, who developed an infarct during the course of the disease.

The delta power increased two days after the alpha drop, probably as an expression of now irreversible ischemia (Figure 4). In this context, the alpha-delta-power ratio could be a valuable parameter for

Predicting the prognosis and predicting the functional outcome of the patient, whereas Alphapower could be the superior parameter to prevent infarction by countermeasures.

Ciaassen and his co-workers described a strong correlation between a drop in the alpha-delta power ratio and DCI (100% sensitivity, 76% specificity), with EEG experts visually pre-selecting artifact-free clips (15). Our approach introduces an automated technique that does not require the presence of an EEG expert to select and analyze the EEG and is therefore suitable for daily use by ICU staff. Vespa et al. chose a similar approach and identified relative alpha variability (RA) as most sensitive to indicating impending ischemia. (12) Although highly sensitive, RA did not respond exclusively to vasospasms / DCI but fell significantly as a result of various other events, eg, increase in intracranial pressure or Herniation, which resulted in low specificity (50%).

Labar et al. used a 2-channel EEG and compared total power changes by applying stable EEG trends from day to day and left versus right against each other. (13) To be considered stable, a trend in this study had to last for 24 hours. The algorithm of our study has already shown a focal alpha decrease over five hours as highly sensitive and specific for vasospasm / DCI.

Our study has several limitations. First, the small number of patients limits the statistical power of the results. The results must therefore be confirmed in larger cohorts. Another restriction of the algorithm is the need for a vasospasm-free baseline EEG. The method was therefore not used in patients who were admitted to our clinic later in the course of their illness, if they had already existing

Vasospasm.

In summary, our study introduces an automated, easy-to-use, investigator-independent, reliable, and robust ischemia-detecting algorithm for SAB patients, which can be done early enough to take countermeasures that may still prevent infarction. In contrast to established methods such as TCD or imaging, the real-time nature of the method provides the basis for an EEG-based alarm system with a high potential for improving the functional outcome of SAB patients. Brief Description:

Fig. 3 - correlation of alpha power, TCD and imaging: patient 9, SAH 111 °, aneurysm of the anterior communicating artery; A. TCD plot of critical vasospasm of right MCA on day 7-10; no frequency increase in the left MCA; B. EEG Alpha Power: significant decrease in aplha power in all channels on days 3 and 4 (black arrow); Recovery of alpha power in channels T4P4 and F8C4 on day 6/7 (gray bearing); no recovery in T3P3 / F7C3; C. CCT and D. Perfusion-weighted MRI on day 10 shows complete infarction of the left ACA and MCA area (CCT and MRI courtesy of A. Doerfler, MD, Department of Neuroradiology, University Hospital Erlangen)

Fig. 4 - Alpha and delta power during various ischemic conditions: patient 3, SAH III, aneurysm of the anterior communicating artery. A. Critical vasospasm in the left and right ACA on day 9, detected by TCD, B. Decrease in alpha power on days 4 and 5 in F8C4 and F7C3 (black arrow), no relevant EEG changes in T4P4 and T3P3; C. increase in delta power in F8C4 and F7C3 on day 7/8 (gray arrow), D. moderate decrease in alpha / delta ratio from day 6 (black arrow).

Fig. 5 - Statistical analysis (supplementary material): Youden index (J) for different frequency bands. Percentage of power decrease (y-axis) plotted against the Dazer's power increase in hours (x-axis). Youden index rendered by gray scale, according to the bar shown on the right. High J values, represented by bright gray values, imply a strong correlation. Best results were in the alpha band (J = 0.79, optimal threshold 40% - drop, 5 hours duration) and in the theta band (J = 0.79, optimal threshold 40% - drop, 6 hours duration). Changes in beta, delta and alpha-delta ratios were less sensitive and specific for DCI (beta: J = 0.59, delta: J = 0.34, alpha-delta ratio: J = 0.48). ROC curve for varying power acceptance values and fixed duration (5 hours); for AUC values see Table 3. references

1. Johnston SC, Selvin S, Gress DR. The burdens, trends, and demographics of mortality from subarachnoid hemorrhage. Neurology 1998; 50: 1413-1418.

2. Rabinstein AA, Friedman JA, Weigand SD, et al. Predictors of cerebral infarction in aneurysmal subarachnoid hemorrhage. Stroke 2004; 35: 1862-1866.

3. Vergouwen MD, llodigwe D, Macdonald RL. Cerebral infarction after subarachnoid hemorrhage to vasospasm-dependent and - independent effects. Stroke 201 1; 42: 924-929.

4. Crowley RW, Medel R, Dumont AS, et al. Angiography vasospasm is strongly correlated with cerebral infarction after subarachnoid hemorrhage. Stroke

201 1; 42: 919-923.

5. Brown RJ, Kumar A, Dhar R, Sampson TR, Diringer MN. The relationship is delayed infarets and angiographic vasospasm after aneurysmal subarachnoid hemorrhage. Neurosurgery 2013; 72: 702-708.

6. Threesome JP. The role of spreading depression, spreading depolarization and spreading ischemia in neurological disease. Nat Med 201 1; 17: 439-447.

7. Schmidt JM, Wartenberg KE, Fernandez A, et al. Frequency and clinical impact of asymptomatic cerebral infarction due to vasospasm after subarachnoid hemorrhage. J Neurosurg 2008; 10911052-1059.

8. Astrup J, Siesjo BK, Symon L. Thresholds in cerebral ischemia - the ischemic penumbra. Stroke 1981; 12: 723-725.

9. O'Gorman RL, Poil SS, Brandeis D, et al. Coupling Between Resting Cerebral Perfusion and EEG. Brain Topogr 2012.

10. Sundt TM, Jr., Sharbrough FW, Anderson RE, Michenfelder JD. Cerebral blood flow measurements and electroencephalograms during carotid

endarterectomy. J Neurosurg 1974; 41: 310-320.

11. Diedler J, Sykora M, Bast T, et al. Quantitative EEG correlates of low cerebral perfusion in severe stroke. Neurocrit Care 2009; 1 1: 210-216.

12. Vespa PM, Nuwer MR, Juhasz C, et al. Early detection of vasospasm after acute subarachnoid hemorrhage using continuous EEG ICU monitoring.

Electroencephalogram Clin Neurophysiol 1997; 103: 607-615.

13. Labar DR, Fish BJ, Pedley TA, Finch ME, Solomon RA. Quantitative EEG Monitoring for patients with subarachnoid hemorrhage. Electroencephalogr Clin Neurophysiol 1991; 78: 325-332.

14. Stuart RM, Waziri A, Weintraub D, et al. Intracortical EEG for the detection of vasospasm in patients with poor-grade subarachnoid hemorrhage. Neurocrit Care 2010; 13: 355-358.

15. Ciaassen J, Hirsch LJ, Kreiter KT, et al. Quantitative continuous EEG for detecting delayed cerebral ischemia in patients with poor-grade subarachnoid hemorrhage. Clin Neurophysiol 2004; 1 1512699-2710.

16. Ciaassen J, Taccone FS, HornP, Holtkamp M, Stocchetti N, Oddo M. Recommendations on the use of EEG monitoring in critically ill patients: consensus statement from the neurointensive care section of the ESICM. Intensive Care Med 2013.

17. Rathakrishnan R, Gotman J, Dubeau F, Angle M. Using continuous electroencephalography in the management of delayed cerebral ischemia following subarachnoid hemorrhage. Neurocrit Care 201 1; 14: 152-161.

18. Rivierez M, Landau-Ferey J, Rough R, Grosskopf D, Philippon J. Value of electroencephalogram in prediction and diagnosis of vasospasm after intracranial aneurysm rupture. Acta Neurochir (Vienna) 1991; 110: 17-23.

19. Ciaassen J, Taccone FS, Horn P, Holtkamp M, Stocchetti N, Oddo M. Recommendations on the use of EEG monitoring in critically ill patients: consensus statement from the neurointensive care section of the ESICM. Intensive Care Med 2013; 39: 1337-1351.

20. Hopfengartner R, Kerling F, Bauer V, Stefan H. An efficient, robust and almost method for the offline detection of epileptic seizures in long-term scalp EEG receptions. Clin Neurophysiol 2007; 18: 2332-2343.

21. Fisher CM, Kistler JP, Davis JM. Relation of cerebral vasospasm to subarachnoid hemorrhage visualized by computerized tomographic scanning. Neurosurgery 1980; 6: 1-9.

22. Kronvall E, Undren P, Romner B, Saveland H, Cronqvist M, Nilsson OG.

Nimodipine in aneurysmal subarachnoid hemorrhage: a randomized study of intravenous or oral administration. J Neurosurg 2009; 1 0: 58-63.

23. Treggiari MM. Hemodynamic management of subarachnoid hemorrhage. Neurocrit Care 2011; 15: 329-335. 24. Aburto-Murrieta Y, Marquez-Romero JM, Bonifacio-Delgadillo D, Lopez I, Hemandez-Curiel B. Endovascular treatment: balloon angioplasty versus nimodipine intra-arterial for medically refractory cerebral vasospasm following aneurysmal subarachnoid hemorrhage. Vase Endovascular Surg 2012; 46: 460-465.

25. Sayama CM, Liu JK, Couldwell WT. Update on endovascular therapies for cerebral vasospasm induced by aneurysmal subarachnoid hemorrhage.

Neurosurg Focus 2006; 21: E12.

26. Kuli LL, Emerson RG. Continuous EEG monitoring in the intensive care unit: technical and staffing considerations. J Clin Neurophysiol 2005; 22: 107-1 18.

27. Hopfengartner R. Automatic Seizure Detection in Long-term Scalp EEG Monitoring Using an Adaptive Thresholding Technique: A Validation Study for Clinical Routine, (submitted) 2013.

28. Youden WJ. Index for rating diagnostic tests. Cancer 1950; 3: 32-35.

29. Young GB, Campbell VC. EEG monitoring in the intensive care unit: pitfalls and caveats. J Clin Neurophysiol 1999; 16: 40-45.

30. Huang Z, Dong W, Yan Y, Xiao Q, Man Y. Effects of intravenous mannitol on EEG recordings in stroke patients. Clin Neurophysiol 2002; 113: 446-453.

31. Huang Z, Dong W, Yan Y, Xiao Q, Man Y. Effects of intravenous human albumin and furosemide on EEG recordings in patients with intracerebral hemorrhage. Clin Neurophysiol 2002; 113: 454-458.

32. Miller CM, Palestrant D. Distribution of delayed ischemic neurological deficits after aneurysmal subarachnoid hemorrhage and implications for regional neuromonitoring. Clin Neural Neurosurg 2012; 114: 545-549.

33. Thut G, Nietzel A, Brandt SA, Pascual-Leone A. Alpha-band electroencephalographic activity over occipital cortex indexes visuospatial attention bias and prelicts Visual target detection. J Neurosci 2006; 26: 9494-9502.

34. Brass LM, Pavlakis SG, DeVivo D, Piomelli S, Mohr JP. Transcranial Doppler measurements of the middle cerebral artery. Effect of hematocrit. Stroke

, 1988; 19: 1466-1469.

35. Ameriso SF, Paganini-Hill A, Meiselman HJ, Fisher M. Correlates of middle cerebral artery blood velocity in the elderly. Stroke 1990; 21: 1579-15 83.

36. Isikay CT, Uzuner N, Gucuyener D, Özdemir G. The effects of hematocrit and age on transcranial Doppler measurements in patients with recent ischemic stroke. Neural India 2005; 53: 51-54; discussion 54.

37. Machado C, Cuspineda E, Valdes P, et al. Assessing acute middle cerebral artery ischemic stroke by quantitative electric tomography. Clin EEG Neurosci 2004; 35: 116-124.

38. Castle House L, Hilker R, Dohmen C, et al. Early electroencephalography in acute ischemic stroke: prediction of a malignant course? Clin Neural Neurosurg

2007; 109: 45-49.

39. Sheorajpanday RV, Nagels G, Weeren AJ, van Putten MJ, De Deyn PP. Quantitative EEG in ischemic stroke: correlation with functional status after 6 months. Clin Neurophysiol 201 1; 122: 874-883.

40. Sheorajpanday RV, Nagels G, Weeren AJ, De Deyn PP. Quantitative EEG in ischemic stroke: correlation with infarct volume and functional status in posterior circulation and lacunar syndromes. Clin Neurophysiol 201 1; 122: 884-890.

Tables 1 to 3:

Table 1: Patient characteristics

Arteria cerebri media; PCA: posterior cerebral artery, PCOM: posterior communicating artery SD: standard deviation; EVD: external ventricular drainage

Table 2: Identification and localization of vasospasm / DCI by different modalities (qEEG, TCD, CCT / MRI)

Patient EEG (reduced alpha power) ^* TCD (enhanced CBF) CT / MRI (infarction)

Channel Day SAH Area Day SAH Area Day SAH

No DCI

1

4; ;

5 - - -

7 - - -

8th - - -

10 P301 3 - -

DCI

2 F4C4 6 R ACA 2

R MCA

R + L

T4P4 6 5

MCA

3 F4C4 4 R ACA 9 R ACA 12

F3C3 5L ACA 9L ACA 12

6 F4C4 3 R MCA 6

9 T4P4 3 R MCA 6

F3C3 3L ACA "9T3P3 3L MCA" 9

11 F3C3 4 L ACA 12

T3P3 4L MCA 9

12 - - everywhere 4 -

**: Patient died after Himinfarkt

* EEG event, defined as a 40% decrease in alpha power for a duration of a 5 hours; L: Left; : Right; ACA: Arteria cerebri anterior; MCA: Arteria cerebri media; PCA: Arteria cerebri posterior; CBF: cerebral blood flow

Table 3: Relationship between the decrease in qEEG power and vasospasms / DCI

Specificity and sensitivity equally weighted. LIST OF REFERENCE NUMBERS

1 device

2 EEG device

3 control computer

4 data transmission line

5 control program

6 electrodes

7 patient couch

8 patient

9 head

10 screen

20 step

21 step

22 step

23 step

24 step

25 step

26 step

command

EEG signal

deviation

alarm

Claims

Expectations

1. Method for measuring the cerebral perfusion of a patient (8) with the following steps:

- Continuous long-term recording and storage of an EEG signal (E) from the patient (8);

- Cleaning the EEG signal (E) from artifacts;

- Power spectral analysis of the artifact-corrected EEG signal in specific frequency bands;

- Cleaning of the traces obtained by the power spectral analysis by a common trend of all traces;

- Detection of deviations (D) of the detrended power values of at least one trace from a baseline; and

- Display of significant deviations (D) and/or trigger an alarm (A) if a significant deviation (D) is detected.

2. Method according to claim 1,

where the EEG signal (E) is recorded for at least 12 hours.

3. Method according to claim 1 or 2,

wherein an electrode arrangement according to the 10-20 system or a subgroup of the electrodes provided according to this system is used to record the EEG signal (E).

4. Method according to one of claims 1 to 3,

in order to remove artifacts from the EEG signal (E), each channel of the EEG signal (E) is divided into time intervals, with an average value of the amplitude per channel of the EEG signal (E) being calculated in each interval. net, this mean value being subtracted from each sample signal value of the EEG signal (E), the maximum value and the minimum value as well as the difference between these two extreme values being determined from this new signal in each interval, and each interval of a channel , in which the amount of this difference exceeds a certain specified threshold, is not taken into account for further analysis.

Method according to one of claims 1 to 4,

whereby the artifact-corrected EEG signal is subjected to power spectral analysis at least in the frequency bands delta (0.5 - 3.5Hz), theta (4-7.5Hz), alpha (8-12Hz) and beta (12.5-22Hz). becomes.

Method according to one of claims 1 to 5,

whereby at least two channels of the power spectrum analysis are used

Artifact-corrected EEG signals that are assigned to different brain regions or vascular flow areas can be considered separately.

Method according to one of claims 1 to 5,

For the trend adjustment, the common fluctuations of all derivatives are calculated and the individual derivatives are then offset against the common fluctuation.

8. Method according to claim 7,

for the trend adjustment, the median of the power curve is first calculated in each track, each track being normalized with its median, the median track being formed from the averaged tracks, and each of the original tracks being divided by the trend track calculated in this way.

9. Method according to one of claims 1 to 9,

wherein the baseline is calculated from a part of the EEG signal (E) recorded at the beginning of the EEG recording, in particular during the first 6 to 24 hours after the start of the EEG recording.

10. Device (1) for measuring the cerebral perfusion of a patient (8), with an EEG device (2) for recording an EEG signal (E) from the patient (8) and with an EEG device (2) Control unit (3) connected for data communication technology, the control unit (3) being set up to automatically carry out the method according to one of claims 1 to 9.