AEEG interpretation system and its methodology

Abstract

The invention relates to the use of artificial intelligence to evaluate amplitude-integrated electroencephalography (aEEG) data in terms of patterns, symmetry, presence of seizures, sleep / wake cycles and artefacts indicating basal brain function of patients during the recording period.

Description

[0001] AEEG INTERPRETATION SYSTEM AND ITS METHODOLOGY

[0002] The technical field related to the invention:

[0003] The invention relates to the use of artificial intelligence to evaluate amplitude- integrated electroencephalography (aEEG) data in terms of patterns, symmetry, presence of seizures, sleep / wake cycles and artefacts indicating basal brain function of patients during the recording period.

[0004] State of the art: aEEG devices are devices that process electrical signals arising from brain activity with the help of electrodes placed on the scalp, and ultimately measure the amplitude of these signals in a certain time period and display them graphicallyParameters such as specific patterns, sleep-wake cycles, symmetry, seizure activity and artefacts should be evaluated in brain function recordings made with aEEG. In current practice, these procedures are performed by trained-experienced clinicians (neonatologists) simultaneously with aEEG recordings or mostly retrospectively from recordings made a few hours / days before. aEEG device is a valuable neuromonitoring tool used in neonatal intensive care units that provides information about the baby's brain functions and allows long-term recordings (96 hours). The electrical signals obtained from the brain of the sick baby, as in EEG, are processed by the aEEG device and generate outputs that require training for evaluation. aEEG data are evaluated during the recording period in terms of patterns indicating basal brain function, symmetry, presence of seizures (convulsions), sleep / wake cycles and artefacts. Training and experience are essential for accurate evaluation of the outputs.

[0005] In the known state of the art, retrospective analyses cannot be performed in detail due to the long period of recording. Simultaneous evaluation is not possible for every patient during follow-up. The placement and fixation of EEG probes, impedance evaluation, recording quality, presence or absence of artefacts, presence of convulsions, presence / absence of symmetry, sleep-wake cycles, and changes in patterns cannot be interpreted correctly in evaluations performed by teams without sufficient experience. This may result in significant negative consequences for patients, such as the risk of under or over-treatment, late detection of conditions requiring urgent treatment, and the risk of misinformation about the patient's future prognosis.

[0006] The invention subject to the application numbered "CN112244871" in the known state of the art includes the interpretation of aEEG and the detection of seizures. The invention makes epilepsy-seizure and classification by analysing aEEG data. However, it does not provide information to the user about artefacts, sleep-wake cycle, or long-term prognosis by analysing both aEEG and visual data.

[0007] The invention subject to the application numbered "CN110811609B" in the known state of the art provides detection of epilepsy by processing EEG signals. The invention is a device for intelligent detection of epilepsy using adaptive template matching and machine learning algorithms. However, it does not provide information to the user about variables such as pattern classification, sleep-wake cycle at certain intervals, seizures, and artefacts by processing aEEG data.

[0008] In the known state of the art, the detection of epilepsy has been achieved through the processing of EEG signals. However, there is no system in place for processing the data in aEEG device screenshots and report documents, analysing spectrogram images for pattern classification, or providing information to the user about variables such as sleep-wake cycles, seizures, and artefacts at specific intervals.

[0009] As a result, a new technology is needed in the relevant technical field due to the above-mentioned negativities and the inadequacy of the existing solutions on the subject.

[0010] Brief Description and Objectives of the Invention

[0011] The invention relates to the use of artificial intelligence to evaluate amplitude integrated electroencephalography (aEEG) data in terms of patterns, symmetry, presence of seizures, sleep / wake cycles and artefacts indicating basal brain function of patients during the recording period. The most important aim of the invention is the simultaneous evaluation of aEEG data, reducing impedance distortion and artefact formation and reducing reading / interpretation errors by providing symmetrical recording from two brain halves (in dual-channel devices).

[0012] Another purpose of the invention is to enable remote monitoring.

[0013] Another aim of the invention is to provide the team with a recommendation regarding drug regulation in the presence of seizures.

[0014] Another aim of the invention is to provide the team with recommendations regarding drug regulation in the presence of seizures.

[0015] Another object of the invention is to provide a warning by reporting transitions between patterns (duration, pattern type) and presence / absence / occurrence time of sleep / wake cycles.

[0016] Another object of the invention is to provide the user with relevant data about longterm prognosis.

[0017] Another purpose of the invention is to enable the interpretation of aEEG document, and aEEG screenshot with mobile application integration.

[0018] Description of Figures:

[0019] FIGURE-1 : A drawing giving a schematic view of the system of the invention.

[0020] Detailed Description of the Invention

[0021] The invention relates to the use of artificial intelligence to evaluate amplitude- integrated electroencephalography (aEEG) data in terms of patterns, symmetry, presence of seizures, sleep / wake cycles and artefacts indicating basal brain function of patients during the recording period.

[0022] With the simultaneous evaluation of data from the aEEG device (110) and ensuring that the aEEG device (110) is properly connected to the patient, the invention reduces impedance issues and artefact formation. Symmetrical recording from both hemispheres (in dual-channel devices) minimises reading / interpretation errors. The system enables remote monitoring, provides real-time recommendations to the team for medication adjustments in the presence of seizures, issues real-time warnings for artefact formation or symmetry loss, reports transitions between patterns (duration, pattern type), and delivers simultaneous warnings. Additionally, it generates reports and warnings about the presence / absence or timing of sleep / wake cycles and provides the user with relevant data for long-term prognosis.

[0023] The invention collects instantaneous data from the aEEG device (110). These data include aEEG device screen data. The data in the report documents, such as the drug treatment applied to the patient via the application (130), are received by the server (120). The sampling time is determined by the expert, and all data are labelled by the expert according to the sampling time. These data are used by the server (120) as artificial intelligence training and test data. The server extracts spectrograms of the data obtained from the aEEG device (110).

[0024] The server (120) evaluates the instantaneous data and report document data received from the aEEG device (110) with deep learning and warns via the application by detecting artefact formation and symmetry loss. The artefact formation is made by analysing the video recording of the patient. If feeding, changing clothes or other care or intervention is being performed at that time, artefacts may occur.

[0025] By reporting the transitions between patterns (duration, pattern type), the server (120) alerts the application about the presence, absence or occurrence time of the sleep / wake cycle. There are sleep-wake brain wave values. Entries are made by evaluating these values and waves. REM and non-REM sleep can be distinguished. The server (120) ensures that when a seizure is detected, recommendations regarding medication regulation are given through the application.

[0026] At the stage of making recommendations regarding drug regulation, the latest neonatal convulsion treatment is valid in seizure treatment (the first option is phenobarbital 20-30 mg / kg load (it can be increased to a daily dose of 50 mg / kg), then 5 mg / kg can be maintained. If seizures could not be controlled, phenytoin loading can be added at 20-30 mg / kg and maintained at a dose of 5 mg / kg. If no response is obtained and it is resistant, midazolam 0.15 mg / kg IV bolus, levatiracetam 40 mg / kg IV bolus and three divided doses can be continued. Finally, lidocaine 2 mg / kg bolus 7 mg / kg / hour 4-hour infusion followed by a half dose (3.5 mg / kg in 12 hours and 1.75 mg / kg in the other 12 hours) can be maintained, and treatment can be continued with phenobarbital.

[0027] The server (120) provides the user with relevant information about the long-term prognosis via the application. The server (120) is remotely connected to the aEEG device (110) using communication protocols. As such, the aEEG device can be monitored remotely.

[0028] For prognosis or prediction, eEEG data should be reviewed by a specialist physician, the infant's clinic should be analysed, and data from diagnosed infants should be used. With the specialist's opinion, the baby's aEEG device (110) data are labelled for model training. This data set will be used in artificial intelligence training, and results will be generated for new samples. With continuous monitoring of the data, warnings will be generated when threshold values are exceeded. For prognosis determination, interpretation will be made with the data obtained from literature studies. The brain electrical activity will be monitored by following the infant whose aEEG data is taken, and thus the labelling will be verified. The diagnosis will be supported by the clinical evaluation of the baby after each recording. Possible artefacts will be prevented by shooting video with a camera during filming. In addition, the babies will be followed up for the first year and their brain electrical activities will be monitored.

[0029] The application (130) operates on any electronic device and enables the alerts sent by the server (120) to be provided to the user via an interface. The application (130) provides an interface where the user can enter the report details of the patient.

[0030] The aEEG interpretation method includes the following steps:

[0031] - Receiving instantaneous data, video recording and screenshots from the aEEG device (110) by the server (120) and training them with artificial intelligence, - Detection of artefact formation and symmetry loss by the server (120) by evaluating the instantaneous data and report document data received from the aEEG device (110) with deep learning and warning on the application,

[0032] - Reporting the transitions between patterns by the server (120), determining the presence, absence or time of occurrence of the sleep / wake cycle and presenting it to the user through the application,

[0033] - When a seizure is detected by the server (120), calculating the recommendation regarding the medication arrangement through the application and sending it to the user through the application,

[0034] - Long-term prognosis assessment by server (120),

[0035] In the process step of receiving instantaneous data and screenshots from the aEEG device (110) by the server (120) and training them with artificial intelligence;

[0036] - Artificial intelligence training of eEEG data obtained from diagnosed infants,

[0037] - Receiving by the server (120) the data in the report documents, including the drug treatment information applied to the patient via the application (130),

[0038] - Determination of the sampling time by the expert and labelling of all data by the expert according to the sampling time,

[0039] - Use of the data labelled according to the sampling time by the server (120) as artificial intelligence training and test data,

[0040] - Extraction of spectrograms by the server (120) of the data obtained by the aEEG device (110).

Claims

CLAIMS1. Amplitude-integrated electroencephalography data is an aEEG interpretation system that enables the evaluation of patients in terms of patterns, symmetry, presence of seizures, sleep / wake cycles and artefacts indicating basal brain functions with artificial intelligence during the recording period, characterised by the following: at least one aEEG device that measures brain electrical activity and includes a camera (110), at least one server that uses data labelled according to sampling time as artificial intelligence training and test data; extracts spectrograms of data obtained via aEEG, detects artefact formation and loss of symmetry by evaluating instantaneous data and report document data received from aEEG device (110) with deep learning, detects the presence, absence or time of occurrence of sleep / wake cycles by reporting transitions between patterns, calculates the recommendation regarding drug regulation when seizures are detected and performs long-term prognosis assessment (120), at least one application that works on an electronic device and presents the information calculated by the server (120) and warnings of artefact formation, loss of symmetry, presence, absence or time of occurrence of sleep / wake cycle to the user via the interface.

2. aEEG interpretation method, characterised by the following process steps:Receiving instantaneous data, video recording and screenshots from the aEEG device (110) by the server (120) and training them with artificial intelligence,Detection of artefact formation and symmetry loss by the server (120) by evaluating the instantaneous data and report document data received from the aEEG device (110) with deep learning and warning on the application,Reporting the transitions between patterns by the server (120), determining the presence, absence or time of occurrence of the sleep / wake cycle and presenting it to the user through the application,When a seizure is detected by the server (120), calculating the recommendation regarding the medication arrangement through the application and sending it to the user through the application,Long-term prognosis assessment by server (120)3. The aEEG interpretation method according to claim 2, characterised in that the process step of receiving instantaneous data, video recording and screen images from the aEEG device (110) by the server (120) and training them with artificial intelligence includes the following process steps:- Artificial intelligence training of eEEG data obtained from diagnosed infants,- Receiving by the server (120) the data in the report documents, including the drug treatment information applied to the patient via the application (130),- Determination of the sampling time by the expert and labelling of all data by the expert according to the sampling time,- Use of the data labelled according to the sampling time by the server (120) as artificial intelligence training and test data,- Extraction of spectrograms by the server (120) of the data obtained by the aEEG device (110)

Images