• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention describes a method for determining in real time the activation level of the trigeminal vascular system. In particular, said invention may be applicable in the field of medical devices capable of determining the activation rate of the trigeminal vascular system based mainly on the use of biomedical signals of a hemodynamic nature. The method establishes objectification criteria for this level and is presented as a result of the application of modeling and data fusion techniques. The method is also based on other types of signals, such as environmental ones, for statistical improvement in real time of the determined level. (Machine-translation by Google Translate, not legally binding)
    公开号:ES2634874A1
    申请号:ES201600158
    申请日:2016-02-29
    公开日:2017-09-29
    发明作者:Ana Beatriz GAGO VEIGA；Mónica SOBRADO SANZ；José Aurelio VIVANCOS MORA；Josué PAGÁN ÓRTIZ；María Irene DE ORBE IZQUIERDO；José Luis Ayala Rodrigo
    申请人:Universidad Complutense de Madrid；Fundacion para la Investigacion Biomedica del Hospital Universitario de la Princesa；
    IPC主号:
    专利说明:

    image 1
    DESCRIPTION
    Method to determine the level of activation of the trigeminal-vascular system. Technical sector
    The present invention fits mainly in the field of medical devices. In particular, it focuses on the modeling and prediction systems applied to neurological diseases and, more specifically, the prediction of the migraine crisis with the aim of becoming more effective in reducing pain with pharmacological treatments. State of the art
    Migraine is a type of headache that is characterized by pain with a series of characteristics defined by: its duration between 4 and 72 hours, a moderately severe intensity, worsening with exercise, pulsatile, which interrupts the patient's activity and is usually accompany symptoms such as nausea, vomiting, photophobia and / or sonophobia. Migraines constitute, in our day, a very important public health problem due to their high prevalence that carries an important burden for patients, family and society. Migraines are considered one of the most disabling pathologies, with important consequences on social and labor activities (absenteeism and loss of labor productivity) and, therefore, with a high socioeconomic cost (Linde M, Rastenyte D, Ruiz de la Torre E , et al. The cost of headache disorders in Europe: The Eurolight project. European journal of neurology: the official journal of the European Federation of Neurological Societies. 2012; 19 (5): 703-711) ...
    In the treatment of migraine pain it is known that an early treatment is much more effective, since once the central sensitization of the trigeminal occurs, within the pain cycle, curbing it is much more complicated. Thus, the fundamental objective in clinical practice is to treat pain before this central sensitization takes place. Consequently, the patient is recommended to take painkillers as soon as the pain begins in order to stop the migraine crisis, since if the medication is taken too late, it is usually not effective.
    Published works show that the dopamine antagonist, domperidone (Waelkens J Dopamine blockade with domperidone: Bridge between prophylactic and aborlive treatment of migraine A dose-finding stud. Cephalalgia. 1984; 4 (2) 85-90), and triptan, naratriptan (Luciani R, Carter D Mannix L. Hemphill M, Diamond M. Cady R. Prevention of migraine during prodrome with naratriptan, Cephalalgia. 2000; 20 (2): 122-126), drugs used in the acute phase, can reach Eliminate pain when administered early, in prodromes. The prediction based on prodromal symptoms by the patient may have a number of limitations, since the prodromal symptoms have a variable time horizon, that is, the patient does not know exactly when the pain will occur, the symptoms are very nonspecific (changes of character, appetite, sleep rhythm, etc.), can occur in any other situation and are very inaccurate (Becker W). The premonitory phase of migraine and migraine management. Cephalalgia 2013; 33 (13): 1117-1121: Rossi P. Ambrosini A. Buzzi MG. Prodromes and predictors of migraine attack Funct Neurol. 2005: 20 (4): 185191; Giffin NJ, Ruggiero L, Lipton RB, et al. Premonitory symptoms in migraine. An electronic diary study Neurology. 2003; 60 (6): 935-940). In this way, we could be less effective in reducing pain or decide to take a medication when a migraine will not occur.
    image2
    For this reason, knowing objectively when the patient is going to have pain before suffering (prediction), it will be possible not only to give early treatment but also more targeted treatment; knowing when the pain can occur and, known the mechanism of action of the drug, you can select the most convenient treatment for the patient, knowing when the pain will occur, you can observe the mechanism of action of the drug and state which treatment is more convenient. The method proposed in the present invention allows to predict the patient's pain and be more effective in the treatment.
    It is noted in the state of the art, the involvement of the autonomic nervous system (ANS) in migraine, demonstrating that there is a dysautonomic dysfunction, which is evidenced as an alteration of the variables controlled by it in the migraine patient and that This is involved in both the genesis and the persistence of migraine.
    There are still many unknowns about the nature of such dysautonomy of the migraine patient, in particular it is unknown if it is the cause or consequence of it, considering it simply as an epiphenomenon. Clinical studies evaluating autonomic function in migraine patients have shown discordant results (Gass JJ, Glaros AG. Autonomic dysregulation in headache patients. Appl Psychophysiol Biofeedback. 2013; 38 (4): 257-263; Mosek A, Novak V, Opfer-Gehrking TL, Swanson JW. Low PA. Autonomic dysfunction in migraineurs. Headache: The Journal of Head and Face Pain. 1999; 39 (2): 108-117; Sanya E, Brown C, Wilmowsky C, Neundorfer B. Hilz M. lmpairment of parasympathetic baroreflex responses in migraine patients, Acta Neurol Scand. 2005; 111 (2): 102-107 and Benjelloun H, Birouk N, Slaoui I, et al. Autonomic profile of patients with migraine. Neurophysiol Clin. 2005; 35 (4): 127-134.), Which describe a hypofunction and hyperfunction of both the sympathetic and parasympathetic systems.
    Only some studies have evaluated the changes between the baseline situation and the symptomatic period in the migraine patient, among which we highlight Duru M, Melek I, Seyfeli E, et al. QTc dispersion and p-wave dispersion during migraine attacks. Cephalalgia 2006; 26 (6). 672-677, which demonstrates the association in the migraine crisis with an increase in QTc interval of the electrocardiogram signal and P wave dispersion compared to asymptomatic periods, also highlighting the studies of
    Ordás CM, Square ML, Rodríguez-Cambrón AB, Casas-Limón J, del Prado N, Porta-Etessam J. lncrease in body temperature during migraine attacks. Pain Medicine 2013; 14 (8): 1260-1264 and Porta-Etessam J, Square ML, Rodriguez-Gomez O, Valencia C, Garcia-Placek S. Hypothermia during migraine attacks. Cephalalgia 2010; 30 (11: 14061407., Where they describe a case of a patient with hypothermia and another with hyperthermia during the pain period. In a study by Dr. Seçil (Seçil Y, Ünde C, Beckmann YY, Bozkaya YT, Ozerkan F. Başoğlu M. Blood pressure changes in migraine patients before, during and after migraine attacks. Pain practice. 2010; 10 (3): 222-227) a tendency to diastolic hypotension was observed. In relation to the state of the art related to the changes that occur at the electroencephalographic level include Bjork's works (Bjørk M, Sand T. Quantitative EEG power and asymmetry increase 36 hours before a migraine attack. Cephalalgia. 2008; 28 (9): 960-968) that show a slower activity and asymmetric prior to the onset of pain in relation also to the duration of the crisis and the intensity.
    However, none of these studies present the results for continuous ambulatory monitoring in real time during the previous period, the migraine phase and the subsequent period, nor do they apply pain prediction algorithms.
    image3
    While patent application WO03063684 (Geatz M. Roiger R.) proposes the use of a system and methodology for the prediction of symptoms through physiological variables with application to diseases of different kinds, one of them is migraine.
    However, the proposed methodology is of a high level of abstraction, not technically motivating the selection of variables to a certain pathology.
    Other patent applications also describe, but in more detail, closed loop control systems with application, in this case, to diabetes disease (ES2334733) and epilepsy (US2011270095). However, none of these proposes adaptive techniques against data loss to make a robust system.
    Application TW201023087 mentions the use of sensor networks for prediction, regardless of not specifying the industrial application, the claims are based on efficient consumption and data acquisition management through knowledge of the prediction status
    On the other hand, US8123683, which refers to headaches in general, aims to detect the type of headache produced; This is achieved through the grouping of different triggers that the patient can indicate to the system.
    However, this document only claims the classification of the crisis and not its prediction.
    In the industrial field, several mobile applications have been developed that serve as migraine calendar and classifiers (Migraine Buddy, My Migraine TriggersTM and My Migraines), however, none of these are used to predict crises in real time, and do not perform ambulatory monitoring of biometric variables.
    In conclusion, none of these documents propose:
    one. A method of predicting migraine crisis through non-invasive ambulatory monitoring of biometric variables and the recording of environmental variables.
    2. A detailed prediction methodology and a selection of the variables to be monitored, justified, based on the medical literature.
    3. A robust method against partial or total loss of data and saturation of sensors, by means of the System of Selection of Dependent Model of the Sensors (SDMS2).
    Four. The linear combination of migraine models through the selection of several trained models for each patient.
    5. A system that implements a hierarchical module of model selection for each patient depending on the availability of sensors.
    A system that implements an expert prediction improvement module for the elimination of spurious detections that could be considered as false positives.
    Preliminary results carried out by the authors of the present invention confirm the following technological advantages:
    image4
    one. Prediction of the neurological pathology of migraine through variables controlled by the autonomic nervous system in a non-intrusive way.
    2. Achievement of prediction time horizons within the period of action of migraine pain medications to make it effective.
    3. Definition of a robust system against data losses and falls in order to maintain a certain prediction horizon.
    Four. Creation of a prediction consequence of the linear combination of the result of several models in order to reduce the variability of the predictions and provide reliability to it.
    5. Use of an expert prediction improvement module capable of eliminating virtually all false alarm prediction. Detailed Explanation of the Invention
    The invention aims to predict migraine crises in real time through a robust and adaptive prediction method against data and / or sensor losses, which makes use of a non-invasive ambulatory monitoring of the patient's biometric variables and environmental variables . To this end, a hierarchical system of prediction models is defined, adapted to each patient according to the set of sensors available at each moment.
    The invention consists of a molasses for the prediction of migraines in real time.
    The method is structured in three phases (Acquisition of data, Training and validation of migratory models and Real-time Prediction) that are subdivided into different modules.
    In the stage of Acquisition of data, the biometric variables (d1), environmental (d5), the subjective sensation of pain of the patient during migraine (d2), information of the patient's activity that can affect migraine episodes (d3) are recorded ) and clinical data of the patient (d4). All this data is transmitted, from monitoring equipment to cloud servers. Such low-cost biometric variable monitoring equipment has limited duration batteries: however, they have sufficient processing capacity so that, being aware of the prediction status, they are able to decide when to temporarily stop ambulatory monitoring and thus reduce The consumption of your batteries.
    The Training and Validation stage of the models is carried out on servers, which are external equipment with great computing capabilities. The data (d1 and d2) are used, in the first instance, for the creation of personalized migraine models for each patient that will be used in real time in the third stage, for the prediction of migraine crises. In this last stage, the prediction in real time of the migraine crisis is carried out in a recursive loop based on the input data (d1, d3-d5). The prediction based on environmental data (d5), activity (d3) and clinical (d4), will be used in the module of the decision system (7).
    The following describes the steps to follow in each stage and the way in which each module intervenes.
    image5 A) Data acquisition
    The migraine crisis prediction method presented in the invention involves the use of the following multi-source data:
    • Biometric data of the patients (d1): hemodynamic signals and brain electrical activity. The biometric data of the patients are collected through a low cost wireless human body monitoring (WBSN) device. Said device is non-invasive and includes the following variables: by means of surface electrodes, the ECG electrocardiogram and the electrodermal activity or galvanic response of the skin (EDA or GSR, h1); the surface temperature of the skin (h2), through a thermistor; blood oxygen saturation (SpO2, h3) and photoplethysmography curve (PPG, h4) using an infrared sensor by reflection. The secondary variables used as inputs to the system are obtained in the signal preprocessing module (1). The heart rate (HR, hs1) is calculated from the ECG signal. Similarly, the pulse transit time (PTT, hs2) is obtained through the combination of ECG and PPG variables. Brain electrical activity or electroencephalography (EEG, h5) is measured at the occipital level. From it the power bands of brain waves are quantified (qEEG, hs3). The choice of these variables is made according to their alteration by the autonomic nervous system, mentioned above.
    • Climatic data of the geographical area in which the patient is located and local environmental data (d5): the climatological data of the geographical area are collected from a state meteorological service, and are: temperature (a1), relative humidity (a2), atmospheric pressure (a3) and rainfall (a4). Local measurements are made through a weather station always near the patient connected to a mobile phone, and include: room temperature (al1), relative humidity (al2), pressure (al3), brightness (al4) and level of sound pressure (al5).
    • Subjective sensation of pain of the patient in each episode of migraine (d2): the patient indicates in the application of data collection in smart mobile device the beginning and the end of the pain, as well as the prodromal symptoms and / or auras that he may suffer. The subjective evolution of pain is also recorded as relative increases and decreases in pain, drawing a curve that reflects the levels of pain intensity that the patient feels at different times of the migraine.
    • Information regarding the activity of the patient (d3) and that may be relevant for the study of migraine will play an important role in the method. The information is collected in an application that runs on a mobile smart phone, and refers to food intake (type: dairy, fruits, alcohol, etc.), physical or mental activity, mood, medication intake or sensations subjective such as prodromal symptoms or auras.
    • The patient's clinical information (d4) is also taken into account. Sex, age, related diseases or prescription medication may be relevant information in the prediction process. This sensitive information is anonymous and is collected at the beginning of the study. a) Signal preprocessing modules (1, 1 · 2 and 2)
    The data collected by wireless monitoring systems may suffer losses due to the cessation of wireless communication or due to sensor drops. To repair the signal losses in these time intervals, signal regeneration techniques are applied in the module (1) for automatic signal regeneration based on its statistical behavior. Before repair, the calculation of the secondary variables (hs1-3) is also performed in this module. In this module, in addition to the calculation of the secondary variables and the repair of lost data, the synchronization of the data is also carried out to adjust the sampling rates of the different signals. The result of the module (1) is the signal (b), which has two different types of signals, those of biometric origin (b [1]) and those of environmental origin (b [2]). Thus, b = b [1] ᴜ b (2].
    image6
    Module (1-2) is responsible for calculating derived variables (c) and knowledge extraction using Machina Learning techniques and runs in parallel to module (1). The derived variables are features or features of the primary (h1-5) and secondary (hs1-3) and global (a1-4) and local (al15) environmental variables. These traits or features chosen may be different for each patient, and the study of which ones are better is automated for each of them. These traits can be temporary variables (hd (t)), or concrete values (figures of merit, hD). The most common features that are sought in these signals are: signal energy, energy in some band of interest of the signal, moving average, maximum, minimum and average values, etc. Traits that vary with time hd = {hd1 (t), h2 (t), ..., hdH (t)}, will be used in the model training module (3). Note that, although no explicit reference is made, the derived signals are the signal (c), and include those of biometric origin (c [1]), and those of environmental origin (c [2]). Thus, c = c [1] ᴜ c [2]. Traits (hD) are calculated and use the module itself which, together with the patient's activity (d3) and clinical (d4) data, gives rise to the signal (g). The signal (g) is the result of the application of Fuzzy Logic or Fuzzy Logic techniques, which will provide classification criteria to help improve the prediction of the expert system (7).
    Environmental regeneration (d5) also applies signal regeneration and calculation of derived variables. Like the biometric variables, migraine prediction models will be trained with the repaired environmental data (b [2]) and with derived environmental data (c [2]).
    These modules are very important, because the goodness of the models obtained (d) depends on the quality of the data.
    To be able to calculate the models, the output signal is needed, which is the pain signal. The pain monitoring module (2) consists of reading the patient's subjective pain records. This module generates a synthetic pain curve by means of a bilateral Gaussian adjustment of the pain evolution points marked by the patient (d2). This evolution is recorded as integer values with no upper or lower limit, which represent relative changes with respect to the last moment marked. This non-limited scale, compared to others that are, allows us to generate a more faithful curve of the patient's pain evolution. Since he does not know a priori when he reaches or how much is his maximum pain, this method avoids having a curve with saturated values to the maximum of a supposed traditional limited scale. The module result
    (2) is a signal (a) relativized to its maximum value in order to normalize all the patient's migraine records. The symptomatic curve of synthesized pain (a) is described by the parameters {(μ1, σ1), (μ2, σ2)} that form the semi-Gaussians. The module (2), as well as the collection of pain data (d2), is only in non-real time, for the creation of the models.
    image7 B) Training and validation stage
    This stage runs offline, in non-real time. During the training stage, the patient is monitored for a period of time in which a sufficient number of migraines (T) is recorded to train the models. These modeling algorithms generate a prediction curve (ŷ) for each migraine from the multi-source data collected. During the training the result curve of the models is compared (signal given in Figure 1), with the one generated (a) from the subjective sensation of pain (d2). The metric used to evaluate the goodness of the prediction models generated is the fit or fit, defined as:
    Equation 1:
    image8
    where y is the calculated symptomatic curve (signal a) marked by the patient, and ŷ is the prediction. a) Training module (3)
    A model is trained for each of the registered migraines. Migraine models are created with module (3). To create the migraine models, the repaired biometric signals (b [1]), those derived with temporal variation (c [1]) and the synthetic pain curve (a) are used. In training, the search for the prediction horizon that best suits each patient will be carried out. The method evaluates different modeling techniques or algorithms; Unlike the patent proposed by Geatz M and Rolger R., which is only based on Artificial Neural Networks, our proposal and previous work has shown that different modeling techniques have to be evaluated for each patient , and choose the best of them all. The models are linear functions, or not, that have as input variables the primary, secondary or some of the calculated biometric signals (all of them included in the sets of variables b [1] and c [1]), and whose result is a prediction
    (d) of the evolution of the migraine crisis to a given horizon.
    Equation 2:
    ŷ (t) = f ([h1 (t), h2 (t), ..., hs1 (t), ..., hs3 (t), hd1 (t), ... hdH (t)])
    Equation 2 represents the predicted symptomatic curve ŷ (t) as a function of primary (h), secondary (hs) or derivative (hd) biometric variables that vary over time, all of them contained in the signal (b [1]) .
    Some valid modeling schemes or algorithms that are used successfully are: State-Space Equations, Artificial Neural Networks or Genetic Programming. An example is a generic equation (Equation 3) of a space system of states of order nx:
    Equation 3:
    x [k + 1] = Ax [k] + Bu [k] + w [k]
    v [k] = Cx [k] + Du [k] + v [k]
    image9
    Where y [k] is the output at time k, which depends on the current state, x [k], and a matrix,
    C. The system order, nx, is determined by the dimension of x [k]. The variable v [k] is the innovation or unexplained part that is added to the prediction, ideally incorrect white noise with the input variables, u (k). The output can also depend on the inputs at that moment, through of vector D, which weights the vector of input variables, u [k]. The following state of the method is defined by the equality of x [k + 1]. The state transition matrix (A) weighs the dependence between states, while the matrix of weights of the variables (B) sets the dependence with the input variables, u [k], w [k] is, ideally, white noise incorrect with the current state, and is the unexplained part or error that is committed in the passage from one state to another.
    This invention proposes to generate different models for different sets of input variables in order to subsequently develop a hierarchy of model selection according to the available sensors or variables. This hierarchy underlies the idea of creating a robust method, fault tolerant and partial sensor drops. At the end of the Training Stage there are Md different trained models and for each different combination of biometric input variables, with d = 1, 2, ..., N, N can be one or several migraines. The amount of Md models generated depends on the type of algorithm selected and its parameters.
    Another set of models (independent of biometric variables) is also trained with the use of environmental variables (sets b [2] and c [2]). Prediction models based on environmental variables are coarse-grained models, and have less time definition than predictions based on biometric variables; Therefore, these predictions will only be used in the decision module (7). b) Model validation and selection module (4 and 4-1)
    Module (4) searches for the best models to predict patient migraines using Cross Validation or Cross Validation techniques. Each Md model, (obtained from the Md set), with i = 1, ..., d, is validated by predicting the remaining unused migraines in the Nth combination that created that model. The validations are carried out at a prediction horizon that does not have to coincide with the one with which the model was trained, but rather the furthest one with the best prediction results.
    The validation involves checking how each of these models fit (according to the fit) to the prediction of other migraines. The result of this module (e) is a set of models that will be sorted and stored in the module (4-1), to give as a final result a batch or set of the Best Models (signal f) that best predict the migraines of the patient in question, if Md> 1. The selection criteria are: the prediction horizon, the robustness of the model against sensor drop or signal saturation and the complexity or order of the model. The selection of a set of models, instead of just one, gives greater stability to the method and mitigates the use of models with overfitting or overfitting.
    For each different combination of biometric variables (b [1] and c [1]) input there is, finally, a batch of models. The same process is followed for environmental variables (b [2] and c [2]). C) Real-time prediction stage. Expert system
    This stage is executed in real-time loop and the data can continue to be sent to the servers, or not. If they are sent, it will be possible to have a real-time control of the patient's condition, and in case of false negatives in the predictions, a record would be available to check how these failures occurred. With the batch of models of each patient, the prediction of migraine attacks is carried out using the primary, secondary and / or derived biometric variables. The steps in which the prediction is executed are detailed below.
    image10
    a) Sensor Dependent Model Selection System (SMDS2) (5)
    This module provides robustness to the method by adaptation, and not by redundancy or a high number of trained models. The migraines of each patient can be defined by a set of variables (v1, v1 = b ᴜ c) different from that of another patient; In addition, in a context of ambulatory monitoring, sensor failure due to unavailability or saturation is a recurring evil. In this scenario, a hierarchy or criterion of model selection is defined by patient and dependent on the necessary and available input variables. This adaptation is what makes the method robust.
    After the validation and selection of a patient's models, it is known which input variables best define their migraines. In this case, the chosen models are used and they have that combination of variables as input ({Mmejores, v1}). If one of the necessary sensors is not available due to failure or breakage, the next set of models with better fit and not dependent on said sensor ({Mmejores, v2}) is passed.
    This module, therefore, indicates the set of models of the module (4-1) to be used in (6), both for biometric variables and for environmental variables. b) Prediction and linear combination of models (6)
    Using the models obtained from (5) and with the biometric signals (b [1] and c [1]), the prediction of migraine will be performed. The prediction horizon is the best that can be obtained with each model. Each of the models in the lot (5) makes a prediction about the data of the necessary biometric variables (b [1] and c [1]). In total you have better predictions. The end result is a linear combination (p) of all these predictions. As prediction models are also trained through environmental variables, they are also used to predict, so the signal (p) is the union of the average predictions obtained with the biometric (p [1]) and environmental inputs (p [1]. Thus, p = p [1] ᴜ p [2]. c) Correction and adjustment of the prediction (7)
    In order to obtain the best predictions, this module makes a correction to the prediction (p [1]) through the biometric variables (b [1] and c [1]), eliminating spurious of said signal. Spurious are detected by defining level and time thresholds.
    This module also makes decisions about the prediction through another prediction (p [2]) made with the second set of variables (g), which includes: environmental variables (b [2]) and (c [2 ]), the information obtained with the techniques of Data Mining and Diffuse Logic about the features of the biometric and environmental variables, and the activity and clinical data of the patient These decisions are weights or weights to the prediction. The result is twofold, on the one hand a prediction curve to a given horizon, and on the other a warning signal of pain onset with probability of occurrence.
    The environmental variables generate longer-term predictions, and with little precision, but serve as a contrast to the prediction made with the biometric variables. These contrasts or decisions are carried out with Fuzzy Logic techniques previously trained for the patient in the module (1-2). This module presents at its output (i) the prediction (p) corrected, adjusted and contrasted. It is returned, apart from the prediction, a decision of at what time and with what probability the pain will occur.
    image11 d) Actuator (8)
    The actuator is the last module of the method, it is a man machine interface (8), and it feeds the patient the prediction made and corrected (i). In this way the patient can take the medication with enough time to take effect and have a total effectiveness before the pain appears. The prediction also reaches local biometric and environmental variable monitoring devices so that they can assess whether or not to stop monitoring for a known period of time. The interface will be the mobile application itself that collects information on the patient's activity, since the patient will have it with him at all times.
    The novelty of this invention lies in the following technical characteristics:
    • Personalized prediction of migraine attacks using the patient's hemodynamic variables and brain electrical activity. In addition, using weather and environmental variables, clinical data and information related to patient activity as support for the prediction.
    • Personalized training of models for each patient with automatic variable selection.
    • Creation of a hierarchical model set. Development of an automated method of selecting the hierarchical model dependent on the available input variables. Robust and fault tolerant method maintaining a certain prediction horizon.
    • Creation of a module to help predict and repair models that improves the goodness of the prediction.
    • Interface or warning system (actuator) for the effective intake of the medication that stops migraine pain before it appears. Brief description of the drawings
    The following drawings illustrate the preceding description:
    Figure 1.- It shows the scheme of creation of models in non-real time. This scheme is executed in the training phase and gives rise to the models (f) that will predict the migraines of each patient. The models are created in module (3) through the biometric variables (b [1] and c [1]) and the synthesized pain signal (a). Different models are created for each different combination of variables The models are validated in module (4), of these the best ones (f) will be chosen in hierarchical order in module (4). At the same time, module (3) also trains prediction models based on environmental variables (b [2] and c [2]), also for each different combination of these variables different models are created and the best ones are chosen.
    Figure 2.- Shows the modeling and prediction scheme in real time of the migraine prediction method through the patient's hemodynamic variables and brain electrical activity (d1), and with the support of local and global environmental variables (d5). Said variables are preprocessed and synchronized in the preprocessing module (1) In this prediction stage (in a loop) the appropriate models for the patient are chosen and depending on the variables available in the module (5) and the prediction is made (p [1]) through the linear combination of biometric variable models in the module (6). The prediction reaches the decision module (7) where spurious events are eliminated and, with the help of the prediction of environmental variables (p [2]) and the result of Machine Learning techniques (g), it is decided whether detected or not a migraine (i). The decision comes to the actuator (8), which alerts the patient to anticipate the intake of the medication and avoid pain.
    image12 Embodiment of the invention
    Monitoring involves recording a sufficient number of migraines for the training stage. In order to train the modeling algorithms, it is considered that from 10 migraine episodes the training may be sufficient (generally the monitoring time ranges between 4 and 6 weeks). The following describes a possible way to proceed with the acquisition of data, for information purposes only and not exhaustively.
    The monitoring of biometric variables (d1) is carried out with commercial ambulatory monitoring devices. The ECG sensor can have as many leads as desired, but with only three electrodes it is enough to extract the HR (hs1); in this case, they will be placed in the horizontal precordial plane at leads V3, V4 and V5. The EDA sensor (h1) placed on the arm serves to measure the relative variations of sweating; like the surface temperature sensor (h2), placed as close as possible to the armpit. One way to acquire SpO2 (h3) and PPG (h4) is by using an oximetry clamp placed on a finger. The EEG electrodes will be placed in the occipital, at the OZ, O1 and O2 reference points (according to the international 10-20 system). The PTT will be calculated through the ECG and PPG signals and applying some of the bibliographic methods (Yoon Y, Cho JungH, Yoon Gilwon, Non-constrained Blood Pressure Monitoring Using ECG and PPG for Personal Healthcare. 2009; 33 (4) : 261-266). The HR can be calculated in 20-second intervals with 10-second overlap. The qEEG is the energy of the bands Alpha, Beta, Gamma, Delta and Theta in non-overlapping intervals of 20 seconds.
    At the same time that biometric variables (d1) are recorded, local and global environmental variables (d5) are recorded. The global environmental variables are taken from the geographical area corresponding to the patient's location, and the national meteorology service can be used. Local environmental variables are monitored through a weather station always near the patient. The good synchronization of all the data must be taken into account, for this purpose a smart mobile phone that captures all the meteorological data can be used. Subjective pain sensation (d2) is recorded through a mobile application. The patient indicates the beginning and beginning of the pain, as well as the subjective evolution of the pain. With the same mobile application, the activity of the patient is recorded (d3) and in addition there is knowledge of some of their relevant clinical data (d4) for the study, such as weight, age, sex or diseases related to migrating it. All data (d1-d5) collected during the training stage are preprocessed (1 and 1-2) and used to train migraine models in (3).
    Both the training of the models, as well as the validation (4), are carried out in non-real time in equipment with high computing capacity. The result of the training stage is a different set of models for each possible combination of variables (f). The models are trained through supervised techniques, where the inputs are the processed variables, and the output to be adjusted is the subjective pain of the preprocessed patient in (2). Models will be trained for all possible combinations of input variables. In validation, the best ones will be ordered and chosen to be used in the real-time prediction stage.
    image13
    In the real-time prediction stage, the sensor-dependent model selection system (5) is used to select the variables of interest of each patient and the hierarchical set of models (f-2) depending on the available sensors. To do this you must know at all times the status of the sensors, and if one is not available, you will change models. Once you have chosen the models, they are applied one by one on the input variables, giving rise to a set of predictions; The final prediction (p) is calculated in (6) as the linear combination of these predictions. The horizon to which the prediction is made will depend on the quality of the model obtained, for example 30 minutes. The prediction correction and adjustment module (7) eliminates the spurious prediction according to criteria of duration and level of detection; In this way they can eliminate possible false alarms. Decision criteria (g) are also applied to weigh the response. The decision criteria weigh the prediction obtained from the biometric variables. These weights are the result of the Fuzzy Logic algorithms that, based on the knowledge of the environmental variables and the patient's activity, regulate the prediction, for example by attenuating or increasing its levels Finally, the prediction (i) is transmitted to the patient through the actuator module (8, the mobile device) so that he can advance the intake of the migraine pain medication, before it begins.
    The retraining of the models can be carried out automatically in a transition period in which, it is followed in the prediction stage in real time, but the set of patient models is updated. Retraining will be necessary when the patient finds that the predictions are no longer correct or a clinical evaluation deems appropriate. In real time patients will not mark the evolution of their pain (d2), so the only record of prediction failures that can be had will be false negatives (migraines that were not detected).
    The hardware of the monitoring devices must have sufficient computing capacity to be able to sample the variables at the required sampling rate and send the data wirelessly. The ability to stop monitoring when they are aware of the prediction status must also be supported by the hardware and firmware of the devices.
    The present invention has its application in the field of medical devices for early warning of migraine pain. The network of electronic ambulatory health monitoring devices validated for medical use is increasingly widespread and established: in addition, their portability and battery life are increasing. The machine-man interface for patient alerting is done through an intelligent mobile terminal, so common today. Therefore, the present invention can be used immediately in the monitoring of migraine patients for prediction of their crises.
    This invention allows migraineurs to make the advanced intake of migraine pain medication so that it has a total effect and thus be able to avoid the migraine pain phase. The use of this method increases the quality of life of patients, as well as reducing the direct and indirect costs caused by the disease worldwide.
    权利要求:
    Claims (8)
    [1]
    image 1
    1. Method for determining the level of activation of the trigeminal-vascular system based on the monitoring of biometric variables comprising the execution of the following phases:
    to. Monitoring of biometric and environmental variables and subjective level of activation of the trigeminal-vascular system for model training, structured in the following subphases:
    i. Preprocessing of the signals through statistical techniques based on the knowledge of the history of each signal, its following values and its distribution (mean and standard deviation, among others)
    ii. Objectification of the subjective measurement of the level of activation of the trigeminal-vascular system by means of normalization of levels and bilateral Gaussian adjustment with reference to the maximum level recorded.
    b. Expansion of the set of significant variables for training the models through the following subphases:
    i. Secondary Signal Generation
    ii. Generation and selection of traits of acquired signals and secondary signals.
    C. Estimation of the level of activation of the trigeminal-vascular system from the monitored variables, secondary variables and traits generated according to the following subphases:
    i. Generation of groups of variables, combinations of at least two of them.
    ii. Training of the models, one for each group of input variables and with reference to the level of activation objectified.
    iii. Choice of models according to the input variables used and the level of similarity of the signal they produce (ŷ) with the level of objectified (and) expressed in the following formula:
    iv. From the available variables and the chosen model, a first estimate of the level of activation of the trigeminal-vascular system is obtained as an output.
    d. Reduction of the error in estimating the level of activation of the trigeminal system by using a second set of variables, which may employ expert knowledge strategies, such as Data Mining and / or Diffuse Logic, for the correction and adjustment of the level of activation of the estimated vascular trigeminal system.
    image2
    14
    image3
    [2]
    2.  Method according to claim 1 for monitoring hemodynamic biometric variables, occipital electroencephalogram, ambient weather signals and local environmental signals to be sent to a cloud storage platform for processing.
    [3]
    3.  Method according to the preceding claims in which the data processing is structured in the following subphases:
    [4]
    Four.  Method according to the preceding claims characterized by the generation of secondary signals and characteristic features of the signals developed in the following subphases:
    to. Synchronization of the different signals with timestamps of all data.
    b. Elimination of out of range data and filtering of the signals.
    C. Application of automatic signal regeneration techniques based on the statistical behavior of the signal.
    d. Tithing of the signals to reduce the amount of input data for the models of Claim 1.
    and. Objectification of the level of activation of the trigeminal-vascular system by means of a non-limited level scale and of a mechanism of continuous Gaussian adjustment of discrete values subjective to the level of activation of the trigeminal-vascular system.
    to. Calculation of the heart rate (HR) by counting the number of events of the ECG signal in 20-second time windows with 10-second overlap by defining waiting times between peaks and level decision criteria for non false positive detection.
    b. Calculation of the transit time between pulses (PTT) for blood pressure determination using regression functions calculated with the detected peaks of the ECG and PPG signal.
    C. Calculation of the qEEG signal through the calculation of the filtering energy band pass without overlapping the EEG signal.
    [5]
    5.  Method according to the preceding claims, characterized by a Sensor Dependent Model Selection System (SDMS2) consisting of the preference of models for estimating the level of activation of the trigeminal-vascular system which can be performed by a mechanism based on statistical confidence .
    [6]
    6.  Method according to the preceding claims characterized by the linear combination of the set of models set forth in Claim 5.
    [7]
    7.  Method according to the preceding claims wherein the reduction of the estimation error in Claim 1 is carried out in three subphases:
    to. Detection and elimination of events by defining a threshold by which, those events that do not determine a level of activation of the trigeminal system with an index greater than 50% with respect to the maximum, will be eliminated.
    b. Detection and elimination of time-based events by defining a time threshold of 60 minutes, in which, events that exceed the level threshold but have a duration less than the time threshold will be eliminated: while the events that are found at a distance less than this threshold
    fifteen
    image4
    5 of another event, will be considered the same.
    C. Application of expert knowledge techniques, such as fuzzy logic algorithms, to give confidence levels to the events of activation of the trigeminal system to the extent of re-feeding the signal to the monitoring system.
    10
    [8]
    8. Method according to claim 1 wherein the monitoring of biometric variables obtained by sensors will develop the following subphases:
    to. The status detection may be carried out by means of a decision taken on the statistics of the data recorded in previous moments.
    b. If a sensor is not available, models that include dependent variables will not be chosen.
    Method according to claim 1 based on mobile devices that communicate the information to the monitoring devices.
    16
    类似技术:
    公开号 | 公开日 | 专利标题
    KR102219913B1|2021-02-24|Continuous stress measurement using built-in alarm fatigue reduction characteristics
    Sarkis et al.2015|Autonomic changes following generalized tonic clonic seizures: an analysis of adult and pediatric patients with epilepsy
    US10055549B2|2018-08-21|Method and apparatus for wireless health monitoring and emergent condition prediction
    Cupples et al.1993|Preexisting cardiovascular conditions and long-term prognosis after initial myocardial infarction: the Framingham Study
    US20170277858A1|2017-09-28|System for predicting risk of onset of cerebrovascular disease
    Koenig et al.2006|Clinical neurophysiologic monitoring and brain injury from cardiac arrest
    EP3380001A1|2018-10-03|Personalized health care wearable sensor system
    Calogiuri et al.2013|Methodological issues for studying the rest–activity cycle and sleep disturbances: a chronobiological approach using actigraphy data
    US11039986B2|2021-06-22|Chronotherapeutic dosing of medication and medication regimen adherence
    Kölling et al.2016|Comparing subjective with objective sleep parameters via multisensory actigraphy in German physical education students
    US20160342764A1|2016-11-24|System, computer-implemented method and computer program product for individualized multiple-disease quantitative risk assessment
    Zhang et al.2012|deStress: Mobile and remote stress monitoring, alleviation, and management platform
    JP2020014841A|2020-01-30|System and method involving predictive modeling of hot flashes
    US20180110462A1|2018-04-26|Device, system and method for detecting illness- and/or therapy-related fatigue of a person
    RU2704787C1|2019-10-30|System and method of determining for determining a stage of sleep of a subject
    ES2634874B2|2018-03-16|Method to determine the level of activation of the trigeminal vascular system
    Niel et al.2020|Actigraphy versus polysomnography to measure sleep in youth treated for craniopharyngioma
    Rana et al.2014|Novel Integrated Sensor based Sleep Apnea Monitoring and Tracking System using Soft Computing and Persuasive Technology for Healthcare Support
    KR102028676B1|2019-11-05|A method, server and program for providing medical after case service
    Regalia et al.2021|Sleep assessment by means of a wrist actigraphy-based algorithm: agreement with polysomnography in an ambulatory study on older adults
    Delaney et al.2021|The feasibility and reliability of actigraphy to monitor sleep in intensive care patients: an observational study
    Boateng2016|StressAware: App for continuously measuring and monitoring stress levels in real time on the amulet wearable device
    Sandmann et al.2016|Relevant biomarkers in the prediction of good and bad days for multiple sclerosis patients
    US20220022788A1|2022-01-27|Seasonal affective disorder determination
    Onorati et al.2021|Prospective Study of a Multimodal Convulsive Seizure Detection Wearable System on Pediatric and Adult Patients in the Epilepsy Monitoring Unit
    同族专利:
    公开号 | 公开日
    WO2017149174A1|2017-09-08|
    EP3425542A1|2019-01-09|
    EP3425542A4|2019-11-20|
    ES2634874B2|2018-03-16|
    US20190046123A1|2019-02-14|
    CN108780664A|2018-11-09|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US7223234B2|2004-07-10|2007-05-29|Monitrix, Inc.|Apparatus for determining association variables|
    CN101365373A|2005-06-21|2009-02-11|早期感知有限公司|Techniques for prediction and monitoring of clinical episodes|
    WO2008092133A2|2007-01-25|2008-07-31|Neurovista Corporation|Methods and systems for measuring a subject's susceptibility to a seizure|
    US8085145B2|2009-04-03|2011-12-27|Sharp Laboratories Of America, Inc.|Personal environmental monitoring method and system and portable monitor for use therein|
    US20100318424A1|2009-06-12|2010-12-16|L2La, Llc|System for Correlating Physiological and Environmental Conditions|
    US20150339363A1|2012-06-01|2015-11-26|Next Integrative Mind Life Sciences Holding Inc.|Method, system and interface to facilitate change of an emotional state of a user and concurrent users|
    US20150324544A1|2014-05-09|2015-11-12|The Regents Of The University Of Michigan|Pain surveying and visualization in a human bodily region|
    法律状态:
    2018-03-16| FG2A| Definitive protection|Ref document number: 2634874 Country of ref document: ES Kind code of ref document: B2 Effective date: 20180316 |
    优先权:
    申请号 | 申请日 | 专利标题
    ES201600158A|ES2634874B2|2016-02-29|2016-02-29|Method to determine the level of activation of the trigeminal vascular system|ES201600158A| ES2634874B2|2016-02-29|2016-02-29|Method to determine the level of activation of the trigeminal vascular system|
    US16/080,241| US20190046123A1|2016-02-29|2017-01-03|Method for determining the degree of activation of the trigeminovascular system|
    CN201780011929.6A| CN108780664A|2016-02-29|2017-01-03|The method for determining trigeminal vascular system activation degree|
    EP17759307.6A| EP3425542A4|2016-02-29|2017-01-03|Method for determining the degree of activation of the trigeminovascular system|
    PCT/ES2017/070004| WO2017149174A1|2016-02-29|2017-01-03|Method for determining the degree of activation of the trigeminovascular system|
    [返回顶部]