CN100446719C - Metabolic monitoring, a method and apparatus for indicating a health-related condition of a subject - Google Patents

Abstract

An apparatus for indicating a health-related condition of a subject has an input interface (20), a predictor (30), and an output interface (25), the input interface (20) for receiving a sequence of samples of a first biological quantity derived by a first measurement method, the first measurement method being an invasive measurement and having a first impact on the subject, and for receiving a sequence of samples of a second biological quantity derived by a second measurement method, the second measurement method being a non-invasive measurement and having a second impact on the subject, wherein the first biological quantity gives a more accurate indication of the health-related condition of the subject than the second biological quantity, wherein the first biological quantity and the second biological quantity have a correlation to the health-related condition of the subject, and wherein the second impact is smaller than the first impact, a predictor (30) for providing, for a certain time, for which no sample for the first biological quantity exists, an estimated value of the first biological quantity using samples for the first biological quantity and, as far as available, samples for the second quantity; and an output interface (25) for outputting the estimated value or data derived from the estimated value so that an indication for the health-related condition of the subject is obtained.

Description

Metabolism monitors, is used for the device of the healthy correlated condition of denoted object

Technical field

The present invention relates to by patient data is filtered, prediction and trend analysis improve the explanation to noisy physiology and biochemical signals, and a kind of method and apparatus and/or computer program are disclosed, its purpose is to improve the patient's who suffers from type 2 diabetes mellitus or diabetes relevant disease enthusiasm, oneself's control and self management.The present invention monitors the coefficient of oxygen utilization of heart, thus monitoring physiological condition and health, and the abuse of indication analeptic and drugs, and psychology and emotion pressure.The invention discloses a kind of painless, non-invasive blood glucose and substitute measuring method, and carry out the blood glucose prediction, and disclose a kind of metabolism performance indications by sparse blood sampling.The invention provides a kind of low cost and easy-to-use long-term metabolism method for monitoring, and make the patient know the metabolic system function relevant, and the user only need pay very little effort with disease in mode intuitively.The application of the invention can reduce the cost and the burden of health care system, prolong the patient life-span, improve its quality of life.

Background technology

Physiology and biochemical signals, for example blood glucose sampling, blood pressure and other mammalian signal that can monitor are very noisy, As time goes on sampling can have very high difference.Therefore before can explaining accurately data, it is very crucial reducing this noise.And biochemical signals is invasive usually in essence, therefore this measurement can make us feeling under the weather, with high costs or be difficult to carry out.The present invention attempts by the accuracy of use suitable filtering method raising to the explanation of sort signal, and by using non-invasive alternative measuring method to reduce immoderation and cost.

Diabetes worldwide increase rapidly, cause huge cost burden to health care.Type 1 diabetes accounts for 10% of whole diabetes cases.Therefore, type 2 diabetes mellitus accounts for the about 90% of whole diabetes cases, and in steady increase.Only in the U.S., estimate at nearly 7% population and may suffer from diabetes.Have 100,000,000 body weight for humans overweight, the risk of therefore suffering from type 2 diabetes mellitus is higher.If this trend continues, so in the year two thousand thirty, 100% U.S. adult will be overweight.The year consumption of U.S.'s diabetics comprises indirect consumption, is estimated as about 1,000 hundred million dollars in 1997.In Saudi Arabia, have according to estimates up to 25% population and may suffer from the diabetes relevant disease.The World Health Organization's prediction, by 2025, the number of whole world diabetics will increase to 300,000,000.People have carried out various trials, attempt to reverse this global fashion trend, but up to now, have all failed.

Type 1 diabetes (being called insulin dependent diabetes mellitus (IDDM) IDDM in the early time) is that it causes insulin definitely to lack usually because the irreversible damage of β cell causes.It is a kind of non-simple disorder that type 2 diabetes mellitus (being called non-insulin-dependent diabetes mellitus NIDDM before the drought) is confirmed to be, and it is believed that it is simultaneously relevant with the h and E factor.Type 2 diabetes mellitus mainly is a kind of and the life style diseases associated, and wherein modern sedentary life style and bad dietary habit are considered to the main root of this problem.The type 2 diabetes mellitus patient does not need insulinize to earn a bare living usually.The classical symptom of type 2 diabetes mellitus is: very thirsty, polyuria, tired, become thin, overweight, gustatory hyperhidrosis, the various dimness of vision, blood sugar increasing, breathing have acetone flavor and glycosuria.The patient investigated to find very typical sitting life style and for the special preference of high saturated fat and refining carbon hydrate recipe.

Insulin resistant is a kind of common Developmental and Metabolic Disorder, and it is to suffer from various physiologic derangements, comprises type 2 diabetes mellitus and obesity, the feature of individuality, and with many cardiovascular and Developmental and Metabolic Disorder and deposit.Insulin resistant is defined as, and health can not be made sufficient response to insulin.X syndrome or metabolic syndrome are also referred to as insulin resistance syndrome, are the metabolism of development of prediction type 2 diabetes mellitus and related cardiovascular disease and the summation of physiological risks factor.It is a feature with five kinds of main abnormal usually: obesity, hypertension, insulin resistant, impaired glucose tolerance and high triglyceride disease (dyslipedaemia).The prevalence rate of western countries' metabolic syndrome is 25-35%.Old and feeble relevant with the reduction of β cell function with insulin resistant, and fat relevant with insulin resistant and hyperinsulinemia.

Diabetic autonomic neuropathy (DAN) is a kind of serious and be modal a kind of in the diabetic syndrome.Most of type 2 diabetes mellitus patients were accompanied by the degeneration of autonomic nervous system (ANS) function before dying from cardiovascular disease.This seldom comes into the picture in early days, makes type 2 diabetes mellitus become a kind of " secret " disease that can slowly develop for many years, and have only usually when being found late that the patient just can notice it.DAN can weaken the ability of carrying out normal daily routines, reduces quality of life, increases dead danger.The systemic many internal organs of DAN system, for example gastronintestinal system, the genito-urinary system cardiovascular system of unifying.The result that the DAN nerve fiber relevant with hyperglycemia level " toxicity " effect destroyed and lost.Therefore in the process that prevents DAN generation and development, it is crucial strengthening glycemic control.By carrying out heart rate difference (HRV) analysis, can successfully detect ANS problem and DAN.

Hypertension is a main health problem among the crowd of west, and relevant with cardiovascular disease.Arteriosclerosis may be hypertensive reason, is again its result, however the prompting of research in recent years, and arteriosclerosis is hypertensive typical omen, and arteriosclerosis has hereditary basis probably.Most of type 2 diabetes mellitus patient (surpassing 50%) suffers from hypertension.Therefore, control of diabetes patient's blood pressure is imperative.In the type 2 diabetes mellitus patient, recommend blood pressure to be remained below 130/80 by the mode of making the life better or Drug therapy or the two combination.

Insulin resistant and type 2 diabetes mellitus are attended by the change of blood plasma lipoprotein level.Type 2 diabetes mellitus patient blood fat disorder up to 70%.Coronary heart disease is type 2 diabetes mellitus patient's underlying cause of death.Dyslipidemia and obesity, hypertension and hyperglycemia are the strong inducements of coronary heart disease.Even slight dyslipidemia also can improve the risk factor of coronary heart disease.Because these risk factors be addition or or even the multiplication, so the strategy of the mode of making the life better should only not be conceived to hyperglycemia, should also be noted that dyslipidemia.Because type 2 diabetes mellitus patient's dyslipidemia is shown as less and more thick LDL granule usually, its easier arteriosclerosis that causes is so the target of cholesterol reducing should comprise very low density lipoprotein (VLDL) (VLDL) and low density lipoprotein, LDL (LDL) and reduce the triglyceride (TG) that has increased.

Psychentonia, hypertension and high heart rate are the common problems of current society.In the work and life style in modern times, sports are less, and the high-tech related work often causes sedentary lifestyle.It is very general to continue high nervous high request work, and is a negative interaction factor that causes the stress disease.As everyone knows, psychentonia can influence metabolism, the blood sugar level that for example raises and increase systolic pressure and heart rate.Various analeptic, for example caffeine, nicotine, ethanol, cocaine and amphetamine also can increase systolic pressure and heart rate.

Modern recipe type, high-energy and fat content, relevant with insulin resistant and associated disorders.Yet the definite cause of disease of insulin resistant it be unclear that.Genetic predisposition and environmental factors comprise quality and quantity fatty in the diet, all can cause can not fully handling blood glucose under normal plasma insulin level.Because high sugar, the time efficiency height that is rich in fat, delicious taste and has a meal, fast food consumption constantly increases.Fast, the increase of high-energy carbohydrate consumption can present, blood glucose is excessive and insulin is excessive, hypoglycemia and sleepiness is arranged afterwards is so require to obtain again the fast food carbohydrate once more.This circulation feedback usually can intense impact metabolism regulating system.It is believed that it is deleterious that this instantaneous stimulation is gone down for a long time, and cause insulin resistant and insulin level to increase, this is beginning in early days of type 2 diabetes mellitus.Now, the above-mentioned problem relevant with life style causes over the huge health problem that never hears about.

Sports and aerobatic exercise are the foundation stones of antagonism type 2 diabetes mellitus relevant disease.Most important task is to increase heart by sports to provide the ability and the efficient of oxygen to cardiovascular system, and improves insulin sensitivity and muscle obtains the ability of oxygen, thereby improves cardiovascular health.Cardiac function is the same with other muscle, can become stronger, more effective by exercise.Body weight reduces by 10% can show positive effect to blood glucose and blood lipid level usually.Particularly importantly reduce stomach fat.

Sports and energy expenditure can be assessed in various manners, and can not limit patient's normal daily routines.Multiple diverse ways is arranged, for example pedometer, accelerometer, cardiotachometer etc.A kind of popular method is to use pedometer to calculate the walking step number or calculates approximate calorie that is consumed by simple formula.The method of the energy expenditure that other calculating is relevant with body kinematics and acceleration adopts single shaft, twin shaft or three axis accelerometer.Another kind method adopts the pulse based on plethysmograph (thereby a kind of equipment of launching light by finger or ear-lobe calculating heart rate and sports) to monitor.Another kind of general equipment---pulsometer---uses the pectoral girdle with electrode to custom-designed wrist-watch computer emission EKG pulse; and measure this EKG signal; thereby can calculate the calorie and other parameter relevant that are consumed with sports; yet; the simplest method of calculating sports is the intensity daily routines of only estimating roughly with routine work of being done and performed sports relevant with the persistent period, for example is divided into the grade of 1-5.More accurate calculating and method for reporting comprise MET form (metabolic equivalent) or formula, and it is the index of precision of sports intensity.Modern times lack movable and sedentary lifestyle is hygienic gymnastics and various heart Related product and is used to improve healthy sports project and has opened huge market.Although these positive trend is arranged, still increase fast with surprising rapidity with the type 2 diabetes mellitus diseases associated.

It is very difficult that the people who mobilizes high-risk, overweight, sitting and easily suffer from diabetes changes biochemical mode.Only inform that the individual faces health risk, it usually is not enough needing sports and/or needs to correct dietary habit and/or reduce pressure.The people of subhealth state checks or can feel under the weather usually when gymnasium is carried out sports being forced by other people.Find that through regular meeting overweight people feels uneasy to their low health level, perhaps for fear of shame, refusal is participated in healthy restoration project.The inventor thinks, the unique method of breaking this dogmatism tendency is, use simply a kind of and intuitively instrument monitor they self metabolic function, preferably carry out at home privately, thereby educate people by experience at hand (hand-on experience).Like this, the individual talent will appreciate that the problem that is run into, and knows the change that should make which content and intensity to life style.

Use individual blood glucose meter to carry out the oneself and monitor that normally 1 type insulin dependent diabetes mellitus (IDDM) (IDDM) patient is necessary, so that help the self-injection of insulin.Yet, lessly utilize blood glucose to monitor to the people who suffers from already or be in the type 2 diabetes mellitus edge indication is provided.Because people's kidney changes of threshold scope is very wide, so it is more or less out-of-date to use the urine meter that monitors glucose in urine to carry out oneself's supervision at present, and less employing.In addition, this method is the glucose level under the energy measurement kidney threshold value not, and time delay is long, and sensitivity is low, so blood glucose monitors it is preferred.

The advantage that the BG after using blood glucose meter to type 2 diabetes mellitus patient feed monitors has been reported in nearest research.This thought is, thereby supervision obtains the knowledge of food for patient's metabolic effect with the GLPP level before the meal.Then, the patient is rule of thumb known the degree that the GLPP level increases, and understands the feedback that the dissimilar food of picked-up changes for glucose.The purpose of this thought is the balance food intake, and the amount that promptly reduces exquisite, quick carbohydrate can alleviate post-prandial glycemia and exceed standard.People understand, and this exceeding standard can be caused long-term damage to autonomic nervous system, finally can cause diabetes and diabetic neuropathy.The oneself of this form monitors relatively more clumsy and can not keep, so the patient lacks power thereby withdraw from this test unrare owing to the intensity to this method.Blood glucose meter and during daily, need to carry, and test result is disclosed in masses sometimes when people have a dinner in the dining room.If in the secular everyday practice of patient, comprise the program of this clumsiness, then unlikely meeting success.In addition, since the consumption of a large amount of blood sugar test paper bars and the cost repeatedly pointing puncture every day and produced also can not ignore.In addition, very small although this test is invasive in essence, still can agonize with very uncomfortable for the patient.Further, it is very few for logic and the visual interpretation that the result provides, so patient's indigestion and management be so that realize therapeutic purposes, and this is an important disadvantages.

The World Health Organization (WHO) and ADA (ADA) have stipulated range of blood sugar and level, so that distinguish the diabetes of different phase.Diagnosis has the fasting glucose concentration (WHO standard, 1999) of patient with sympotoms as follows.Sample concentration of glucose unit is mmol/L on an empty stomach:

Unfortunately, when assessing blood sugar level clinically, people can ignore intensive biology of diversity and the existence of diagnosis diversity very at large.Therefore there is significant difference between the observed result, may be misread, thereby cause disease classification and diagnostic accuracy to reduce by unfamiliar doctor.When blood sample collection clinically, many factors can influence the precision of measurement result, for example:

1. the optimization of owing of clinical diagnostic device is calibrated.See the example of Fig. 1.

2. blood sample is owing to glycolysis wears out, because the glucose antiseptic can not be avoided glycolysis fully.

3. " white clothing hyperglycemia " (White-coat Hyperglycemia) is because " needle tubing phobia " patient's anxiety causes the BG level to increase.

4. the empty stomach BG value that continue to descend, its with daytime time lengthening relevant.

5. time dependent insulin sensitivity causes sensitivity difference between every day.

6. the women is because the periodicity hormone variation that menstrual onset causes.

7.BG can be owing to temporary transient actute infection, trauma stress even simply flu or influenza change.

Because top uncertain factor, the inventor believes, uses enough accurate blood glucose meter to be aided with suitable post processing and filter method, carries out blood glucose at home and monitor the precision that can improve diagnostic classification under controlled condition.The inventor believes, this with carry out clinical laboratory and measure and compare better with current practice.

Although before the final apparition of blood sugar level for a long time in, will occur that insulin level increases (hyperinsulinemia) in the blood flow; But the hyperglycemia level remains classical type 2 diabetes mellitus symptom grader.Except the clinical research purpose, insulin level seldom, perhaps never can, as risk of diabetes index or diagnostic tool, this be one significant true.Therefore, blood sugar level is low can not get rid of the probability that disease exists.

Monitor that oxygen saturation is a kind of common operation that patient in emergency treatment and the operative treatment is carried out.Before present widely used pulse anoxia-photometer (a kind of instrument that utilizes infrared Absorption to monitor hemoglobin saturation with oxygen) invention, people calculate the systolic pressure heart rate product (Rate-Pressure-Product) of patient in the operation usually (RPP) to determine cardiac conditions of patients and coefficient of oxygen utilization.RPP (be also referred to as two products (Double Product)) is a kind of quite accurate measuring method to the heart coefficient of oxygen utilization, draws (RPP=sBPxHR/100) by systolic pressure and heart rate are multiplied each other.After introducing the pulse anoxia-photometer, the use of RPP is now seldom still used in sports medical science sometimes, the coefficient of oxygen utilization of its indication heart during treadmill is tempered test etc.RPP can also indicate nervous and anti-depressant use.

In order to alleviate patient's burden, the inventor announces to have only the fasting glucose sampling to be only the type 2 diabetes mellitus relevant disease is carried out accurate monitored for prolonged periods of time and treats necessary.According to one embodiment of present invention, even sparse blood glucose sampled measurements for example weekly, also enough is used to the BG in the daytime that calculates to a nicety.Stronger and clumsy blood glucose monitors that for example carry out blood glucose measurement before the meal and after the meal every day, it is essential not think, because fasting blood glucose level is generally indicated the relative size of post-prandial glycemia drift.Therefore, the higher level of postprandial blood sugar that reflects of fasting blood glucose level is higher, and vice versa.This can be by being confirmed down, promptly during intervening the improvement life style respectively at 0h, Fig. 3 is seen in the oral carbohydrate tolerance OGTT test of 1h and 2h sampling carrying out repeatedly 3 samples.As seen, the life style improvement is accompanied by BG reduction on an empty stomach, and the BG value is on a declining curve after the meal simultaneously.Yet, thinking when being necessary, also can adopt 1h after the meal the BG measurement result replace BG on an empty stomach.But this is clumsy more, and therefore little actual, reason is as indicated above.

In another embodiment of the present invention, preferably only use blood pressure and heart rate (systolic pressure heart rate product) prediction BG level, make except initial calibration and set up the predictive value program, no longer need painful finger puncture or painful invasive program.In another embodiment of the present invention, the frequency that needs of painful finger puncture is reduced.

The present invention provides a kind of method of measuring intuitively and analyzing special physiological parameters, for example sports intensity, blood glucose, blood pressure and heart rate for the patient.In addition, can preserve and handle important patient data, for example blood lipid level, T-CHOL, triglyceride, body temperature, body weight, body-mass index and waist-to-hipratio.After these are measured, utilize suitable filter algorithm data are handled and to be optimized, afterwards the immediate feedback of indicating its behavior, process and result to the patient in mode intuitively.

The preferred embodiments of the present invention comprise the steps:

Preferably estimate every day or the active level of measurement physical culture, and preferably this information is collected in the data base.

Higher or lower frequency ground, intensive or sparsely sampled measurements and preferably this information is collected in the data base on an empty stomach and/or level of postprandial blood sugar.

Frequently, intensive sampling is measured systolic pressure and heart rate, and preferably this information is collected in the data base.

Calculate systolic pressure heart rate product by heart contraction blood pressure and heart rate.

Measure any other related physiological parameters, for example body weight, body temperature, blood fat etc., and preferably this information is collected in the data base.

Using statistics and/or signal processing method that above-mentioned data are carried out low-pass filter, enhancing, error correction and missing data inserts.

Preferably adopt Forecasting Methodology by systolic pressure heart rate product prediction blood sugar level.

By suitable algorithm the gained data are made up and/or filter, so that reduce noise, explaination and raising gained information.

Provide this process processing, enhancing and/or data predicted as a kind of trend with a kind of mode of understanding directly perceived and easy to the patient, thereby make the patient understand these parameters easily.

By on can be obvious, the metabolism of diabetes relevant disease monitors for estimating that the present at least state of object is very important.Important biological parameter is carried out intensive sampling have many advantages.Main advantage is that object can be known the current state of oneself continuously, thereby its health status can not worsened.Another advantage is, object obtains the variation of its current state or the overview of trend continuously, and the sports that for example are in unfavorable conditions lack or lack good nutrition, the sports that perhaps are in good condition fully and diet control good.Another advantage is that object can obtain the feedback of its state immediately, and can regulate its life style according to the trend of benign development.The prerequisite that carries out effective metabolism supervision according to the present invention is that object monitors important biological parameter.For example wake up morning measurement of glucose levels, blood pressure and heart rate are measured sports etc. by day.

Accurate blood glucose monitors and requires intrusive mood to measure, although the finger puncture can think that invasive is very small.At present, there is not the accuracy of other method to compare with the intrusive mood measurement.The a spot of blood sample of object puncture finger collection is measured in analytical equipment subsequently, this equipment output blood glucose value.Even it also is the cost costliness that the most small intrusive mood is measured, and feels under the weather through regular meeting, therefore can cause negative effect to patient and disease control.

Summary of the invention

An object of the present invention is to provide a kind of improvement conception that is used for the healthy correlated condition of denoted object, it is compared with traditional method and is more readily understood, and executory cost is lower, and is more comfortable and bigger to the promotion power of object.

For realizing goal of the invention of the present invention, a kind of device that is used for the healthy correlated condition of denoted object is provided, comprise: inputting interface (20), it is used to receive the primary sample sequence of first Biomass that obtains by first measuring method, be used to receive the primary sample sequence of second Biomass that obtains by second measuring method, this first measuring method is that intrusive mood is measured, object is had first to be impacted, this second measuring method is that non-intrusion type is measured, object is had second impact, each Biomass has useful difference and useless difference; Wherein first Biomass provides the indication than the healthy correlated condition of the accurate more object of second Biomass, and wherein the healthy correlated condition of first Biomass and second Biomass and object has dependency, and wherein second impact is impacted less than first; Predictor (30), it is used to utilize the sample of second Biomass and the sample of first Biomass that obtains as far as possible, and in certain time that does not have first Biomass, the estimated value that first Biomass is provided is as the prediction sample; Filter (22), it is used to filter the sequence with the first Biomass sample and at least one prediction sample, sequence after the filtration with filter before compare useless difference with useful difference and minimizing, and output interface (25), it is used for output at least and increases indication, reduces the indication or the indication that remains unchanged, as data trend, the useful difference of the healthy correlated condition of this trend representative object.

The present invention is devoted to not accommodate cost by what introduce that a kind of new alternative measurement and prediction reduce the user.

The present invention is according to following discovery, and promptly high-precision intrusive mood measuring method can partly replace with a kind of alternate non-intrusion measurement method.Aspect the influence of object, high-precision intrusive mood measuring method often is a kind of with high costs, uncomfortable and measuring method of " difficulty ", and the non-intrusion type measuring method then is a kind of low cost, the comfortable and measuring method of " tenderness ".

According to the intrusive mood data of sparse sampling and the non-intrusion type data of intensive sampling, predictor (predictor) can produce the intrusive mood data of intensive sampling.Therefore, object needn't bear painful finger puncture every day or needn't often carry out as prior art, but can reduce the frequency of finger puncture, and is for example weekly.Object only needs for example to carry out every day once non-invasion blood pressure measurement of correlation simple, no pain, so this influence to the patient is little.

In a further advantageous embodiment, predictor can feed back amount biology that is obtained by non-intrusion measurement more than.

According to the present invention, two kinds measure or biology amount unique prerequisite be that two kinds of measurements all have dependency with the healthy correlated condition of object.

Further, the present invention is devoted to by using the low-pass filter method to extract useful signal difference and removing useless signal difference to improve the explanation precision to noisy physiological signal.

Description of drawings

Now, the present invention will be described with reference to the accompanying drawings for the general, and it just illustrates, do not limit the scope of the invention or spirit, wherein:

Fig. 1 has shown the empty stomach BG measurement result from two different occasions (occasion) and clinic.Be each occasion relatively its laboratory measured value and average ( bar 1,2 and 4 is considered to correct) with three measurement results of three high-quality BG instrument of same brand.

Fig. 2 has shown the influence of three different test occasions " needle tubing phobia ", and wherein the BG value significantly raises when the nurse uses needle tubing.Measurement result is the meansigma methods of three high-quality BG instrument of same brand.

Fig. 3 has shown 3 OGTT of 3 different occasions.Only just can illustrate well that with 3 samples BG is dynamic.

Original fasting glucose measurement result and the trend (low pass filtered signal) that provides by case study (point) has been provided Fig. 4.Provided the WHO limit simultaneously.

Fig. 5 has shown the limit according to WHO, when the patient is diagnosed, has intensive uncertainty in typical clinical BG measurement result, because patient's diagnosis depends on the occasion when testing very much.

Fig. 6 has shown the estimated value of the auto-correlation function (acf) of original empty stomach BG measurement result (by case study).Acf clearly points out, signal life period dependency.

Fig. 7 has shown the distribution histogram of original BG measurement result, shows that they are approximated to normal distribution.

Fig. 8 has shown BG instrument 1﹠amp in the case study; 2,1﹠amp; 3 and 2﹠amp; Measurement difference between 3.

Fig. 9 has shown that all c phases of original BG measurement result (by case study) scheme.Most of energy are in the low-frequency band.Therefore, higher frequency contains seldom or does not have Useful Information, therefore can cast out.

Figure 10 has shown the frequency response of low pass filter.Notice that this cut-off frequency is a typical example.

Figure 11 shows, filters empty stomach BG sample with the cut-off frequency between the 0-1 and can produce residual error, perhaps produces difference between primary sample and filtered sample.For each cut-off frequency, the meansigma methods of residual error square produces the curve among Figure 11.This curve has the intersection breakpoint, and shown in two cross linears, suitable cut-off frequency has been selected in expression.

Figure 12 has shown the periodogram of the empty stomach BG sample of handling by low pass filter.

Figure 13 has shown original heart contraction blood pressure sample and the trend by forming with top identical low-pass filter method.

Figure 14 has shown the trend of empty stomach BG and sports, shows to have a kind of dependency.

Figure 15 has shown that the rectangle by 100 samples moves the BG of window generation and the dependency between the RPP trend (dotted line).Relevant significance (solid line, 1-P) will be understood that＞0.95 for significantly.

Figure 16 has shown by the BG of the rectangle moving window generation of 100 samples and the dependency between the RPP trend derivative (dotted line).Relevant significance (solid line, 1-P) will be understood that＞0.95 for significantly.

Figure 17 has shown the system that is verified of the enough flight data recorder method representations of energy.

Figure 18 has shown BG prediction result behind the low-pass filter.In this example, prediction filter upgraded once in per seven days.

Figure 19 has shown the trend of metabolism performance exponential sum sports, shows to have dependency.

Figure 20 has shown the screen message dump of computer program page 1.

Figure 21 is the block diagram of filter/trend instrument;

Figure 22 a is the block diagram of second processor of equipment shown in Figure 21;

Figure 22 b is the block diagram of the first processor of equipment shown in Figure 21; With

Figure 23 is the block diagram of one embodiment of the invention.

The specific embodiment

Figure 21 has shown the block diagram that is used for diagram filter/trend instrument, just is used for the block diagram of the device of the healthy correlated condition of denoted object.This device comprises inputting interface 20, is used to receive and the healthy relevant Biomass primary sample sequence of object, and wherein Biomass has useful difference and useless difference (arrow 21 of Figure 21).

According to concrete execution, inputting interface by manual input for example by keyboard, obtain these Biomass samples from analytical equipment by cable, radio, infrared or other method, form original series, thereby wherein analytical equipment is analyzed for example electronic buffer, memorizer or similar device output blood glucose value, blood pressure, heart rate, sports or other the interested Biomass of blood sample in inputting interface 20.Therefore, can obtain original series at the outfan of inputting interface 20, as sample sequence, it can be input to the filter 22 by arrow 23 indications of connection box 20 and frame 22.

Selectively, perhaps additionally, original series can also be input in the first processor 24, handles sequence thereby be used for that original series is handled acquisition, and it is filtered in filter plant 22 after being handled by processor 24.

First processor 24 can comprise predictor, inserter or any other device, and it is configured to utilize the original series by frame 20 outputs to draw the processing sequence.In this connects, first processor can also comprise with general introduction in the back and close device, be used to merge two or more original series, thereby obtain the original series that merges, it is filtered with filter 22 then.

The processing sample sequence that filter 22 is used to filter the primary sample sequence or derives from the primary sample sequence, thus filtration sequence obtained.Here it is noted that filter, it is low pass filter preferably, obtains filtration sequence thereby be used to reduce useless difference, wherein the influence of useful difference compare with the influence of useless difference stronger, and useless difference even can eliminate fully.

This device further comprises output interface 25, the enhancement sequences that is used for the output filtering sequence or derives from filtration sequence, wherein output interface is configured to export at least and increases indication, reduces indication or keep not becoming indication, as the trend of filtration sequence or enhancement sequences, the useful difference of this trend representative object and healthy correlated condition.Can be obvious from Figure 21, output interface 25 is handled the filtration sequence of directly exporting from filter 22 with arrow 26 indications.Selectively, apparatus of the present invention further comprise second processor 27, and being used for derives enhancement sequences by filtration sequence.According to a certain environment, second processor 27 can comprise and close device, shown in Figure 22 a, perhaps can comprise any Signal Conditioning Equipment, and for example amplifier etc. obtains enhancement sequences to be exported thereby be used to revise filtration sequence.

About output interface 25, it is noted that here the trend indication can be pictorial display as shown in figure 20 certainly, shows filtering or enhanced sequence fully.Selectively, output interface can also only show trend, when trend shows arrow upwards or colored light up the time or by other indicating means, perhaps show downward arrow or different colored light or other indicating means down the time when trend, perhaps when indicating the symbol that does not move or indicate other when remaining unchanged.

Nature, for for example blind person or deaf mute, this can also express the interface by sense organ and be realized, and the sense organ indication that its output is specific is used for indicating status and improves, reduces or remain unchanged.Nature can also send this index signal by machinery, for example represents to increase with loud or intensive vibration, and reduction is represented in faint vibration, or very faint vibration is represented to remain unchanged.Selectively, for each indication, can make the frequency difference of vibration.Selectively, except vibrational indication means, can also use any other mechanical flag, for example improve a key, thereby the user can being felt compare this key with the state that key is not elevated is elevated.

Figure 22 a has shown an embodiment of second processor 27 among Figure 21.In this embodiment, and to close device be relevant sample (sample-wise) and close device, is used for for example heart rate being filtered back sequence sample and filters back sequence sample with blood pressure and multiply each other, thereby obtain the enhancement sequences that after-contraction pressure heart rate product is filtered in representative.

Figure 22 b has shown an embodiment of first processor 24, and blood pressure and heart rate merge but for example be used for before filtering, multiplying each other of relevant sample just, situation.This means that the embodiment of Figure 22 b forms original rate pressure product earlier, filter by filter 22 subsequently, thereby reduce the useless difference of original rate pressure product.

Figure 23 has shown that according to innovative means of the present invention, it comprises predictor 30, is used for providing in a certain moment that does not have the first Biomass sample estimated value of first Biomass.Preferably, use the measured value of the measurement result prediction intrusive mood measurement of one or more non-intrusion measurements, this will get in touch Figure 17 and Figure 20 is described in detail.According to a certain situation, predictor can be free-running operation predictor or the predictor that upgraded with rule or irregular spacing.

The present invention is that the simple idea of basis " knowledge provides power " makes up, and encourages the patient to improve life style.The present invention is by the usage trend analysis, and it is based on intact proof and traditional patient's measuring method, carries out patient monitoring in a kind of brand-new mode, and provides new and method improved indication patient's states.This improvement information can be used for patient and/or its doctor, is used to formulate treatment plan and follow-up.The present invention can promote and the education patient by using, thereby they can be made progress in the life style correction in force feedback.

In the type 2 diabetes mellitus relevant disease, doctor's practice at present is to inform that needs of patients changes dietary habit and life style, but is difficult to allow the patient judge usually and understands change required degree.Because " silence " property of this disease often is difficult to make the patient produce power.If the life style correction is carried out too radically, may cause power to exhaust and reduce, the patient may abandon.On the other hand, if carry out fully inadequately, then can not produce desired effects.Advantage of the present invention is clearly to have indicated the proper level of life style change in a kind of mode intuitively, thereby avoided losing the courage of effort.

The inventor believes, by the level of this method indication " just enough ", is the key of long-term promotion and successful rehabilitation.Its realization is by using new multi-parameter physiology method for monitoring in conjunction with trend indication clearly, thereby encourages oneself's control and make the patient by making great efforts to obtain good way of act, and provides negative indication when the patient is no longer progressive.This instant indication that can force feedback is compared with traditional medical practice and is wanted much superior and fully opposite, and the individual that wherein traditional medical practice only is the doctor by rareness visits and provides very secular " feedback ".

The present invention has illustrated a kind of new method and/or new equipment, and it needs minimum agreement of patient and effort, the some of them patient parameter can frequently sample, once a day or even weekly, the sample frequency of other parameter is lower.The parameter of frequent sampling can easily be carried out at bedside early morning every day, and does not need to operate any equipment or instrument.The patient parameter of low frequency sampling can be carried out in for example clinic.

In intensive or sparse mode, equidistantly or equidistantly the frequent sampling physiological parameter of sampling can not comprise:

Blood glucose

Sports

Blood pressure

Heart rate

Body temperature

Body weight

The body quality coefficient

Very the patient parameter of low frequency sampling can comprise:

HbAlc

Insulin

Blood fat

Albumin level

Interested other relevant parameter

When estimated blood sugar level at home, importantly, the analysis diversity of measurement device is low, and more much smaller than patient's difference biology.Otherwise measure nonsensical.Unfortunately, some human-like blood glucose meter have unacceptable high analyte diversity, make that they are not too reliable and can not be used for accurately blood glucose measurement.Yet, find that some commercially available low costs human-like blood glucose meter has enough precision, can obtain reliably for example fasting glucose measurement, and data are carried out suitable post processing.On the other hand,, can for example in several minutes, carry out twice or repeatedly successive measurement, in post processing, average subsequently if wish higher precision.Also can use a plurality of BG instrument and the result is average abreast, so that reduce diversity.This can carry out in the high-precision clinical research of needs, and once is used to verify the present invention under study for action.

Because blood sugar level has intensive biology of difference and the analysis difference of blood glucose measurement instrument, can obtain very in disorder data, makes that noisy signal is difficult to explain.See Fig. 4, it illustrates one and typically goes through about 10 months fasting glucose sequence.Note, because data are very noisy, thereby show high difference biology, so be difficult to accurately the patient be diagnosed.As time goes on the data that obtain are scattered in wide range, thus the span of patient B G from normal value to the diabetes value.If for our case study patient counting satisfies the natural law of each WHO standard, can obtain an interesting figure, see Fig. 5.There is 37% natural law estimation patient healthy fully in 257 days.There is 57% natural law to estimate that he suffers from impaired fasting glucose (IFG).There is 7% natural law to estimate that he suffers from tangible diabetes.

According to the strong discrepancy of top BG, the inventor thinks strongly, and present diabetes standard can cause the diagnosis of owing to optimize, and therefore needs to revise.More accurate for BG is explained, need carry out low-pass filter to a large amount of BG data.Yet, importantly do not want filtering data too much, because this can reduce the details of short term variations and reduction difference.Optimize filtration and avoid too much filtration to be realized by the residual analysis of back with explanation.

Although the BG measurement result looks extremely noisy, it can not be characterized as being white noise.For clear, can have a look the auto-correlation function (acf) of estimation, wherein dependency is clearly (see Fig. 6, wherein the acf of Gu Jiing is the long-term empty stomach BG according to the inventor).In this case study, measurement result roughly becomes the normal distribution (see figure 7).If measurement result has bigger difference, then be likely logarithm normal distribution.

Because the inventor's measurement strategies uses the high-quality BG instrument of three same brand can calculate analytical error.This is to realize that by comparing two BG instrument results at a time it produces the case that three similar normal state distribute, and standard deviation is about 0.35mmol/L (see figure 8).By the DS BG1 that three instruments produce, BG2 and BG3 independently of one another and be N (m, σ).BG is an arithmetic mean of instantaneous value, is expressed as follows:

$\stackrel{\‾}{\mathrm{BG}}\∈N(m,\frac{\mathrm{\σ}}{\sqrt{n}})---\left(1\right)$

Wherein standard deviation compares approximately equal for each, is 3 to number (n).By adopting following statistics principle, promptly the DS of two normal distributions can addition, and we obtain

$\sqrt{{2\mathrm{\σ}}^2}\≈0.35---\left(2\right)$

The standard deviation of three measuring instrument meansigma methodss using in case study here is approximately 0.14mmol/L.

Express for the clear trend that obtains these noisy data, must handle data with low pass filter, it can be realized by spectrum analysis.Can find out that from the periodogram of Fig. 9 most of energy are positioned at low-frequency band.Low-pass filter is handled by carry out multiplying in frequency domain.

$S_{\mathrm{LP}}\left(e^{\mathrm{j\ω}}\right)=H\left(e^{\mathrm{j\ω}}\right)\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{BG}\left(t\right)e^{-\mathrm{j\ωt}}---\left(3\right)$

Wherein H is the FIR low pass filter (see Figure 10, be used for the cut-off frequency that picks up is at random carried out frequency response) in the frequency domain, and BG (t) is the original measurement result through Fourier transform.S then _LPTransform back into time domain by inverse-Fourier transform.Therefore, can produce residual error.

ΔBG _fd(t)＝BG(t)-BG _LP(fd)(t)(4)

For certain the cut-off frequency fd (discrete frequency) between the 0-1.When fd when 0 is increased to 1, we can calculate the meansigma methods of residual error square, wherein N is the length of each fd value residual error vector.

$\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{\ΔBG}{\left(t\right)}_0^2\·\·\·\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{\ΔBG}{\left(t\right)}_1^2---\left(5\right)$

This will produce the curve (seeing Figure 11) of describing the behavior of different fd residual error.In order to find only cut-off frequency, the intersection that should be among Figure 11 is selected frequency.The main purpose of Figure 11 cathetus is to illustrate the position that the residual error curve disconnects.Identical residual analysis can be used in measures and signal other biology.When the low pass filter deal with data of design like this, multiply each other and to remove undesirable altofrequency by zero of LP filter being filled up (zero-padded) Fourier transform and BG measurement result.

Figure 12 and Fig. 4 have shown the LP filtering result of selecting cut-off frequency to carry out by in frequency domain and time domain respectively.

As a selection, can in time domain, carry out similar filtration with convolution.Those skilled in the art also can use the low pass filter of other type.

Blood pressure can be measured at both arms, carries out low-pass filter then to reduce difference.Blood pressure can also be in wrist, finger or other position measurement.Pulse wave transit time (PWTT) estimates also to can be used in the measurement blood pressure, it is by measuring pulse wave estimated blood pressure transit time, detect at finger with plethysmograph, from heart produce EKG R ripple for example begin to since the blood impulse density changes impulse wave produce slight transmission difference till.In addition,, can advantageously calculate mean arterial pressure (MAP) and pulse pressure (PP), and figure ground shows these data by heart contraction, diastole and pulse data.

According to similar mode and since daily routines than big-difference and proximate estimation, the sports data are normally dispersive.Therefore, be by this data that As time goes on obtain being carried out low-pass filter, because this can make the easier explanation of sports data easily to top similar mode.Sports can be estimated that wherein this yardstick can comprise following daily routines classification simply according to intensity dimension:

Extremely light (rest, reading, seat, driving etc.)

Gently (take a walk, clean, play the piano, be careful)

In (hurry up, jog, at leisure by bike, the retraining of skating, act rashly)

Heavy (swimming, running, high strength are jogged, bicycle race, football, basketball etc.)

Extremely heavy (box, row the boat, climb the mountain, heavily weight training exercises)

For more accurate estimation, can use MET (metabolic equivalent).Energy expenditure when 1MET equals tranquillization, gentle activity＜3MET, the moderate activity is 3-5.9MET, and the severe activity is 6-8.9MET, and utmost point taskwork is moved＞9MET.The MET hotlist can be used for simplifying incendiary calorie of institute of calculating (kcal), and its product by MET value, body weight and activity time obtains.Estimate that the active cost-effective method of physical culture is to use pedometer.The activity data of collecting in instance graph of the present invention has just used pedometer, and itself and built-in timer coupling are calculated approx by day or carried out during the sports incendiary calorie.With incendiary calorie of indication energy expenditure is feasible, because this is generally that use an and understandable term.

Owing to have bigger difference between every day, so heart rate data also is dispersive.Therefore, be with top similar methods this data being carried out low-pass filter, making the easier explanation of heart rate data easily.

In additional embodiment of the present invention, preferably measure heart contraction and diastole blood pressure data and heart rate data at both arms every day.Then that the data of both arms are average, and carry out low-pass filter to reduce difference.The product that calculates heart contraction blood pressure and heart rate obtains systolic pressure heart rate product (RPP), so that estimate the patient's body situation.RPP=heart contraction BP* heart rate/100.Except the coefficient of oxygen utilization of approximate indication heart, RPP also can show the existence of stimulant substance, for example caffeine, nicotine, cocaine and amphetamine, and spirit and emotion pressure.Therefore, the inventor thinks that RPP is an important parameter assessing and set up patient's general health correlated condition with BG.Have trend and each component of himself for what obtain RPP, we can use a kind of low-pass filter method, and it is similar to the low-pass filter method that is used to produce BG trend.This estimates between the arm of the left and right sides that according to the long-term average separately of each arm blood pressure difference also is valuable arbitrarily for the doctor.

According to similar mode, because existence is than the diversity of big-difference and analysis between every day, the blood pressure monitoring data during tranquillization also is dispersive.The blood pressure measurement that allows patient or doctor carry out single-point seems not too meaningful, because also there is bigger noise level in the BP data.Therefore need carry out low-pass filter to this data, this can make blood pressure data more accurately and easily explain, sees Figure 13.

By the designation data of more filtering sports and BG level simultaneously, but can find out sports blood sugar lowering level, for example blood sugar level and sports are inversely proportional to.Yet in fact sports extremely have retroaction in some cases, can increase blood sugar level.Therefore, by simultaneously providing these for example to use figured data to the patient, he can easily adopt and makes great efforts to carry out sports and the effort relevant with the mode of making the life better, thereby pursues predetermined target.This does not need to pay too much effort, and just every day blood glucose, RPP and exercise goal is carried out shirtsleeve operation according to the program in the figure in periodic mode being realized in accurate and intuitively mode now, sees Figure 14.Should be noted that and in curve chart, also can see new interesting decorrelation, see Figure 15.For example, when the patient suffered from influenza or viral infection, the BG value can be unexpectedly, be independent of the enhancing of sports and raise.When sports strengthened, the BG value also can raise, and RPP reduces, and dependency becomes negative.Perhaps when the patient ran into tense situation, the rising degree of RPP surpassed BG.Therefore can suspect and have the negative correlation incident in this case.Therefore, the time window correlation coefficient that calculates between RPP, BP and the exercise can provide a new interesting index about patient's states, and experienced user can according to this negative correlation indicate the conclusion that makes new advances.

In another embodiment of the present invention, provide a kind of new method, wherein the inventor finds, RPP is dynamic and the BG dynamic correlation is good, be inversely proportional to the sports level, so the fluctuation that RPP can be used to predict BG is seen Figure 15 and Figure 16 with dynamic.

In another one embodiment of the present invention, RPP can with the prediction filter coupling, calculate the alternative measurement result of daily BG.When the patient to blood sampling feel can not, unactual or agonize and when inconvenience, can advantageously adopt this new BG Forecasting Methodology.In this case, BG only measures and uses between initial processing or intervention period, is used to calibrate RPP predictor and BG value.After this calibration was finished, the patient can only get back to RPP and measure, and only just carried out BG and measure when going to see the doctor.In another one embodiment of the present invention, as mentioned below, the frequency ratio that the BG prediction filter upgrades is more sparse, for example weekly.Therefore, the initial training of prediction filter at first needs the measurement result data sequence of intensive sampling.The length of this training sequence can be 1 week to 1 month for example.Yet the renewal frequency of prediction filter can be sparse.Have dependency if detect between signal x1 and the x2, then the Forecasting Methodology that is proposed can be used in any signal x1 (noticing that x2 can be the combination of a plurality of measuring amount) that prediction draws from signal x2.

Therefore, BG and RPP can both be as the important indicators of oneself's control that improves type 2 diabetes mellitus relevant disease patient and life style improvement.RPP shows relevant with BG, particularly in the transition period that life style changes, for example changes to more active life style from sedentary lifestyle, perhaps during the intensity cycle that changes sports.Therefore, the derivative of the derivative of BG and RPP has intensive dependency (seeing Figure 16).In this case, the trend of RPP and BG parameter all changes in a similar manner, shows to have high correlation.Under steady-state conditions, when people are in " metabolic balance ", the dependency between RPP and the BG, its importance reduces, because the too much noise from other metabolic process is arranged in the data.Therefore, prediction filter is used for from the BG data of RPP data prediction every day of intensive sampling.It is noted that by using RPP to estimate that BG is the method for economy and no pain, do not need or only need use seldom blood sugar test paper bar or finger puncture.Therefore, the measurement blood pressure is used to calculate RPP and needs consumptive material unlike the BG test.The Forecasting Methodology that this paper proposes can also be used for other with BG measuring method in the future, when these methods are considered to clumsy, unactual or uneconomic.These methods can comprise from tear, saliva or by measuring the BG value with the instrument of contact skin etc.

The present invention only use sparse blood specimen collection by RPP with high-precision forecast BG every day.This prediction can be handled by two kinds of diverse ways.ARX and FIR-Wiener.As previously mentioned, the Forecasting Methodology data sequence that need be used to train.The BG value of this sparse sampling is used to upgrade advanced filtration predictor.Therefore, for the patient that blood sampling or finger puncture are had the wound sensitivity, it is for example weekly that the behavior of this misery can reduce to, and the BG value by sparse sampling still can be by predictor BG value every day that calculates to a nicety.This system utilizes transfer function and BG, input signal x and white noise to be verified.X can be the vector of a variable, perhaps the matrix of a plurality of variablees.The example of the variable that can measure comprises systolic pressure heart rate product, heart contraction blood pressure, diastole blood pressure, pulse, mean arterial pressure, pulse pressure or sports.This checking can realize, because we suppose that BG and x are subjected to the some effects of identical potential parameter.We find that these parameters comprise, for example sports, dietary habit, pressure, virus and overweight.Therefore, we can provide this system by following hypothesis.

BG(t)＝G(θ，q)x(t)+H(θ，q)e(t) (6)

It is that a linear system is expressed, and wherein noise part e (t) is a white noise at random, E[e (t)]=0.On wider meaning, this system can enough flight data recorder principles be illustrated (seeing Figure 17).G and H are transfer functions, and θ is the vector that contains multinomial coefficient.And q is a displacement operator.More preferably use systolic pressure heart rate product, because it and BG have the highest dependency.Therefore, following example uses BG and RPP data.

An important preprocessing during verification system is the deduction meansigma methods.By under provide:

$\stackrel{\‾}{\mathrm{BG}}=\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{BG}\left(t\right),\stackrel{\‾}{\mathrm{RPP}}=\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{RPP}\left(t\right)---\left(7\right)$

Wherein N measures number of times.Have several different methods be used to estimate transfer function G (θ, q) and H (θ, q), comprising model ARX is arranged, ARMAX, OE and Box-Jenkins.In the example that ARX is discussed, a direct prediction algorithm is provided, be called linear regression.Box-Jenkins is complicated model, and two other model is its special case.Test shows that the magnitude of deviation between the different model methods is very little.

The ARX model can be write as:

BG(t)+a ₁BG(t-1)+...+a _naBG(t-n _a)＝b ₁RPP(t-1-nk)+...+b _naRPP(t-n _b-nk+1)+e(t) (8)

Wherein multinomial coefficient can gather and write:

θ＝[a ₁...a _nab ₁...b _nb] ^T (9)

And equation 2 can be rewritten as:

A(q)BG(t)＝B(q)RPP(t-nk)+e(t) (10)

Wherein:

$G(q,\mathrm{\θ})=\frac{B\left(q\right)}{A\left(q\right)}$

And $H(q,\mathrm{\θ})=\frac{1}{A\left(q\right)}---\left(11\right)$

Nk postpones.

The best constituent element of given vector theta (optimal element) and old BG and RPP value might be predicted BG.Prediction is to know θ and the regression vector that contains have been friends in the past BG and RPP value

In time, calculate.

Notice that an e (t) is not

A constituent element.Further, θ and

Product provide this prediction.

In example, predictor is designed to an one-step prediction device, and become adaptive because it carries out retraining to prediction each time.Those skilled in the art also can use the predictor of other step-length or the predictor of other type.

Each result of calculation of θ during for t-1 is guessed or predictive value for one that all will produce BG (t).Therefore, when time t, might calculate forecast error

$\mathrm{\ϵ}(t,\mathrm{\θ})=\mathrm{BG}\left(t\right)-\hat{B}G\left(t\right|\mathrm{\θ})---\left(14\right)$

For length is the training sequence of N, can obtain quadratic criterion

$V_N\left(\mathrm{\θ}\right)=\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ϵ}}^2(t,\mathrm{\θ})---\left(15\right)$

Therefore, directly extract θ, it provides

${\widehat{\mathrm{\θ}}}_N=\arg \underset{\mathrm{\θ}}{\min }{V}_N\left(\mathrm{\θ}\right)---\left(16\right)$

(" arg min " is minimum argument)

Our forecast error is

Therefore, quadratic criterion (11) can be write as

$\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{BG}}^2\left(t\right){-2\mathrm{\θ}}^T{f}_N{+\mathrm{\θ}}^T{R}_N\mathrm{\θ}---\left(18\right)$

Wherein

And

If R _NReversible, then formula can be write as

$V_N=\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{BG}}^2\left(t\right){-f}_N^T{R}_N^{-1}{f}_N+(\mathrm{\θ}-R_N^{-1}{f}_N{)}^T{R}_N(\mathrm{\θ}-R_N^{-1}{f}_N)---\left(21\right)$

If $\mathrm{\θ}=\widehat{\mathrm{\θ}}=R_N^{-1}{f}_N---\left(22\right)$

Then the decline of (19) is always zero.And because R _NBe positive definite, so have minima.Therefore, when satisfying equation (22), can provide V _NBest minima (θ) is because remaining is independent of θ.In order to improve this predictor, can use residual information, it can obtain when adopting true BG sample.This residual error can so that add in the prediction that is about to carry out, be followed the tracks of to improve amplitude by weighting exponentially.

As an example, Figure 18 has shown the predictor result when upgrading with true BG sample in only per 7 days.In another embodiment of the present invention, predictor can free-running operation, and just upgraded with a short BG sequence at first.

As another prediction example, can carry out FIR-Wiener and filter, it is the powerful predictor that can suppress noise best.Can be write as the variable description of BG sample in the future (BG is an example of this problem certainly, can with the variable replacing BG of other sparse sampling relevant) with the intensive sampling variable.Therefore, we generate the BG predictive value as follows:

x _k＝BG(n+k) (23)

We generate the vector that contains BG and RPP measurement result.

Y[BG(n)...BG(n-tM)RPP(n+k)...RPP(n-T+k)] (24)

Wherein t is old BG value number, and M is the test interval of BG.T represents the old value of RPP, and k is predicted step number (k＜M).

Further, we estimate to contain the matrix of auto-correlation function and cross-correlation function.In order to calculate this estimated value, we use the sequence of known measurement as training sequence.

${\hat{R}}_{\mathrm{yy}}=E\left[y^{T}y\right]---\left(25\right)$

We also estimate cross-correlation

r _xy＝E[x _ky ^T] (26)

Now can by under produce filter (predict, k is the index of prediction steps at every turn)

$h_k={\hat{R}}_{\mathrm{yy}}^{-1}{r}_{\mathrm{xy}}---\left(27\right)$

It causes predictive value

$\hat{B}G(n+k)=h_k{y}^T---\left(28\right)$

Should be used for the analysis of the time series data of dependency, data loss information or the outer value of group that the meansigma methods of use adjacent data surpasses threshold value can be deducted.This is most important, because when the manual interpretation measurement result, often forgets measurement sometimes or occurs mistake sometimes.When object is forgotten its instrument or when object is spent a holiday, needed to insert long data loss in this case.The linear insertion also is an alternative method of linear regression Forecasting Methodology.For example, measure if carry out BG weekly, the vector of then known BG data is that sample every day one of BG vector reduces sampled version.For M days (or the sample) in any insertion interval, it can be expressed as

BG _M(n)＝BG(nM) (29)

Pass through at BG then _MConstituent element between the straight line used based on M-1 sample carry out linear interpolation.As an example, Figure 18 has shown the interpolation result of per 7 days true BG sample.Also can carry out linear interpolation for the not equidistant data that missing data is arranged.

Further, in another embodiment of the present invention, preferably automatically between linear interpolation and prediction, switch according to the missing data sequence.Appropriate point in the time of can determining finishing switching by the residual analysis of past given data.Residual error results from two kinds of situations: situation 1, wherein linear interpolation data and the initial data in the particular sequence compares.Situation 2, wherein very the prediction data of identical sequence and initial data compare.This certainly realizes with same intervals/sequence, the useful comparative result between two kinds of situations is provided.Its objective is that to make residual error as much as possible little, therefore, can determine switching point in the position of inserting from the residual error mean square of two kinds of situations.

The summation that BG and RPP factor decompose can be used as the metabolism performance indications, the inventor is called metabolism performance index (MPI), this index can be crossed over a large amount of unusual and diseases, therefore be the index that is used to improve type 2 diabetes mellitus relevant disease oneself control and life style improvement clearly, see Figure 19.Early stage indicated number, MPI index can also be advantageously used for athlete's etc. training.

The inventor believes, to become the very valuable resource that the patient carries out self management according to metabolism supervision of the present invention and indicating means, and be that the doctor is clear and accurately visit and follow the tracks of the new tool of patient's states, and can be used as valuable handling implement.Figure 20 has shown the screen printing of the software product of certain version.We also believe, this multiparameter metabolism supervision and index instrument can be used in and monitor your condition and progress, athlete for example, as runner and swimmer etc., and be used for any mammal, for example horse racing or dog race, wherein the trainer can advantageously use dateout to instruct further training and raising ability.

Substitute as new hardware development, can use the verified technology of standard and mass-produced user's medical monitoring instrument to collect data, wherein can use a computer program and computer (desktop computer, kneetop computer, palmtop computer or trimline) collect, download, analyze and with actual and intuitively mode information is provided.In addition, can make up a kind of intelligent blood glucose monitor, perhaps contain the blood glucose and the blood pressure monitor intelligence combined unit of the microprocessor that is useful on sports and accelerometer and display screen.The present invention can be embodied as software, hardware chip and DSP, is used for various uses, for example calculates, stores and/or send signal, simulation or numeral.

The embodiment of this paper explanation only is the illustration of the principle of the invention.Should be appreciated that the configuration that those skilled in the art is can obvious this paper illustrated and the various modifications and the change of details.Therefore, the present invention is only limited by the scope of accessory claim book, and the restriction of the detail that provides by the description and interpretation to this paper embodiment is not provided.

Claims (20)

1. device that is used for the healthy correlated condition of denoted object comprises:

Inputting interface (20), it is used to receive the primary sample sequence of first Biomass that obtains by first measuring method, be used to receive the primary sample sequence of second Biomass that obtains by second measuring method, this first measuring method is that intrusive mood is measured, object is had first to be impacted, this second measuring method is that non-intrusion type is measured, and object is had second impact, and each Biomass has useful difference and useless difference;

Wherein first Biomass provides the indication than the healthy correlated condition of the accurate more object of second Biomass, and wherein the healthy correlated condition of first Biomass and second Biomass and object has dependency, and wherein second impact is impacted less than first;

Predictor (30), it is used to utilize the sample of second Biomass and the sample of first Biomass that obtains as far as possible, and in certain time that does not have first Biomass, the estimated value that first Biomass is provided is as the prediction sample;

Filter (22), it is used to filter the sequence with the first Biomass sample and at least one prediction sample, the sequence after the filtration with filter before compare useless difference with useful difference and minimizing, and

Output interface (25), it is used for output at least and increases indication, reduction indication or the indication that remains unchanged, as data trend, the useful difference of the healthy correlated condition of this trend representative object.

2. according to the device of claim 1,

Wherein first measuring method is that blood or plasma glucose are measured, and

Wherein second measuring method is heart rate measurement, blood pressure measurement or the method that is used to obtain heart rate and blood pressure product.

3. according to the device of claim 1, wherein inputting interface (20) is set to receive the first sample subsequence and second sample sequence, as the sample sequence of second Biomass, and

Wherein this device further comprises and closes device (27), is used to merge the first sample subsequence and second sample sequence, thereby obtains the second Biomass sample sequence, should and close the product that device is set to carry out relevant sample.

4. according to the device of claim 1, wherein first Biomass is blood sugar level, blood lipid level or the blood insulin levels of object.

5. according to the device of claim 1, wherein healthy correlated condition is the metabolism disorder that diabetes are relevant or glucose is relevant or insulin is relevant.

6. according to the device of claim 1, wherein first Biomass is a blood sugar level.

7. according to the device of claim 1, wherein predictor (30) is set to receive a plurality of samples that obtain from first measuring method as training sequence, and after the training stage, receives one or more samples that obtain from second measuring method at run duration.

8. according to the device of claim 7, wherein this runtime continues up to the sample that is drawn by first measuring method and is received till moment as the forecast updating value, perhaps wherein the runtime is unlimited, makes that predictor is the predictor of free-running operation after the training stage.

9. according to the device of claim 1, further comprise datin, it uses one or more previous samples or one or more subsequently moment of sample to provide the insertion sample for the disappearance sample at Biomass, with the processing sequence that obtains to have original series and insert sample.

10. according to the device of claim 1, its middle filtrator is a low pass filter.

11. according to the device of claim 10,

Wherein this low pass filter has cut-off frequency, and it is set the low frequency energy that makes original series or processing back sequence frequency for and accounts for predetermined portions original or processing back sequence gross energy.

12. according to the device of claim 10 or 11, further comprise the cut-off frequency computer, it is set to use following steps to determine cut-off frequency:

Determine different cut-off frequencies for low pass filter, be used for filtered sample to obtain to filter the back test signal;

Filter the back test signal for each, derive residual values, to obtain the residual error expression based on original series and the difference of filtering the back test signal;

Express based on residual error, determine to be adapted to separately the cut-off frequency of primary sample sequence.

13. according to the device of claim 12, wherein the cut-off frequency computer is used for following definite cut-off frequency:

Utilizing residual energy is that low cut-off frequency is determined first line;

Utilize residual energy to determine second line for higher cutoff frequency; With

Seek first line and second-line cross point, cut-off frequency is represented in this cross point.

14. device according to claim 1, wherein output interface (25) is set to indicate this trend by sound indicator, optical indicator or pop-up indicator, makes reduction indicate, increases the indication acoustics ground of indicating or remain unchanged, differs from one another optically or mechanically.

15. according to the device of claim 1, wherein output interface (25) is set to derive and export this trend from the actual value that filters back sequence or enhancement sequences and the in good time advance value of filtering back sequence or enhancement sequences.

16. according to the device of claim 1, wherein output interface is set to the demonstration of figure ground and filters back sequence or enhancement sequences.

17. according to the device of claim 1, wherein output interface (25) is set to figure ground demonstration predictive value sequence or filters back predictive value sequence, in addition, also shows further filtering or enhanced sequence at least.

18. according to the device of claim 17, wherein the predictive value sequence is the blood glucose value sequence of object, and extra biological parameter is the rate-pressure product that is used for the blood glucose value prediction of object.

19. according to the device of claim 1, wherein first measuring method is a biochemical measuring method or based on the measuring method of liquid measure, and wherein second measuring method is physiological measurements method or the method measured based on on-liquid.

20. according to the device of claim 1,

Wherein predictor (30) is set to carry out regression algorithm, and

Wherein predictor (30) further was set to receive updating value from first measuring method after the runtime, and an interval comprises at least that the sample of second measuring method is received by inputting interface 3 times.

Images