An odor feedback method based on artificial intelligence

Abstract

The present invention discloses an artificial intelligence-based odor feedback method, which relates to the technical field of odor feedback. The method comprises the following steps: dynamically calculating the clearance rate of odor components in patients with liver and kidney dysfunction by using a CYP2A6 enzyme kinetic model and a renal replacement therapy adsorption equation; combining the liver and kidney metabolic overload threshold constraints of a contraindication mapping table to reduce toxicity risks and solve the metabolic uncontrollability problem of traditional methods; designing a pulse release waveform based on an olfactory sensitivity attenuation model, coupling reverse airflow flushing, shortening the receptor recovery period and the EEG response latency, thereby overcoming the desensitization effect caused by sustained release.

Description

Technical Field

[0001] The present invention relates to the technical field of odor feedback, and in particular to an odor feedback method based on artificial intelligence. Background Art

[0002] At present, the incidence of nocturnal delirium in ICU patients is as high as 80% due to environmental stimulation, pain, metabolic disorders and other factors. Existing aromatherapy mostly uses sedative scents such as lavender and chamomile, which are released continuously for 6-8 hours at night through a nebulizer at a fixed concentration, such as 2% essential oil solution, to regulate limbic system activity (amygdala-prefrontal cortex connection) through the olfactory pathway, thereby reducing delirium-related agitation.

[0003] However, patients' physiological states fluctuate dramatically at night, such as pain outbreaks and drug metabolism peaks, but the intensity of odor release is fixed. In this case, nurses need to manually turn off the nebulizer and rely on manual supervision. In addition, the continuous stimulation of a single odor causes olfactory fatigue, and the intervention effect will drop sharply. Studies have shown that the sedative effect of lavender decays by 62% after 45 minutes. Although intermittent release can be used, the cycle setting relies on experience and cannot match the differences in individual receptor recovery speeds.

[0004] It can be seen that the current aromatherapy is essentially open-loop control, which simplifies the complex neurometabolic system into a linear model and intervenes in the dynamic pathophysiological process through fixed parameters, with poor results. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides an odor feedback method based on artificial intelligence to solve

[0007] Existing aromatherapy cannot adjust to the pain outbreak period and drug metabolism peak when treating symptoms such as nocturnal delirium; and the continuous stimulation of a single smell causes olfactory fatigue.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] The embodiment of the present invention provides an odor feedback method based on artificial intelligence, which includes:

[0010] Acquiring a real-time physiological data set of a target patient, wherein the data set includes at least biochemical indicators of liver and kidney function, respiratory rate time series data, and bispectral index;

[0011] Constructing a knowledge graph containing the metabolic pathways of odor molecules, which correlates the patient's individual metabolic capacity with the clearance rate of odor components;

[0012] generating a dynamic release strategy based on the real-time physiological data set and the knowledge graph, wherein the strategy includes pulsed release waveform parameters and an odor component taboo mapping table;

[0013] The dynamic release strategy is executed by the atomization control module, and the execution process includes adjusting the particle size distribution of the atomized particles according to the real-time nasal mucosal humidity feedback.

[0014] As a preferred solution of the artificial intelligence-based odor feedback method of the present invention, the real-time physiological data set acquisition device includes:

[0015] an embedded biosensor array that adheres to the outer wall of the nasal cavity and measures the rate of change of mucosal impedance;

[0016] The non-contact millimeter-wave radar is installed 20 to 30 cm above the bed and detects the millimeter-level displacement of the chest.

[0017] As a preferred solution of the artificial intelligence-based odor feedback method described in the present invention, the method for constructing the knowledge graph includes:

[0018] Establishing a corresponding relationship library of odor molecule metabolizing enzymes, wherein the relationship library includes decomposition rate parameters of linalool by CYP2A6 enzyme;

[0019] Identify metabolic compensation pathways for patients with impaired liver and kidney function, including molecular adsorption rate correction factors during renal replacement therapy;

[0020] Integrate real-time hemodynamic data to calculate dynamic estimates of vascular permeability of odor components.

[0021] As a preferred embodiment of the artificial intelligence-based odor feedback method of the present invention, in the knowledge graph, metabolic enzyme relationship modeling is performed, an enzyme-substrate kinetic model is constructed, and the linalool decomposition rate equation is defined as:

[0022]

[0023] Among them, v c represents the catalytic rate of linalool by CYP2A6, in μmol / min. The subscript c refers specifically to CYP2A6 enzyme, and k cat is the catalytic constant for the decomposition of the enzyme-substrate complex into products, in units of s -1 , [E] is the free CYP2A6 enzyme concentration, in nM, [S] is the linalool substrate concentration, in mM, K m is the Michaelis constant, which represents the affinity between enzyme and substrate, and its unit is mM. β is the factor affecting the activity of competitive inhibitor on enzyme, which is expressed as β = 0.01·e 0.5·ΔGThe calculations show that [I] is the concentration of the competitive inhibitor in plasma, expressed in μM, and ΔG is the change in the binding free energy of the inhibitor to the enzyme active site, expressed in kcal / mol;

[0024] Dynamically calculate vascular permeability and establish the adsorption kinetic equation for renal replacement therapy. The formula is:

[0025]

[0026] Among them, C eff It indicates the effective concentration of adsorption clearance during renal replacement therapy, in mg / L, C blood Indicates the real-time concentration of molecules to be cleared in the blood, in mg / L, k ad represents the binding rate constant of the adsorbent material to the target molecule, in min-1, t represents the duration of a single renal replacement therapy, in min, η cr is a dynamic correction factor based on creatinine clearance, according to η cr =0.8+0.02·(CrCl-30), effective when CrCl≥30mL / min;

[0027] Dynamic calculation of vascular permeability, the calculation formula is:

[0028]

[0029] Among them, C tis Indicates the concentration of odor molecules in the target tissue, in μg / g, P v Indicates the dynamic vascular permeability coefficient in cm / s, according to P v =0.1·CO·(1+0.05·ΔRR) is updated in real time. A represents the effective capillary exchange area in cm 2 , calculated based on ultrasound blood flow imaging data, ΔC is the concentration gradient between blood and tissue, in μg / mL / cm, τ met represents the local metabolic time constant of the tissue, measured in minutes by microdialysis technology. CO represents the real-time cardiac output, measured in L / min. ΔRR represents the deviation of the current respiratory rate from the baseline value, measured in breaths / min.

[0030] As a preferred embodiment of the odor feedback method based on artificial intelligence described in the present invention, in the dynamic release strategy, olfactory sensitivity attenuation modeling is performed to establish a correlation model between drug concentration and olfactory threshold:

[0031]

[0032] Among them, S(t) represents the olfactory sensitivity at time t, in AU, which is calibrated by the electrophysiological signal of the olfactory bulb, S0 is the baseline sensitivity, and k decay is the neural adaptation attenuation coefficient, in min -1 , corrected to k by the bispectral index decay =6·BIS+3, C drug represents the concentration of sedative drugs in plasma, in μg / mL. The pharmacokinetic curve was read from the electronic medical record system. λ represents the olfactory receptor recovery rate constant, in min-1, which is negatively correlated with age. The update formula is: λ = 0.02·(1-0.005·Age);

[0033] To predict the risk of respiratory depression, we constructed an LSTM-ATT prediction model, which is expressed as:

[0034] h t =LSTM(x t ,h t-1 ;W h ,b h ),

[0035]

[0036] Among them, h t Represents the hidden state at time t, dimension 128, W h represents the weight matrix (256×128), b h is the bias term, ΔRR t Indicates the time series change of respiratory rate, in times / min 2 , e t is the attention matching degree, v is the attention weight vector, 64×1, W a is the attention matrix 192×64, α t is the temporal attention weight, R risk is the risk probability of respiratory depression in the next 10-15 minutes (0-1), and σ is the Sigmoid function.

[0037] As a preferred embodiment of the artificial intelligence-based odor feedback method of the present invention, the step of generating the dynamic release strategy specifically includes:

[0038] Historical sedative drug use records were extracted from the electronic medical record system to establish a drug half-life olfactory sensitivity attenuation model;

[0039] Extract features from the respiratory rate time series data based on the LSTM network to predict the risk level of respiratory depression within the next 10 to 15 minutes;

[0040] When the respiratory depression risk level is detected to exceed the threshold, the odor component replacement mechanism is automatically triggered to switch the release formula to an essential oil combination that does not contain monoterpenes;

[0041] The triggering conditions of the odor component replacement mechanism further include:

[0042] When the real-time blood oxygen saturation decrease rate ΔSpO2 / Δt ≤ -0.5% / min, the current release pulse is immediately interrupted;

[0043] After the interruption, the low-flow pure oxygen carrier gas flush is started in the expiratory phase of the first respiratory cycle, and the flush time is t = 3*RR -1 , where RR is respiratory rate, in beats / minute.

[0044] As a preferred embodiment of the artificial intelligence-based odor feedback method of the present invention, the pulsed release waveform parameters include:

[0045] The pulse width interval is set to [T min ,T max ], where T min ≥10 seconds and T max ≤60 seconds;

[0046] Interval time dynamic adjustment coefficient α = 1 + 0.2 * current BIS value - target BIS value, where α∈[0.8,1.5];

[0047] The peak concentration constraint is C peak ≤0.03*CrCl+0.5, where CrCl is creatinine clearance, in mL / min.

[0048] As a preferred embodiment of the odor feedback method based on artificial intelligence described in the present invention, in the dynamic release strategy, dynamic optimization of pulse parameters is performed, and the pulse parameter generation equation under the constraint conditions is established as follows:

[0049]

[0050] C peak =min(0.03·CrCl+0.5,0.8·S(t)·e -0.1·(BIS-40) ),

[0051] Among them, T width Indicates pulse width, C peak Indicates the peak concentration, T base is the reference pulse width, which is preset to 30 seconds, and clip is used to constrain the value to [T min ,T max ] is the truncation function within the interval, S(t) is the current olfactory sensitivity;

[0052] Perform real-time updates of taboo mapping and construct molecular interaction constraints as follows:

[0053]

[0054] in, is a set of taboo molecules,

[0055] is the odor molecule m i With m j The second derivative of the concentration interaction in tissues, ξ, is the liver and kidney metabolic overload threshold, in (μg / g) 2 , calculated according to the knowledge graph is ξ=2.5·(1-e 0.1·BIS ).

[0056] As a preferred solution of the artificial intelligence-based odor feedback method of the present invention, the execution process of the atomization control module further includes:

[0057] During the interval between adjacent pulses, reverse airflow is started, and the reverse airflow velocity V=0.5Q+2, where Q is the total amount of odor molecules released in the previous pulse stage, in μL;

[0058] According to the ambient temperature and humidity sensor data, the oscillation frequency of the atomizer is dynamically adjusted to f = 200 + 10 (T-25) + 5* (H-60), where T is the temperature in ° C and H is the humidity in % RH.

[0059] As a preferred embodiment of the artificial intelligence-based odor feedback method of the present invention, the method further comprises establishing an olfactory-visual cross-modal compensation mechanism:

[0060] The coordinates of the patient's gaze area are captured through an eye tracking device. When the patient's gaze falls on the alarm device area for 5 or more consecutive times, the release of the menthol component is activated;

[0061] When the eye movement frequency is detected to be greater than 8 Hz during rapid eye movement sleep, the release base carrier is switched to a tetradecane microcapsule suspension with phase-change cooling properties.

[0062] The beneficial effects of the present invention are as follows: the present invention dynamically calculates the clearance rate of odor components in patients with liver and kidney dysfunction through the CYP2A6 enzyme kinetic model and the renal replacement therapy adsorption equation, and combines the liver and kidney metabolic overload threshold constraint of the contraindication mapping table to reduce the toxicity risk and solve the metabolic uncontrollable problem of traditional methods; the pulse release waveform is designed based on the olfactory sensitivity attenuation model, and coupled with reverse airflow flushing, the receptor recovery period is shortened, the EEG response latency is shortened, and the desensitization effect caused by continuous release is broken through; through the atomization frequency temperature and humidity compensation formula and the phase change microcapsule carrier, the atomization particle size fluctuation is compressed from ±2.1μm to ±0.3μm under extreme conditions, and the deposition efficiency is increased from 52% to 89%, overcoming the influence of environmental parameter drift; through multimodal data fusion and dynamic constraint model, millisecond-level spatiotemporal synchronization of odor intervention and real-time mapping of metabolic safety boundaries are achieved, systematically solving the static control defects in the background technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0064] Figure 1 Schematic diagram of the process of the artificial intelligence-based odor feedback method in Example 1. DETAILED DESCRIPTION

[0065] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0066] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0067] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0068] Example 1, with reference to Figure 1 This embodiment provides an artificial intelligence-based odor feedback method, comprising the following steps:

[0069] Step S1, obtaining a real-time physiological data set of a target patient, wherein the data set at least includes biochemical indicators of liver and kidney function, respiratory rate time series data, and bispectral index;

[0070] The real-time physiological data set acquisition device includes:

[0071] an embedded biosensor array that adheres to the outer wall of the nasal cavity and measures the rate of change of mucosal impedance;

[0072] A non-contact millimeter-wave radar is installed 20 to 30 cm above the bed and detects millimeter-level displacement of the chest cavity;

[0073] Step S2, constructing a knowledge graph containing the metabolic pathways of odor molecules, wherein the knowledge graph associates the patient's individual metabolic capacity with the clearance rate of odor components;

[0074] The method for constructing the knowledge graph includes:

[0075] Establishing a corresponding relationship library of odor molecule metabolizing enzymes, wherein the relationship library includes decomposition rate parameters of linalool by CYP2A6 enzyme;

[0076] Identify metabolic compensation pathways for patients with impaired liver and kidney function, including molecular adsorption rate correction factors during renal replacement therapy;

[0077] Integrate real-time hemodynamic data to calculate dynamic estimates of vascular permeability of odor components;

[0078] In the knowledge graph, metabolic enzyme relationship modeling is performed, an enzyme-substrate kinetic model is constructed, and the linalool decomposition rate equation is defined as:

[0079]

[0080] Among them, v c represents the catalytic rate of linalool by CYP2A6, in μmol / min. The subscript c refers specifically to CYP2A6 enzyme, and k cat is the catalytic constant for the decomposition of the enzyme-substrate complex into products, in units of s -1 , [E] is the free CYP2A6 enzyme concentration, in nM, [S] is the linalool substrate concentration, in mM, K m is the Michaelis constant, which represents the affinity between enzyme and substrate, and its unit is mM. β is the factor affecting the activity of competitive inhibitor on enzyme, which is expressed as β = 0.01·e 0.5·ΔG The calculations show that [I] is the concentration of the competitive inhibitor in plasma, expressed in μM, and ΔG is the change in the binding free energy of the inhibitor to the enzyme active site, expressed in kcal / mol;

[0081] Dynamically calculate vascular permeability and establish the adsorption kinetic equation for renal replacement therapy. The formula is:

[0082]

[0083] Among them, C eff It indicates the effective concentration of adsorption clearance during renal replacement therapy, in mg / L, C blood Indicates the real-time concentration of molecules to be cleared in the blood, in mg / L, k ad Indicates the binding rate constant of the adsorbent material to the target molecule, in min -1 , t represents the duration of a single renal replacement therapy, in min, η cr is a dynamic correction factor based on creatinine clearance, according to η cr =0.8+0.02·(CrCl-30), effective when CrCl≥30mL / min;

[0084] Dynamic calculation of vascular permeability, the calculation formula is:

[0085]

[0086] Among them, C tis Indicates the concentration of odor molecules in the target tissue, in μg / g, P v Indicates the dynamic vascular permeability coefficient in cm / s, according to P v =0.1·CO·(1+0.05·ΔRR) is updated in real time. A represents the effective capillary exchange area in cm 2 , calculated based on ultrasound blood flow imaging data, Δc is the concentration gradient between blood and tissue, in μg / mL / cm, τ met represents the local metabolic time constant of the tissue, in minutes, measured by microdialysis technology; CO represents the real-time cardiac output, in L / min; ΔRR represents the deviation of the current respiratory rate from the baseline value, in breaths / min;

[0087] In the dynamic release strategy, olfactory sensitivity attenuation modeling was performed to establish a correlation model between drug concentration and olfactory threshold:

[0088]

[0089] Among them, S(t) represents the olfactory sensitivity at time t, in AU, which is calibrated by the electrophysiological signal of the olfactory bulb, S0 is the baseline sensitivity, and k decay is the neural adaptation attenuation coefficient, in min -1 , corrected to k by the bispectral index decay =6·BIS+3, C drugrepresents the concentration of sedative drugs in plasma in μg / mL. The pharmacokinetic curve is read from the electronic medical record system. λ represents the olfactory receptor recovery rate constant in min. -1 , negatively correlated with age, the update formula is: λ = 0.02·(1-0.005·Age);

[0090] To predict the risk of respiratory depression, we constructed an LSTM-ATT prediction model, which is expressed as:

[0091] h t =LSTM(x t ,h t-1 ;W h ,b n ),

[0092]

[0093] Among them, h t Represents the hidden state at time t, dimension 128, W h represents the weight matrix (256×128), b h is the bias term, ΔRR t Indicates the time series change of respiratory rate, in times / min 2 , e t is the attention matching degree, v is the attention weight vector, 64×1, W a is the attention matrix 192×64, α t is the temporal attention weight, R risk is the probability of respiratory depression risk in the next 10-15 minutes (0-1), and σ is the Sigmoid function;

[0094] Step S3, generating a dynamic release strategy based on the real-time physiological data set and the knowledge graph, wherein the strategy includes pulsed release waveform parameters and an odor component taboo mapping table;

[0095] The steps of generating the dynamic release strategy specifically include:

[0096] Historical sedative drug use records were extracted from the electronic medical record system to establish a drug half-life olfactory sensitivity attenuation model;

[0097] Extract features from the respiratory rate time series data based on the LSTM network to predict the risk level of respiratory depression within the next 10 to 15 minutes;

[0098] When the respiratory depression risk level is detected to exceed the threshold, the odor component replacement mechanism is automatically triggered to switch the release formula to an essential oil combination that does not contain monoterpenes;

[0099] The triggering conditions of the odor component replacement mechanism further include:

[0100] When the real-time blood oxygen saturation decrease rate ΔSpO2 / Δt ≤ -0.5% / min, the current release pulse is immediately interrupted;

[0101] After the interruption, the low-flow pure oxygen carrier gas flush is started in the expiratory phase of the first respiratory cycle, and the flush time is t = 3*RR -1 , where RR is respiratory rate, in beats / min;

[0102] The pulsed release waveform parameters include:

[0103] The pulse width interval is set to [T min ,T max ], where T min ≥10 seconds and T max ≤60 seconds;

[0104] Interval time dynamic adjustment coefficient α = 1 + 0.2 * current BIS value - target BIS value, where α∈[0.8,1.5];

[0105] The peak concentration constraint is C peak ≤0.03*CrCl+0.5, where CrCl is creatinine clearance, in mL / min;

[0106] In the dynamic release strategy, the pulse parameters are dynamically optimized, and the pulse parameter generation equation under the constraint conditions is established as follows:

[0107]

[0108] C peak =min(0.03·CrCl+0.5,0.8·S(t)·e -0.1·(BIS-40) ),

[0109] Among them, T width Indicates pulse width, C peak Indicates the peak concentration, T base is the reference pulse width, which is preset to 30 seconds, and clip is used to constrain the value to [T min ,T max ] is the truncation function within the interval, S(t) is the current olfactory sensitivity;

[0110] Perform real-time updates of taboo mapping and construct molecular interaction constraints as follows:

[0111]

[0112] in, is a set of taboo molecules,

[0113] is the odor molecule mi With m j The second derivative of the concentration interaction in tissues, ξ, is the liver and kidney metabolic overload threshold, in (μg / g) 2 , calculated according to the knowledge graph is ξ=2.5·(1-e 0.1·BIS );

[0114] Specifically, the strategy achieves precise intervention through four-level dynamic constraints; the decay model couples the pharmacokinetic integral with the neural adaptive decay to solve the hysteresis problem of the traditional static threshold model;

[0115] The LSTM-ATT model uses respiratory rate changes as contextual features for the attention mechanism, improving risk prediction accuracy by 23%. Pulse parameter generation introduces dual constraints, incorporating the antagonistic effects of CrCl and BIS into peak concentration calculations to avoid liver and kidney metabolic overload. Taboo mapping captures molecular synergy through second-order derivatives, and its dynamic threshold mechanism can predict metabolic conflicts 15 minutes in advance.

[0116] Step S4, executing the dynamic release strategy through the atomization control module, the execution process includes adjusting the atomized particle size distribution according to real-time nasal mucosal humidity feedback;

[0117] The atomization control module execution process also includes:

[0118] During the interval between adjacent pulses, reverse airflow is started, and the reverse airflow velocity V=0.5Q+2, where Q is the total amount of odor molecules released in the previous pulse stage, in μL;

[0119] According to the ambient temperature and humidity sensor data, the oscillation frequency of the atomizer is dynamically adjusted to f = 200 + 10 (T-25) + 5 * (H-60), where T is the temperature in ° C and H is the humidity in % RH;

[0120] The method also includes establishing an olfactory-visual cross-modal compensation mechanism:

[0121] The coordinates of the patient's gaze area are captured through an eye tracking device. When the patient's gaze falls on the alarm device area for 5 or more consecutive times, the release of the menthol component is activated;

[0122] When the eye movement frequency is detected to be greater than 8 Hz during rapid eye movement sleep, the release base carrier is switched to a tetradecane microcapsule suspension with phase-change cooling properties.

[0123] Example 2 provides a temperature and humidity compensation mechanism for the oscillation frequency of the atomizer in an artificial intelligence-based odor feedback method. This mechanism addresses the problem of atomized particle size distribution drifting due to ambient temperature and humidity fluctuations, resulting in a drop of more than 40% in nasal deposition efficiency. The mechanism includes:

[0124] Configure the hardware:

[0125] Piezoelectric atomizer: Material is PZT-5H (piezoelectric constant d 33 =650×10 -12 C / N, frequency temperature coefficient β T =-0.02% / °C);

[0126] Environmental sensor: SHT35 temperature and humidity sensor (detection accuracy ±0.2℃ / ±2%RH) is used.

[0127] Compensation logic: Calculate the oscillation frequency of the atomizer, f = 200 + 10 × (T-25) + 5 × (H-60),

[0128] Temperature compensation term 10×(T-25):

[0129] Experimental measurements show that the resonant frequency of the piezoelectric piece shifts by 0.3% for every 1°C increase in temperature at 25°C (verified by an impedance analyzer).

[0130] Humidity compensation item 5×(H-60):

[0131] According to ASTM E104-02, for every 10% RH increase in humidity, the surface tension of the solution decreases by 1.2 mN / m, resulting in an increase in the mass median diameter (MMAD) by 1.8 μm.

[0132] Implementation steps:

[0133] Calibration stage: Under 25℃ / 60%RH conditions, adjust the driving voltage so that the resonant frequency of the atomizer is 200Hz;

[0134] Real-time adjustment: When the ambient temperature rises to 28°C and the humidity is 75%, calculate the target frequency:

[0135] f=200+10×(28-25)+5×(75-60)=305Hz;

[0136] The driving voltage amplitude is adjusted by the PID controller so that the actual frequency error is ≤±2Hz.

[0137] Experimental verification was performed, see Table 1;

[0138] Table 1:

[0139]

[0140] Example 3 provides a temperature-sensitive microcapsule carrier system in an artificial intelligence-based odor feedback method, which addresses the problem of nasal mucosal temperature rising by 0.5-1.2°C during rapid eye movement (REM) sleep, leading to an imbalance in the volatility of conventional essential oils, including:

[0141] Microcapsule preparation parameters:

[0142] Core material composition: Tetradecane to Hexadecane molar ratio 7:3 (eutectic point 34±0.5℃);

[0143] Wall material structure: gelatin-gum arabic composite film (cross-linking degree 85%, thickness 2.8±0.2μm);

[0144] Particle size distribution: D10 = 10 μm, D50 = 28 μm, D90 = 50 μm (measured by laser particle size analyzer).

[0145] Carrier switching control:

[0146] Trigger condition: eye movement frequency > 8 Hz for 10 seconds (detected by Tobii Pro Nano eye tracker);

[0147] The comparison of carrier parameters is shown in Table 2;

[0148] Table 2:

[0149]

[0150] Phase change characteristics verification data:

[0151] Differential Scanning Calorimetry (DSC):

[0152] Melting onset temperature: 33.7°C;

[0153] Melting enthalpy: 218 J / g (testing instrument: TA Q2000, heating rate 5°C / min).

[0154] Thermal response test: When the temperature of the nasal mucosa rises from 33.5°C to 34.2°C, the release rate of the microcapsules increases by 180%.

[0155] Example 4 provides a precise flushing mechanism for the expiratory phase in an artificial intelligence-based odor feedback method. This addresses the problem of increased alveolar dead space caused by the misalignment of traditional flushing strategies with the respiratory cycle phase, including:

[0156] Time calculation model: flushing time, t = 3 × RR -1 ,

[0157] Physiological basis: Adult anatomical dead space volume: 150 mL;

[0158] Flushing air flow rate: 500 mL / s (according to ISO 80601-2-55 standard);

[0159] Theoretical flushing time: 150mL / 500mL / s=0.3s, covering one exhalation cycle (RR=20 times / min, cycle is 3s).

[0160] Implementation steps:

[0161] Dynamic monitoring: Detect the starting point of the expiratory phase through the respiratory flow sensor (triggered when the peak expiratory velocity drops to -200mL / s);

[0162] Interruption condition: When the rate of decrease of blood oxygen saturation ΔSpO2 / Δt ≤ -0.5% / min, the odor release is stopped immediately;

[0163] Washout execution: Start 500 mL / s pure oxygen flow during the expiratory phase, with a duration of t = 3 / RRmin (e.g., when RR = 18 times / min, t = 10 seconds).

[0164] Lung function compensation correction:

[0165] Adding a correction factor t for patients with chronic obstructive pulmonary disease (COPD) adj :

[0166] t adj =t×(1+0.05×FEV1%), where FEV1% is the percentage of forced expiratory volume in one second to the predicted value.

[0167] Comparison of clinical effects is shown in Table 3;

[0168] Table 3:

[0169]

[0170] All experimental data of Examples 2-4 are based on ISO / IEC 17025 certified laboratory testing and comply with medical device clinical validation standards.

[0171] In summary, it can be seen that the present invention dynamically calculates the clearance rate of odor components in patients with liver and kidney dysfunction through the CYP2A6 enzyme kinetic model and the renal replacement therapy adsorption equation, and combines the liver and kidney metabolic overload threshold constraint of the contraindication mapping table to reduce the toxicity risk and solve the metabolic uncontrollable problem of traditional methods; based on the olfactory sensitivity attenuation model, the pulse release waveform is designed, coupled with reverse airflow flushing, so as to shorten the receptor recovery period and the EEG response latency, and break through the desensitization effect caused by continuous release; through the atomization frequency temperature and humidity compensation formula and the phase change microcapsule carrier, the atomization particle size fluctuation is compressed from ±2.1μm to ±0.3μm under extreme conditions, and the deposition efficiency is increased from 52% to 89%, overcoming the influence of environmental parameter drift; through multimodal data fusion and dynamic constraint model, millisecond-level spatiotemporal synchronization of odor intervention and real-time mapping of metabolic safety boundaries are achieved, systematically solving the static control defects in the background technology.

[0172] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based odor feedback method, characterized by: include, Acquiring a real-time physiological data set of a target patient, wherein the data set includes at least biochemical indicators of liver and kidney function, respiratory rate time series data, and bispectral index; Constructing a knowledge graph containing the metabolic pathways of odor molecules, which correlates the patient's individual metabolic capacity with the clearance rate of odor components; generating a dynamic release strategy based on the real-time physiological data set and the knowledge graph, wherein the strategy includes pulsed release waveform parameters and an odor component taboo mapping table; The dynamic release strategy is executed by the atomization control module, and the execution process includes adjusting the atomized particle size distribution according to the real-time nasal mucosal humidity feedback; The method for constructing the knowledge graph includes: Establishing a corresponding relationship library of odor molecule metabolizing enzymes, wherein the relationship library includes decomposition rate parameters of linalool by CYP2A6 enzyme; Identify metabolic compensation pathways for patients with impaired liver and kidney function, including molecular adsorption rate correction factors during renal replacement therapy; Integrate real-time hemodynamic data to calculate dynamic estimates of vascular permeability of odor components; In the knowledge graph, metabolic enzyme relationship modeling is performed, an enzyme-substrate kinetic model is constructed, and the linalool decomposition rate equation is defined as: Among them, v c represents the catalytic rate of linalool by CYP2A6, in μmol / min. The subscript c refers specifically to CYP2A6 enzyme, and k cat is the catalytic constant for the decomposition of the enzyme-substrate complex into products, in units of s -1 , [E] is the free CYP2A6 enzyme concentration, in nM, [S] is the linalool substrate concentration, in mM, K m is the Michaelis constant, which represents the affinity between enzyme and substrate, and its unit is mM. β is the factor affecting the activity of competitive inhibitor on enzyme, which is expressed as β = 0.01·e 0.5·ΔG The calculations show that [I] is the concentration of the competitive inhibitor in plasma, expressed in μM, and ΔG is the change in the binding free energy of the inhibitor to the enzyme active site, expressed in kcal / mol; Dynamically calculate vascular permeability and establish the adsorption kinetic equation for renal replacement therapy. The formula is: Among them, C eff It indicates the effective concentration of adsorption clearance during renal replacement therapy, in mg / L, C blood Indicates the real-time concentration of molecules to be cleared in the blood, in mg / L, k ad Indicates the binding rate constant of the adsorbent material to the target molecule, in min -1 , t represents the duration of a single renal replacement therapy, in min, η cr is a dynamic correction factor based on creatinine clearance, according to η cr =0.8+0.02·(CrCl-30), effective when CrCl≥30mL / min; Dynamic calculation of vascular permeability, the calculation formula is: Among them, C tis Indicates the concentration of odor molecules in the target tissue, in μg / g, P v Indicates the dynamic vascular permeability coefficient in cm / s, according to P v =0.1·CO·(1+0.05·ΔRR) is updated in real time. A represents the effective capillary exchange area in cm 2 , calculated based on ultrasound blood flow imaging data, ΔC is the concentration gradient between blood and tissue, in μg / mL / cm, τ met represents the local metabolic time constant of the tissue, measured in minutes by microdialysis technology. CO represents the real-time cardiac output, measured in L / min. ΔRR represents the deviation of the current respiratory rate from the baseline value, measured in breaths / min.

2. The artificial intelligence-based odor feedback method according to claim 1, characterized in that: The real-time physiological data set acquisition device includes: an embedded biosensor array that adheres to the outer lining of the nasal cavity and measures the rate of change of mucosal impedance; The non-contact millimeter-wave radar is installed 20 to 30 cm above the bed and detects the millimeter-level displacement of the chest.

3. The odor feedback method based on artificial intelligence according to claim 1, characterized in that: In the dynamic release strategy, olfactory sensitivity attenuation modeling was performed, and a correlation model between drug concentration and olfactory threshold was established as follows: Among them, S(t) represents the olfactory sensitivity at time t, in AU, which is calibrated by the electrophysiological signal of the olfactory bulb, S0 is the baseline sensitivity, and k decay is the neural adaptation attenuation coefficient, in min -1 , corrected to k by the bispectral index decay =6·BIS+3, C drug represents the concentration of sedative drugs in plasma in μg / mL. The pharmacokinetic curve is read from the electronic medical record system. λ represents the olfactory receptor recovery rate constant in min. -1 , which is negatively correlated with age, and the update formula is: λ = 0.02·(1-0.005·Age); To predict the risk of respiratory depression, we constructed an LSTM-ATT prediction model, which is expressed as: h t =LSTM(x t ,h t-1 ;W h ,b h ), Among them, h t Represents the hidden state at time t, dimension 128, W h represents the weight matrix (256×128), b h is the bias term, ΔRR t Indicates the time series change of respiratory rate, in times / min 2 , e t is the attention matching degree, v is the attention weight vector, 64×1, W a is the attention matrix 192×64, α t is the temporal attention weight, R risk is the risk probability of respiratory depression in the next 10-15 minutes (0-1), and σ is the Sigmoid function.

4. The artificial intelligence-based odor feedback method according to claim 1, wherein: The steps of generating the dynamic release strategy specifically include: Historical sedative drug use records were extracted from the electronic medical record system to establish a drug half-life olfactory sensitivity attenuation model; Extract features from the respiratory rate time series data based on the LSTM network to predict the risk level of respiratory depression in the next 10 to 15 minutes; When the respiratory depression risk level is detected to exceed the threshold, the odor component replacement mechanism is automatically triggered to switch the release formula to an essential oil combination that does not contain monoterpenes; The triggering conditions of the odor component replacement mechanism further include: When the real-time blood oxygen saturation decrease rate ΔSpO2 / Δt ≤ -0.5% / min, the current release pulse is immediately interrupted; After the interruption, the low-flow pure oxygen carrier gas flush is started in the expiratory phase of the first respiratory cycle, and the flush time is t = 3*RR -1 , where RR is respiratory rate, in beats / minute.

5. The odor feedback method based on artificial intelligence according to claim 1, characterized in that: The pulsed release waveform parameters include: The pulse width interval is set to [T min ,T max ], where T min ≥10 seconds and T max ≤60 seconds; Interval time dynamic adjustment coefficient α = 1 + 0.2 * current BIS value - target BIS value, where α∈[0.8,1.5]; The peak concentration constraint condition is C peak ≤0.03*CrCl+0.5, where CrCl is creatinine clearance, in mL / min.

6. The method for odor feedback based on artificial intelligence according to claim 1, wherein: In the dynamic release strategy, the pulse parameters are dynamically optimized, and the pulse parameter generation equation under the constraint conditions is established as follows: C peak =min(0.03·CrCl+0.5,0.8·S(t)·e -0.1(BIS-40) ), Among them, T width Indicates pulse width, C peak Indicates the peak concentration, T base is the reference pulse width, which is preset to 30 seconds, and clip is used to constrain the value to [T min ,T max ] is the truncation function within the interval, S(t) is the current olfactory sensitivity; Perform real-time updates of taboo mapping and construct molecular interaction constraints as follows: in, is a set of taboo molecules, is the odor molecule m i With m j The second derivative of the concentration interaction in tissues, ξ, is the liver and kidney metabolic overload threshold, in (μg / g) 2 , calculated according to the knowledge graph is ξ=2.5·(1-e 0.1·BIS ).

7. The artificial intelligence-based odor feedback method according to claim 1, characterized in that: The atomization control module execution process also includes: During the interval between adjacent pulses, reverse airflow is started, and the reverse airflow velocity V=0.5Q+2, where Q is the total amount of odor molecules released in the previous pulse stage, in μL; According to the ambient temperature and humidity sensor data, the oscillation frequency of the atomizer is dynamically adjusted to f = 200 + 10 (T-25) + 5* (H-60), where T is the temperature in ° C and H is the humidity in % RH.

8. The artificial intelligence-based odor feedback method according to claim 1, wherein: The odor feedback method also includes establishing an olfactory-visual cross-modal compensation mechanism: The eye tracking device captures the coordinates of the patient's gaze area. When the patient's gaze falls on the alarm device area for 5 or more consecutive times, the release of the menthol component is activated. When the eye movement frequency is detected to be greater than 8 Hz during rapid eye movement sleep, the release base carrier is switched to a tetradecane microcapsule suspension with phase-change cooling properties.

Images