Alzheimer's disease progression virtual simulation method and system based on exosome biomarkers

Abstract

The application provides an Alzheimer's disease course virtual simulation method and system based on an exosome biomarker, relates to the technical field of computer-aided medical treatment, and the method comprises the following steps: acquiring an exosome p-Tau181 concentration time sequence and an exosome A beta 42 oligomer concentration time sequence output by a cerebrospinal fluid exosome multi-modal combination sensor; and performing processing on the exosome p-Tau181 concentration time sequence and the exosome A beta 42 oligomer concentration time sequence. In the application, concentration transient peak point identification is used to replace smoothing processing, so that the time delay of simulation output lagging behind the actual deterioration of the disease condition is reduced.

Description

Technical Field

[0001] This invention relates to the field of computer-aided medical technology, and in particular to a virtual simulation method and system for the course of Alzheimer's disease based on exosome biomarkers. Background Technology

[0002] In clinical trials of targeted drug intervention for familial early-onset Alzheimer's disease, the existing system continuously collects cerebrospinal fluid samples from patients through an implantable cerebrospinal fluid microdialysis catheter. After the exosomes are enriched by a nanoporous exosome capture microfluidic chip, the time series of exosome p-Tau181 concentration and the time series of exosome Aβ42 oligomer concentration are detected by a surface plasmon resonance multichannel biosensor array.

[0003] In clinical drug monitoring, when abnormal pulsed discharges occur in the neurons of a patient's brain, the concentration of exosome markers should surge rapidly, forming a sharp concentration peak in the time series. The frequency and sharpness of these peaks directly reflect the intensity of neuronal release activity and are key signals for determining whether the condition is transitioning from the preclinical stage to mild cognitive impairment.

[0004] However, some existing systems with built-in static amyloid cascade reaction compartment models do not identify these spikes after receiving concentration data. Instead, they smooth the data using a fixed clearance half-life parameter. The smoothed curve weakens the spikes, turning them into a slowly changing, flat curve. Therefore, when abnormal neuronal firing has intensified and concentration spikes appear frequently, the predicted curve output by the simulation model still changes smoothly and slowly until the spikes accumulate enough to change the overall trend of the smoothed curve, at which point the model belatedly reflects the turning point of the disease. This "slow response" means that doctors adjust the dosage of targeted drugs based on model predictions later than the actual deterioration point of the disease, missing the optimal window for intervention. Summary of the Invention

[0005] This invention provides a virtual simulation method and system for the course of Alzheimer's disease based on exosome biomarkers. It replaces smoothing processing with the identification of transient concentration peaks, thereby reducing the time delay between the simulation output and the actual deterioration of the disease.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a virtual simulation method for the course of Alzheimer's disease based on exosome biomarkers is applied to a simulation server equipped with heterogeneous computing nodes, the method comprising: Acquire the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration output from cerebrospinal fluid exosome multimodal coupled sensing; For the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration, the ratio of the concentration value at each sampling point to the preset reference concentration value is calculated as the normalized concentration value. Taking each sampling point as the current point, the previous and next sampling points are obtained. On the normalized plane formed by the time coordinate and the normalized concentration value, the current point and the previous sampling point are connected to form the first line segment, and the current point and the next sampling point are connected to form the second line segment. The angle between the first and second line segments is calculated. If the included angle is less than the preset angle threshold, and the normalized concentration value of the current point is greater than the normalized concentration values ​​of the previous sampling point and the next sampling point, then the current point will be identified as the concentration transient peak point. All identified transient peak points of concentration are arranged in chronological order. The pulse release frequency sequence is extracted based on the time interval between two adjacent transient peak points. The pulse release amplitude sequence is extracted based on the concentration value of each transient peak point. At the same time, the included angle corresponding to each transient peak point of concentration is retained as a pulse release intensity measurement sequence. The pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence are used as input features and input into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease course. The dynamic simulation model obtains the virtual simulation results of the disease course based on the dynamic evolution rate of amyloid protein cascade reaction and synaptic damage driven by the pulse release features.

[0007] Secondly, a virtual simulation system for the course of Alzheimer's disease based on exosome biomarkers includes: The acquisition module is used to acquire the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration output from the cerebrospinal fluid exosome multimodal coupled sensing. The calculation module is used to calculate the ratio of the concentration value of each sampling point to the preset reference concentration value as the normalized concentration value. On the normalized plane formed by the time coordinate and the normalized concentration value, each sampling point is used as the current point. The previous sampling point and the next sampling point are obtained. The current point and the previous sampling point are connected to form a first line segment, and the current point and the next sampling point are connected to form a second line segment. The angle between the first line segment and the second line segment is calculated. The identification module is used to identify the current point as a transient peak point of concentration if the included angle is less than a preset angle threshold and the normalized concentration value of the current point is greater than the normalized concentration values ​​of the previous sampling point and the next sampling point. The extraction module is used to arrange all the identified transient peak points of concentration in chronological order, extract the pulse release frequency sequence based on the time interval between two adjacent transient peak points, extract the pulse release amplitude sequence based on the concentration value of each transient peak point, and retain the included angle corresponding to each transient peak point as a pulse release intensity measurement sequence. The driving module is used to input the pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence as input features into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease course. The dynamic simulation model obtains the virtual simulation results of the disease course based on the dynamic evolution rate of amyloid protein cascade reaction and synaptic damage driven by the pulse release features.

[0008] The above-described solution of the present invention has at least the following beneficial effects: By integrating multimodal sensors to synchronously acquire the temporal concentrations of exosome p-Tau181 and Aβ42 oligomers, the problem of parallel detection of dual biomarkers is overcome, enabling simultaneous monitoring of dual biomarkers. By calculating the angle between adjacent line segments on the normalized plane and combining the dual conditions of peak values, the transient peak point of concentration is accurately identified, overcoming the defect of existing smoothing processes that lose pulse information and cause lag in the prediction of inflection points in the course of the disease, thus achieving high-precision localization of peak events. By extracting frequency from peak intervals, amplitude from peak concentrations, and intensity from peak angles through multidimensional feature separation, the problem that single dimensions cannot depict the full picture of pulse dynamics is overcome, achieving joint characterization of release frequency, amplitude, and intensity. By using the three-channel synergistic driving mechanism of frequency regulating amyloid production, amplitude regulating Aβ oligomerization, and intensity regulating synaptic damage, the problem of static disease course models lacking dynamic regulation of neuronal activity is overcome, realizing cascade reactions driven by pulse features and dynamic simulation of synaptic damage. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the virtual simulation method for the course of Alzheimer's disease based on exosome biomarkers provided in an embodiment of the present invention.

[0010] Figure 2 This is a schematic diagram of a virtual simulation system for the course of Alzheimer's disease based on exosome biomarkers provided in an embodiment of the present invention.

[0011] Figure 3 This is a flowchart illustrating the process of identifying the current point as a transient peak point of concentration if the included angle is less than a preset angle threshold and the normalized concentration value of the current point is greater than the normalized concentration values ​​of the previous and next sampling points. Detailed Implementation

[0012] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0013] like Figures 1 to 3 As shown, embodiments of the present invention propose a virtual simulation method for the course of Alzheimer's disease based on exosome biomarkers, which is applied in clinical trials of targeted drug intervention for familial early-onset Alzheimer's disease. Subjects are patients confirmed by genetic testing to carry the presenilin 1 gene mutation. The method is applicable to all stages of the course of familial early-onset Alzheimer's disease, including the preclinical stage, the mild cognitive impairment stage, and the dementia stage. The preclinical stage refers to the stage where pathological changes such as Aβ plaque deposition and tau protein hyperphosphorylation have appeared in the brain, but the cognitive function assessment scale score has not yet reached the diagnostic criteria for mild cognitive impairment.

[0014] A cerebrospinal fluid (CSF) microdialysis catheter was implanted into the frontal horn of the patient's lateral ventricle via stereotactic surgery. The catheter's outlet end was led through a subcutaneous tunnel to the back of the ear and connected via a microfluidic conduit to the inlet of a nanoporous exosome capture microfluidic chip placed outside the patient's body. Downstream of the microfluidic chip, it was aligned with the incident light path and reflected light detector of a surface plasmon resonance (SPR) optical detection stage via an optical fiber coupler. The entire combined platform was deployed in the bedside detection unit of the clinical trial center. The ambient temperature of the detection unit was maintained within a preset temperature range by a thermostat to minimize the impact of temperature fluctuations on the SPR optical detection baseline.

[0015] The method in this embodiment runs on a simulation server equipped with heterogeneous computing nodes. These heterogeneous computing nodes are those configured with both a general-purpose central processing unit (CPU) and a dedicated parallel computing acceleration unit. The CPU is responsible for data preprocessing, logic control of the peak detection algorithm, and process scheduling; the parallel computing acceleration unit is responsible for the parallel numerical solution of the differential equations in the dynamic simulation model. The simulation server communicates with the data acquisition workstation of the joint platform via the hospital's intranet, receiving the real-time time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration from the joint platform.

[0016] The platform continuously outputs data from the p-Tau181 concentration channel and the Aβ42 oligomer concentration channel at a preset fixed sampling frequency. Within each sampling period, each channel outputs a concentration value, along with the start timestamp of that sampling period. The data acquisition workstation merges the concentration data from the two channels into a two-dimensional data stream with a unified time axis, and pushes it to the simulation server in real time via the hospital's intranet. After receiving the data stream, the simulation server appends each newly arriving sampling point to the end of the corresponding concentration time series and triggers the subsequent pulse release feature extraction process. For ease of description, the processing procedures for the exosome p-Tau181 concentration time series and the exosome Aβ42 oligomer concentration time series are described together below; the processing logic for both is identical, only the input data sources differ.

[0017] Step 1: Obtain the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration output from the cerebrospinal fluid exosome multimodal coupled sensing. The specific implementation process is as follows.

[0018] The aforementioned cerebrospinal fluid exosome multimodal sensing refers to a coupled detection platform consisting of an implantable cerebrospinal fluid microdialysis catheter, a nanoporous exosome capture microfluidic chip, and a surface plasmon resonance multichannel biosensor array connected in series. A cerebrospinal fluid microdialysis catheter is implanted in the lateral ventricle of the patient. This catheter is a double-lumen catheter with a semi-permeable membrane probe at its end. The semi-permeable membrane only allows solute molecules with a hydrodynamic radius below the cutoff threshold to pass through, thereby establishing a concentration gradient-driven material exchange interface between the cerebrospinal fluid and the perfusion fluid. The perfusion fluid is continuously pumped into the inlet chamber of the catheter at a constant flow rate. As it flows through the semi-permeable membrane region of the probe, exosomes and other small molecule solutes in the cerebrospinal fluid diffuse across the membrane into the outlet chamber of the catheter, and are continuously output to the outside with the outflow. In this embodiment, the constant flow rate is preset to 0.5 μL / min. The selection principle for this flow rate is to ensure the cerebrospinal fluid exchange efficiency while avoiding the cerebrospinal fluid dilution effect caused by excessively high perfusion flow rates, and to match the upper limit of the injection flow rate of the downstream microfluidic chip.

[0019] Cerebrospinal fluid samples containing exosomes are continuously output at a constant flow rate and directly introduced into the inlet of a nanoporous exosome capture microfluidic chip via a catheter outlet. The reason for using a constant flow rate instead of intermittent sampling is that neuronal pulse release activity is sudden and transient. If intermittent sampling is used, the concentration peak of the pulse release may be missed, resulting in the omission of subsequent peak detection.

[0020] The nanoporous exosome capture microfluidic chip integrates two levels of functional regions.

[0021] The first stage is a tunable dielectric electrophoresis electrode array. Dielectric electrophoresis refers to a technique that uses a non-uniform alternating electric field to induce directional movement of dielectric particles in a liquid. The principle is as follows: when an AC voltage is applied to the electrode pairs arranged in the microchannel, a non-uniform electric field is formed between the electrodes. Exosomes suspended in the liquid experience different electric forces than other particles due to their different dielectric properties. By adjusting the frequency of the applied voltage to a specific frequency band where the exosomes respond significantly while other particles respond weakly, the exosomes are subjected to a stronger electric force and are drawn to the capture area on the electrode surface. Most other particles, experiencing a weaker electric force, flow with the liquid through the electrode area into the waste liquid chamber. In this embodiment, the AC voltage frequency applied to the tunable dielectric electrophoresis electrode array is preset to 500 kHz. The selection of this frequency is based on the following: In preliminary experiments, exosome concentration was detected in the enriched samples within a frequency range of 10 kHz to 10 MHz. The results showed that the number of exosomes captured on the electrode surface reached its peak in the range of 500 kHz to 800 kHz, while leukocyte and erythrocyte fragments responded very weakly in this frequency range due to their dielectric relaxation frequency being below 100 kHz. In this embodiment, the lower limit of this range, 500 kHz, was selected as the operating frequency.

[0022] The second stage involves a polycarbonate nanoporous membrane, which is a polymer film with uniformly distributed cylindrical pores. In this embodiment, the pore size of the polycarbonate nanoporous membrane is 20 nm. This pore size is chosen because the typical particle size distribution of exosomes ranges from 30 nm to 150 nm. A 20 nm pore size can effectively retain the vast majority of exosomes while allowing free proteins and other particles smaller than 20 nm to pass through the membrane pores and enter the downstream waste liquid chamber. The polycarbonate nanoporous membrane is positioned on the sidewall of the microfluidic channel. The sample solution flows tangentially across the membrane surface. Exosomes are retained on the upstream side surface of the membrane, while the filtrate and molecules smaller than the pore size pass through the membrane pores and flow into the downstream waste liquid chamber. This tangential flow configuration allows the retained exosomes to form an enrichment layer on the membrane surface while the shear force of the liquid flow continuously acts on the membrane surface, preventing complete blockage of the pores. The exosome suspension, initially enriched by dielectric electrophoresis, is guided to flow across the surface of the polycarbonate nanoporous membrane, where exosomes gradually accumulate, forming an exosome-enriched sample layer.

[0023] Dielectric electrophoresis electrode arrays first aggregate exosomes dispersed in cerebrospinal fluid to the electrode surface region, achieving initial concentration of sample volume. Subsequently, the enriched exosomes are propelled by the fluid flow to the surface of a polycarbonate nanoporous membrane for secondary physical retention. Compared with relying solely on nanoporous membrane retention (processing throughput is limited by membrane pore blockage) or relying solely on dielectric electrophoresis (limited capture efficiency for extremely low concentrations of exosomes), the cascaded two-stage functional regions allow exosomes to first undergo volume concentration under the action of an electric field, and then achieve high-rate enrichment through physical sieving, thereby achieving effective capture of low-concentration exosomes while maintaining high throughput processing capacity.

[0024] After the exosome-enriched sample flows out from the downstream side of the nanoporous membrane, it enters a surface plasmon resonance (SPR) multichannel biosensor array coupled to the downstream detection chamber of the microfluidic chip. Surface plasmon resonance refers to the physical phenomenon where the intensity of reflected light decreases sharply at a specific incident angle when incident light excites collective oscillations of free electrons on the metal surface at the interface between the metal thin film and the dielectric. This specific incident angle is extremely sensitive to changes in the refractive index near the surface of the metal thin film. When biomolecules bind or dissociate on the surface of the sensing chip, the resulting local refractive index change can be detected in real time as the shift of the resonance angle.

[0025] The surface of the SPR sensor chip is pre-divided into multiple independent detection channels, with at least two detection channels modified with the following two specific capture antibodies: p-Tau181 specific capture antibody refers to a monoclonal antibody that specifically recognizes and binds to phosphorylated threonine residue 181 of the tau protein, used to capture the p-Tau181 protein marker carried on the surface of exosomes. Aβ42 oligomer specific capture antibody refers to a monoclonal antibody that specifically recognizes and binds to the conformational epitope of β-amyloid oligomers composed of 42 amino acid residues, used to capture the Aβ42 oligomer marker carried on the surface of exosomes. The term "specific recognition" means that the antibody's binding affinity to the target antigen is much higher than its binding affinity to non-target proteins, thus enriching only the target molecule in a mixed sample.

[0026] The sample enriched with exosomes flows through the surface of each detection channel at a constant flow rate. When the p-Tau181 exosome in the sample flows through the p-Tau181 channel, it is captured by the p-Tau181-specific capture antibody on the channel surface, causing a local increase in mass and a change in refractive index on the channel surface. The SPR optical system records the resonance angle shift of the channel over time in real time (i.e., the resonance angle shift sensing map). Similarly, the resonance angle shift sensing map caused by Aβ42 oligomer capture is recorded synchronously in the Aβ42 oligomer channel. The detection in each channel is independent and performed in parallel.

[0027] In actual testing, factors such as temperature fluctuations, slight changes in the composition of the perfusion fluid, and non-specific adsorption on the sensor chip surface can cause a slow baseline drift in the resonance angle offset sensor map. This baseline drift refers to the signal background trend where the resonance angle still changes slowly over time even in the absence of exosome capture events. It reflects non-specific signal fluctuations caused by environmental and instrumental factors and is unrelated to exosome capture events.

[0028] To accurately extract the response component caused by the exosome capture event from the raw sensor signal, it is necessary to first determine the specific value of the baseline drift at each sampling time and subtract it from the sensor map. The operation process consists of the following four steps: The first step is to select a baseline reference period. On the time axis of the entire resonance angle offset sensor map, identify all periods in which no exosome capture events occurred. The criteria for determining whether no capture events occurred within a period are: the resonance angle within that period does not exhibit the typical binding-dissociation response curve pattern of rising from low to high and then falling back to the baseline level, and the fluctuation amplitude of the resonance angle does not exceed the noise level of the sensor map. Typically, the periods at the beginning of detection when the exosome-containing sample has not yet flowed to the surface of the detection channel, and the periods between two adjacent capture events when the response has returned to a stable level, are considered baseline reference periods in which no capture events occurred.

[0029] The second step is to determine the baseline reference value for each baseline reference period. For each baseline reference period selected in the first step, the arithmetic mean of the resonance angle values ​​corresponding to all sampling points within that period is taken, and this arithmetic mean is used as the baseline reference value for that baseline reference period. Simultaneously, the midpoint of that period is taken as the time reference point corresponding to that baseline reference value. After this step, several sets of time reference point-baseline reference value data pairs are obtained. Each data pair represents the level at which the resonance angle should be at a specific moment, without interference from captured events.

[0030] The third step is to determine the baseline value for each sampling point. For each sampling moment in the resonance angle offset sensing map, the baseline value corresponding to that moment is determined according to the following rules: If the sampling time happens to fall within a certain baseline reference period, the baseline value at that time is directly taken as the baseline reference value corresponding to that baseline reference period.

[0031] If the sampling time falls between two adjacent baseline reference time periods, determine its baseline value one by one according to the following sub-steps: The forward time distance is obtained by calculating the difference between the time coordinate of the sampling time and the time coordinate of the time reference point of the previous baseline reference period. The total span of the reference points is obtained by calculating the difference between the time coordinate of the time reference point of the next baseline reference period and the time coordinate of the time reference point of the previous baseline reference period. Both the forward time distance and the total span of the reference points are in seconds, and their ratio is dimensionless. The ratio of the forward time distance to the total span of the reference points is calculated to obtain the time position ratio. The time position ratio is a dimensionless value between 0 and 1. When the sampling time is exactly equal to the previous time reference point, the time position ratio is 0; when the sampling time is exactly equal to the next time reference point, the time position ratio is 1; when the sampling time falls between the two, the time position ratio takes on a linear value between 0 and 1. The total difference in reference values ​​is obtained by calculating the difference between the baseline reference value corresponding to the next baseline reference period and the baseline reference value corresponding to the previous baseline reference period. The product of the total difference in reference values ​​and the time position ratio is calculated to obtain the reference value increment. The sum of the previous baseline reference value and the reference value increment is calculated, and the result is the baseline value at that sampling time. If the sampling time is before the very beginning of the entire time axis, that is, earlier than the start of the first baseline reference period, then the baseline value at that sampling time is directly taken as the baseline reference value of the first baseline reference period.

[0032] If the sampling time is located after the very end of the entire timeline, i.e. after the end of the last baseline reference period, then the baseline value at that sampling time is directly taken as the baseline reference value of the last baseline reference period.

[0033] The fourth step involves calculating the difference between the original resonance angle value and the baseline value determined in the third step for each sampling time in the resonance angle offset sensor map. This yields the baseline-removed resonance angle offset response value for that sampling time. All baseline-removed resonance angle offset response values ​​for all sampling times are arranged in chronological order, forming the baseline-removed resonance angle offset response component. This component only reflects the resonance angle change caused by the exosome capture event.

[0034] Since the p-Tau181 channel and the Aβ42 oligomer channel are independent optical detection channels, each with its own independent sensor output, the p-Tau181 resonance angle shift response and the Aβ42 oligomer resonance angle shift response are naturally separated, eliminating the need for further component separation at the signal level.

[0035] Converting the resonance angle offset into a concentration value requires relying on the resonance angle offset characteristics caused by unit concentration on the sensor chip surface, i.e., the calibration relationship. This calibration relationship refers to a standard reference describing the quantitative mapping relationship between the target concentration and the resonance angle offset, and its form is that the concentration value and the resonance angle offset satisfy a linear mapping. In this embodiment, the calibration relationship is pre-established through the following least-squares linear fitting process.

[0036] Under the same buffer conditions, several standard exosome samples with known p-Tau181 concentrations were injected sequentially, and the steady-state resonance angle offset corresponding to each sample was recorded to obtain several sets of concentration-offset data pairs.

[0037] Step 1: Take the arithmetic mean of the concentration values ​​of all data pairs, and record the result as the first mean; take the arithmetic mean of the resonance angle offset of all data pairs, and record the result as the second mean.

[0038] Step 2, for each concentration-offset data pair, perform the following operations: Calculate the difference between the concentration value of the data pair and the first mean to obtain the concentration difference of the pair; calculate the difference between the offset of the data pair and the second mean to obtain the offset difference of the pair; calculate the product of the concentration difference and itself to obtain the square of the concentration difference of the pair; calculate the product of the concentration difference and the offset difference to obtain the difference product of the pair.

[0039] Step 3: Sum the squares of the concentration differences of all data pairs to obtain the sum of the squares of the concentration differences; sum the differences of all data pairs to obtain the sum of the product of the differences.

[0040] Step 4: Calculate the quotient obtained by dividing the sum of the product of differences by the sum of the squares of the concentration differences. This quotient is the slope parameter of the calibration line. The slope parameter represents the resonance angle offset corresponding to a unit change in concentration, and its dimension is the angle offset divided by the concentration unit, i.e., degrees per nanogram per milliliter.

[0041] Step 5: Calculate the difference between the second mean and the product of the slope parameter and the first mean. This difference is the intercept parameter of the calibration line. The dimensions of the intercept parameter are the same as those of the resonance angle offset.

[0042] This determines the slope and intercept parameters, which together define a calibration line that minimizes the sum of the squares of the differences between the actual and predicted offsets of all data pairs across all possible parameter combinations. This calibration line is the concentration-resonance angle offset calibration curve.

[0043] In actual detection, for the resonance angle shift response component of the p-Tau181 channel after baseline removal, the following transformation operation is performed for each sampling time: The resonance angle shift at that time is taken, the intercept parameter is subtracted from the shift to obtain the difference, and then this difference is divided by the slope parameter. The quotient is the p-Tau181 concentration value at that time. Arranging the concentration values ​​from all sampling times in chronological order yields the exosome p-Tau181 concentration time series.

[0044] The calibration line for the Aβ42 oligomer channel was independently established using the process described above, with the injected standard sample being a standard exosome sample with a known Aβ42 oligomer concentration. During actual detection, the resonance angle offset of the Aβ42 oligomer channel at each sampling time was converted into Aβ42 oligomer concentration values ​​in the same manner, and arranged chronologically to obtain the time series of exosome Aβ42 oligomer concentrations.

[0045] The concentration time series refers to a numerical sequence formed by arranging concentration values ​​measured at consecutive equally spaced sampling times in chronological order. Each element of the sequence contains two attributes: a sampling timestamp and the corresponding concentration value.

[0046] Step 2: For the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration, calculate the ratio of the concentration value at each sampling point to the preset reference concentration value, and use it as the normalized concentration value. Sequentially, take each sampling point as the current point, obtain the previous sampling point and the next sampling point of the current point, connect the current point and the previous sampling point to form the first line segment, and connect the current point and the next sampling point to form the second line segment. Calculate the angle between the first line segment and the second line segment. The specific implementation process is as follows.

[0047] For the p-Tau181 concentration time series and the Aβ42 oligomer concentration time series, the following normalization operation was performed. The median of the concentration values ​​at all sampling points in the concentration time series was taken as the reference concentration value for the series, with units of ng / mL. For each sampling point in the concentration time series, the concentration value at that point was divided by the reference concentration value of the series; the quotient was the normalized concentration value for that sampling point. The normalized concentration value is a dimensionless value, and its physical meaning is the multiple of the concentration at that sampling point relative to the typical concentration level of the series. The median, rather than the arithmetic mean, was used as the reference concentration value because the median is not sensitive to individual extreme peaks and can more stably reflect the baseline level of the concentration series, avoiding the problem of an overestimation of the reference concentration due to individual extremely high pulse peaks, which in turn leads to an underestimation of the overall normalized concentration value.

[0048] After normalization preprocessing, each sampling point in the concentration time series has two attributes: a time coordinate and a normalized concentration value. The time coordinate refers to the offset of the sampling time of that sampling point relative to the start time of the concentration time series, in seconds. The normalized concentration value is the quotient obtained by dividing the concentration value of that sampling point by the reference concentration value.

[0049] Establish a normalized plane. The normalized plane refers to a two-dimensional rectangular coordinate system with the normalized time coordinate as the horizontal axis and the normalized concentration value as the vertical axis. The normalized time coordinate is determined as follows: Divide the time coordinate of each sampling point by the sampling period of the concentration time series. The sampling period refers to the fixed time interval between two adjacent sampling points in the series, in seconds. Since both the time coordinate and the sampling period are in seconds, the normalized time coordinate obtained after division is a dimensionless pure numerical value, which is equal to the index number of the sampling point in the series, starting from zero. For example, the normalized time coordinate of the first sampling point in the series is zero, the normalized time coordinate of the second sampling point is one, and so on.

[0050] The normalized concentration value is the ratio of the concentration value at that sampling point to the reference concentration value. The index number, rather than the actual time, is used as the x-axis because subsequent angle calculations involve geometric measurements of the x and y axes. If the actual time were used as the x-axis, the same morphological curve at different sampling frequencies would have different angle values ​​due to time axis scaling, causing the peak identification result to depend on the sampling frequency, a non-physiological parameter. Using the index number as the x-axis ensures that the x-coordinates of all sampling points are evenly distributed, and the angle size is entirely determined by the morphology of the concentration value change, eliminating the influence of sampling frequency differences on the identification result.

[0051] Therefore, the values ​​represented by the horizontal and vertical axes of the normalized plane are dimensionless pure numerical values. When calculating the distance between any two points or the angle between any two line segments on the normalized plane, the spans of the horizontal and vertical axes involved in the calculation are dimensionless numbers. The calculated distances and angles are not affected by the absolute value of the original sampling interval, maintaining consistency across multiple devices with different sampling frequencies. After adopting the normalized plane, the time series acquired by each device at different sampling frequencies are mapped to a unified index number-normalized concentration space. The calculation of the cosine value of the angle is not affected by the absolute value of the original sampling interval, ensuring the comparability of peak identification results across different acquisition devices.

[0052] Each sampling point in the concentration time series corresponds to a unique location on the normalized plane. The x-coordinate of this location is equal to the normalized time coordinate of the sampling point, and the y-coordinate is equal to the normalized concentration value of the sampling point. Starting from the first sampling point of the concentration time series, each sampling point is taken as the current sampling point in chronological order, and subsequent angle calculation and peak determination operations are performed on the current sampling point. For each current sampling point: The preceding sampling point is the sampling point in the concentration time series that is closest to the current sampling point in terms of sampling time. The following sampling point is the sampling point that is closest to the current sampling point in terms of sampling time in terms of sampling time. Since there are no sampling points before the first sampling point in the concentration time series, the first sampling point is skipped and does not participate in the angle calculation or peak identification. Similarly, since there are no sampling points after the last sampling point in the concentration time series, the last sampling point is also skipped. The angle calculation process only proceeds when both a preceding and following sampling point exist for the current sampling point.

[0053] On the normalized plane, for the three location points corresponding to the current sampling point, the previous sampling point, and the next sampling point, construct two line segments according to the following operations: Connect the current sampling point with the previous sampling point using a straight line segment; this segment is called the first line segment. Similarly, connect the current sampling point with the next sampling point using a straight line segment; this segment is called the second line segment. The first and second line segments intersect at the current sampling point, which is the common endpoint of the two line segments. The two line segments form an angle at the current sampling point, the size of which reflects the degree of curvature of the normalized concentration time series at that point.

[0054] To quantitatively determine the degree measure of the angle between the first and second line segments, calculate step by step as follows.

[0055] Sub-step a: Determine the span of the first line segment in the normalized time direction and the normalized concentration direction.

[0056] Take the normalized time coordinate of the current sampling point and the normalized time coordinate of the previous sampling point, and calculate the difference between these two normalized time coordinates to obtain the normalized time span of the first line segment. Take the normalized concentration value of the current sampling point and the normalized concentration value of the previous sampling point, and calculate the difference between these two normalized concentration values ​​to obtain the normalized concentration span of the first line segment.

[0057] Sub-step b: Determine the span of the second line segment in the normalized time direction and the normalized concentration direction.

[0058] Take the normalized time coordinate of the next sampling point and the normalized time coordinate of the current sampling point, and calculate the difference between these two normalized time coordinates to obtain the normalized time span of the second line segment. Take the normalized concentration value of the next sampling point and the normalized concentration value of the current sampling point, and calculate the difference between these two normalized concentration values ​​to obtain the normalized concentration span of the second line segment.

[0059] Sub-step c: Calculate the product of the normalized time span of the first line segment and itself to obtain the first time squared term. Calculate the product of the normalized concentration span of the first line segment and itself to obtain the first normalized concentration squared term. Calculate the sum of the first time squared term and the first normalized concentration squared term to obtain the first sum of squares. Calculate the square root of the first sum of squares to obtain the length of the first line segment.

[0060] Sub-step d: Calculate the product of the normalized time span of the second line segment and itself to obtain the second time squared term. Calculate the product of the normalized concentration span of the second line segment and itself to obtain the second normalized concentration squared term. Calculate the sum of the second time squared term and the second normalized concentration squared term to obtain the second sum of squares. Calculate the square root of the second sum of squares to obtain the length of the second line segment.

[0061] Sub-step e: Calculate the inner product of the direction vectors of the two line segments.

[0062] Calculate the product of the normalized time span of the first line segment and the normalized time span of the second line segment to obtain the first time product term. Calculate the product of the normalized concentration span of the first line segment and the normalized concentration span of the second line segment to obtain the first normalized concentration product term. Calculate the sum of the first time product term and the first normalized concentration product term to obtain the direction vector inner product.

[0063] Sub-step f: Calculate the product of the length of the first line segment and the length of the second line segment to obtain the length product.

[0064] Sub-step g: Calculate the cosine of the included angle.

[0065] The ratio of the dot product of the direction vectors to the product of their lengths is used to obtain the cosine value of the included angle. This cosine value is the cosine function of the angle between the first and second line segments, and its value ranges from -1 to 1. The inverse cosine function of the cosine value is then calculated to obtain the angle between the first and second line segments. There is a monotonic correspondence between the cosine value and the magnitude of the included angle. When the cosine of the included angle is 1, the two line segments are in the same direction and the included angle is 0. 0 This corresponds to the case where the three sampling points are completely collinear on the normalization plane and the concentration changes monotonically in the same direction; when the cosine of the included angle is 0, the two line segments are perpendicular to each other, and the included angle is 90°. 0 When the cosine of the included angle approaches -1, the directions of the two line segments are almost completely opposite, and the included angle approaches 180°.0 This corresponds to the situation where the concentration first rises and then falls, forming a smooth arc transition at the current sampling point. In the case of a concentration peak, the normalized concentration value at the current sampling point is much higher than that of the two sampling points before and after it. The first line segment points diagonally upward from the lower concentration to the concentration peak, and the second line segment points diagonally downward from the concentration peak to the lower concentration, forming a sharp acute angle between the two line segments. Because the sharper the concentration peak, the greater the difference in direction between the two line segments, and the smaller the relative dot product of the direction vectors, the smaller the included angle.

[0066] It should be noted that if the first or second line segment length calculated in sub-step c or sub-step d is 0, meaning the position of the previous or next sampling point completely coincides with the current sampling point, then the current sampling point will not participate in the angle calculation and will be skipped directly. This situation will not occur in normal equally spaced sampling and is only used as a protection measure when data is abnormal.

[0067] Step 3: Identify the transient peak point of concentration based on the angle and concentration comparison results. The specific operation is as follows.

[0068] After calculating the cosine of the angle between the first and second line segments at the current sampling point, determine whether the current sampling point is a transient concentration peak point according to the following rules.

[0069] Take the concentration value of the current sampling point, the concentration value of the previous sampling point, and the concentration value of the next sampling point. Compare the concentration value of the current sampling point with the concentration values ​​of the previous and next sampling points one by one: If the concentration value of the current sampling point is greater than the concentration value of the previous sampling point, and the concentration value of the current sampling point is also greater than the concentration value of the next sampling point, then the current sampling point meets the concentration peak condition. If the concentration value of the current sampling point is not greater than the concentration value of the previous sampling point, or the concentration value of the current sampling point is not greater than the concentration value of the next sampling point, then the current sampling point does not meet the concentration peak condition and is directly determined as a non-concentration transient peak point, and the angle condition is no longer determined.

[0070] The smaller the angle, the more abrupt the change in the concentration-time curve at the current sampling point, resulting in a sharper concentration peak; conversely, the larger the angle, the smoother the curve change. Therefore, an angle less than a preset angle threshold is used as a condition for identifying sharp transient concentration peaks.

[0071] The specific degree of the preset angle threshold. The selection criteria for the angle threshold are: the smaller the angle threshold, the sharper and fewer the concentration peaks identified; the larger the angle threshold, the smoother and more numerous the concentration peaks identified. The angle threshold is set at 30 degrees. 0 Up to 60 0 Between, the 30 0 Up to 60 0The range was determined based on statistical analysis of the pre-experimental pulse peak angle distribution in multiple cases of familial early-onset AD patients, with the angle of true positive pulse peaks concentrated at 10°. 0 Up to 40 0 The angles of the pseudo-peaks in the baseline fluctuations are concentrated at 65°. 0 Up to 85 0 Therefore, the threshold is set within the interval between the two distributions. Within this range, the angle threshold can effectively distinguish true pulse release events (angle approximately 10°). 0 Up to 40 0 ) and spurious peaks formed by random fluctuations in concentration (the angle between them is usually greater than 60 degrees) 0 ).

[0072] Since the cosine function is at 0 0 Up to 180 0 Within the range, the cosine value is monotonically decreasing. The larger the angle, the smaller the cosine value, and the smaller the angle, the larger the cosine value. Therefore, the included angle being less than the preset angle threshold is equivalent to the included angle cosine value being greater than the included angle cosine judgment threshold.

[0073] Compare the cosine value of the included angle calculated in sub-step g with the cosine threshold for judging the included angle: If the cosine value of the included angle is greater than the cosine threshold, the included angle is determined to meet the sharpness condition; if the cosine value of the included angle is not greater than the cosine threshold, the included angle is determined not to meet the sharpness condition, and the current sampling point is determined to be a non-concentration transient peak point.

[0074] The current sampling point is marked as a concentration transient peak point if and only if both of the following conditions are met: The concentration value at the current sampling point is greater than both the concentration values ​​at the previous and next sampling points; the cosine value of the angle calculated at the current sampling point is greater than the cosine threshold for angle determination, meaning the angle between the first and second line segments is less than a preset angle threshold. This sampling point is marked as a transient peak concentration point.

[0075] For each sampling point in the concentration time series that satisfies the condition of having adjacent sampling points, the above operations are repeated until the penultimate sampling point of the concentration time series is processed, thus obtaining all the transient peak points of the concentration time series and their associated three pieces of information. The above operations are performed independently for the p-Tau181 concentration time series and the Aβ42 oligomer concentration time series, respectively, and each set of sequences generates an independent set of transient peak points of concentration.

[0076] Step 4: Arrange all the identified transient peak points of concentration in chronological order, extract the pulse release frequency sequence based on the time interval between two adjacent transient peak points, extract the pulse release amplitude sequence based on the concentration value of each transient peak point, and retain the included angle corresponding to each transient peak point as the pulse release intensity measurement sequence. The specific implementation process is as follows.

[0077] Extract the time coordinate of each concentration transient peak point from the set of concentration transient peak points, and arrange all concentration transient peak points in ascending order of their time coordinates. The peak point with the smallest time coordinate, i.e., the earliest occurrence at the sampling time, is placed at the beginning; the peak point with the largest time coordinate, i.e., the latest occurrence at the sampling time, is placed at the end. The ordered list after this arrangement is called the peak point sequence. The concentration transient peak points at the beginning of the peak point sequence correspond to earlier neuronal pulse release events, and those at the end correspond to later pulse release events.

[0078] Extract the pulse release frequency sequence, as follows.

[0079] The pulse release frequency sequence is composed of the time interval between adjacent transient peak points of concentration. The smaller the time interval, the more concentrated the pulse release and the higher the frequency. Starting from the first transient peak point of the peak point sequence, two adjacent transient peak points are combined into an adjacent peak point pair. For a peak point sequence containing N transient peak points of concentration, a total of N-1 adjacent peak point pairs are formed.

[0080] Perform the following operations sequentially on each pair of adjacent peak points: Extract the time coordinates of the latter transient peak point and the former transient peak point from the adjacent peak point pair; calculate the difference between the latter and former time coordinates to obtain the pulse release cycle interval corresponding to the adjacent peak point pair. The smaller the pulse release cycle interval, the shorter the time distance between two adjacent pulse releases, and the more frequent the pulse releases. Arrange the obtained N-1 pulse release cycle intervals in order of appearance of each adjacent peak point pair in the peak point sequence to form a pulse release frequency sequence. The k-th element in the sequence corresponds to the time interval between the k-th and (k+1)-th transient peak points. If the peak point sequence contains only one transient peak point, there are no adjacent peak point pairs, no pulse release cycle interval is generated, and the pulse release frequency sequence is an empty sequence.

[0081] For each transient peak in the peak point sequence, the determined concentration value of that transient peak is extracted and used as the corresponding peak concentration value. The peak concentration value is a measure of the highest level of exosome marker concentration increase during a single neuronal pulse release event; the higher the concentration value, the higher the peak concentration of that pulse release event. The peak concentration values ​​corresponding to the N transient peaks are arranged sequentially according to the order in which they appear in the peak point sequence, forming a pulse release amplitude sequence. The k-th element in the sequence corresponds to the peak concentration value of the k-th transient peak.

[0082] If the peak point sequence is empty and the pulse release intensity measurement sequence is empty, the following traversal operation will not be performed.

[0083] For each transient concentration peak in the peak point sequence, the angle between the first and second line segments calculated at the time of identification is extracted, and this angle is used as the pulse release sharpness metric for that transient concentration peak. The pulse release sharpness metric is a morphological index that characterizes the sharpness of the bend in the concentration-time curve of the transient concentration peak using an angle value. The smaller the angle, the more abrupt the transition between the concentration rise and fall segments, the sharper the concentration peak, and the higher the corresponding neuronal pulse release intensity. The pulse release sharpness metrics corresponding to N transient concentration peaks are arranged sequentially according to the order in which they appear in the peak point sequence, forming a pulse release intensity metric sequence. The k-th element in the sequence corresponds to the pulse release sharpness metric of the k-th transient concentration peak.

[0084] There is a definite temporal correspondence between the pulse release frequency sequence, the pulse release amplitude sequence, and the pulse release intensity measurement sequence: The k-th element of the pulse release frequency sequence corresponds to the time interval between the k-th concentration transient peak and the (k+1)-th concentration transient peak; the k-th element of the pulse release amplitude sequence corresponds to the peak concentration value of the k-th concentration transient peak; and the k-th element of the pulse release intensity measurement sequence corresponds to the pulse release sharpness measurement value of the k-th concentration transient peak.

[0085] The above operations were performed independently on the p-Tau181 concentration time series and the Aβ42 oligomer concentration time series. Each of the two channels outputs a set of pulse release frequency sequences, a set of pulse release amplitude sequences, and a set of pulse release intensity measurement sequences, for a total of six sequences.

[0086] Step 5: The pulse release interval sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence are used as input features and input into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease course. The dynamic simulation model is based on the pulse release feature-driven dynamic evolution rate of amyloid cascade reaction and synaptic damage to obtain the virtual simulation results of the disease course. The specific implementation process is as follows.

[0087] For each concentration channel, elements within the same time window are extracted from its pulse release interval sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence to construct a multidimensional input feature set for that time window.

[0088] The preset time window width and sliding step size are both in seconds. The preset window width is 180 seconds, and the preset sliding step size is 60 seconds. The selection of the 180-second window width is based on autocorrelation analysis of cerebrospinal fluid exosome concentration monitoring data from 5 familial early-onset AD patients over 24 hours. The results showed that the average interval between concentration pulse events was approximately 45 seconds. The 180-second window width can cover approximately 4 consecutive pulse events, which is sufficient to statistically analyze the frequency and amplitude distribution characteristics of pulse release. The 60-second step size ensures that adjacent windows overlap for 120 seconds (overlap rate of approximately 67%), maintaining the temporal continuity of the simulation output while avoiding missed pulse events due to window boundary truncation. The window width is an integer multiple of the sampling interval and covers the time span of at least 3 to 5 typical pulse release events. The step size is approximately one-third of the window width, ensuring that adjacent windows overlap by two-thirds, achieving a balance between simulation response speed and output smoothness. The sliding step size is smaller than the window width. The time window refers to a time interval defined by tracing back one window width from the current simulation moment, i.e., the time period ending at the current simulation moment and starting at the difference between the current simulation moment and the window width. The window width determines the time range of historical pulse events included in each simulation calculation. An excessively large width causes the simulation response to lag behind current disease changes, while an excessively small width results in insufficient available pulse event samples and increased fluctuations in simulation statistics. Starting from the monitoring start moment, the first time window is extracted using the window width. Subsequently, the window slides forward along the time axis, forming a new window with each slide of one step. Because the step size is less than the window width, there is an overlapping area between adjacent windows, with the overlap length equal to the difference between the window width and the step size. This overlapping design ensures a smooth temporal transition between the outputs of two adjacent simulation moments, avoiding abrupt changes in simulation results due to sudden changes in window boundaries. For each time window, the following three extraction operations are performed.

[0089] The first extraction step is the pulse release interval subsequence. From the pulse release interval sequence, select all elements that meet the following conditions: the normalized time coordinate of the adjacent peak point pair corresponding to the element falls within the time window. Arrange these selected time interval elements in the order they appear in the original sequence to form the pulse release interval subsequence for this window.

[0090] The second extraction step is the pulse release amplitude subsequence. From the pulse release amplitude sequence, select all elements that meet the following condition: the normalized time coordinate of the concentration transient peak point corresponding to the element falls within the time window. Arrange these selected peak concentration values ​​in chronological order of their corresponding peak points to form the pulse release amplitude subsequence for this window.

[0091] The third extraction step is the pulse release intensity measurement subsequence. From the pulse release intensity measurement sequence, all elements that meet the following condition are selected: the normalized time coordinate of the concentration transient peak point corresponding to the element falls within the time window. These selected sharpness measurement values ​​are arranged in chronological order of their corresponding peak points to form the pulse release intensity measurement subsequence for this window. Each element in this subsequence is a pulse release sharpness measurement value, and both refer to the same physical quantity.

[0092] The above three sub-sequences are combined to form the multidimensional input feature set of this window. The multidimensional input feature set refers to a three-dimensional feature array composed of three pulse release feature sub-sequences within the same time window and the same concentration channel. Its three dimensions describe the time interval distribution characteristics of pulse release, the concentration intensity characteristics of pulse release, and the sharpness characteristics of pulse release within the window, respectively.

[0093] If a subsequence is empty, that feature dimension takes a null value in the current window. For feature dimensions with null values, the corresponding rate adjustment path is not updated; the rate output of this path uses the calculation result of the previous time window until the next non-empty window appears. For the simulation's starting window, if a subsequence is empty, the corresponding rate adjustment path uses the preset baseline rate. The above operations are performed independently for the p-Tau181 channel and the Aβ42 oligomer channel, with each channel generating one multidimensional input feature set in each time window.

[0094] The pulse-release feature-driven dynamic simulation model of Alzheimer's disease progression is a computational model built upon the amyloid cascade hypothesis. The model uses three state variables: an Aβ monomer accumulation pool, an Aβ oligomer accumulation pool, and a synaptic damage accumulation index. All three are numerical quantities that are continuously updated as the simulation time window progresses. The model uses statistics from a multidimensional input feature set as time-varying input parameters, with each time window as a step unit, calculating the disease progression trajectory through discrete-time steps. Internally, the model defines three rate-regulating pathways, each corresponding to one pulse-release feature dimension, regulating one rate of disease progression.

[0095] Before describing the three rate regulation pathways, the preset parameters involved in the pathways are explained in a unified manner as follows.

[0096] The baseline cleavage rate refers to the basal rate at which β-secretase and γ-secretase cleave amyloid precursors to produce Aβ monomers in a resting state without abnormal neuronal impulse release activity. This parameter is taken as the mass of newly generated Aβ monomers per milliliter of cerebrospinal fluid per second from the cleavage of amyloid precursors. In this embodiment, this parameter is taken from published literature data on the metabolic flux of amyloid precursors measured by a stable isotope labeling kinetic method in a cohort of patients with familial early-onset Alzheimer's disease, and is taken as 0.12 ng / (mL·s).

[0097] The cutting ratio constant is a preset proportional coefficient used to map the intensity of pulse release to a cutting rate adjustment factor, and its dimension is seconds. The more intense the pulse release and the smaller the equivalent average interval, the larger the ratio of the cutting ratio constant to the equivalent average interval, and the greater the cutting rate factor, thus driving the cutting rate to increase proportionally from the baseline cutting rate. In this embodiment, the cutting ratio constant is exemplarily set to 0.8s.

[0098] The baseline polymerization rate refers to the basic rate at which Aβ monomers spontaneously polymerize into soluble oligomers under physiological conditions without concentration pulse perturbation. This parameter is taken as the mass of oligomers formed per second from monomer polymerization per milliliter of cerebrospinal fluid sample. In this embodiment, this parameter is taken from the initial polymerization rate data determined by in vitro aggregation kinetics experiments of Aβ42 in published literature, and is taken as 0.05 ng / (mL·s).

[0099] The unit oligomer damage coefficient refers to the baseline amount of synaptic damage caused per nanogram of Aβ oligomer per milliliter of cerebrospinal fluid per second. In this embodiment, the value is taken as 0.03 (ng / mL). -1 ·s -1 .

[0100] The Aβ monomer clearance rate constant characterizes the rate at which Aβ monomers are cleared from the brain parenchyma via cerebrospinal fluid circulation and degradation by proteases such as insulin-degrading enzymes and enkephalins within the brain parenchyma, and its dimension is per second. In this embodiment, the value is taken as 1.2 × 10⁻⁶. -4 s -1 This value was determined by referencing the half-life data of Aβ monomers obtained from brain parenchymal microdialysis combined with radioactive tracing in a transgenic mouse model of Alzheimer's disease, and the corresponding clearance rate constant was calculated from the half-life.

[0101] The Aβ oligomer clearance rate constant characterizes the rate at which soluble Aβ oligomers are cleared from the brain parenchyma via microglia phagocytosis, astrocyte uptake, and enzymatic degradation, and its dimension is per second. In this embodiment, the value is taken as 8.0 × 10⁻⁶. -5 s -1 .

[0102] The synaptic repair rate constant characterizes the rate at which the damaged postsynaptic membrane recovers its function through endogenous repair mechanisms such as receptor recirculation and synaptic scaffold protein recombination, and its dimension is per second. In this embodiment, an exemplary value of 5.0 × 10⁻⁶ is used. -5 s -1 .

[0103] The first rate regulation pathway regulates the cleavage rate of amyloid precursor. This pathway takes the pulse release time interval subsequence as input and drives the rate at which amyloid precursor is cleaved to produce Aβ monomers. The operation process is as follows.

[0104] Take the reciprocal of each time interval element in the pulse release time interval subsequence to obtain the reciprocal value of each element.

[0105] Calculate the arithmetic mean of all reciprocals to obtain the average reciprocal value; calculate the quotient of 1 divided by the average reciprocal value to obtain the equivalent average interval, in seconds. The smaller the equivalent average interval, the more concentrated the pulse release within the window; calculate the ratio of the cutting ratio constant to the equivalent average interval to obtain the cutting rate factor. The cutting rate factor is a dimensionless rate adjustment coefficient whose value dynamically changes according to the equivalent average interval of pulse release; the more concentrated the pulses, the smaller the equivalent average interval, and the larger the cutting rate factor. Calculate the product of the baseline cutting rate and the cutting rate factor to obtain the actual cutting rate of the window.

[0106] The biological logic of this pathway is as follows: when neuronal impulse release activity is enhanced, the presynaptic membrane depolarization frequency increases, calcium ion influx increases, and the cleavage activity of β-secretase and γ-secretase on amyloid precursors is activated. The more intensive the impulse release, the higher the rate at which amyloid precursors are cleaved to produce Aβ monomers.

[0107] The second rate regulation pathway regulates the polymerization rate of Aβ monomers. This pathway uses the pulse release amplitude subsequence as input to drive the polymerization rate of Aβ monomers into oligomers, and the operation process is as follows.

[0108] Calculate the arithmetic mean of all peak concentration values ​​in the pulse release amplitude subsequence to obtain the window average peak concentration value, in ng / mL.

[0109] Secondly, a reference concentration value is determined by taking the median of the concentration values ​​at each sampling point in the concentration time series. The ratio of the average peak concentration value of the window to this reference concentration value is calculated to obtain the polymerization rate factor. The polymerization rate factor is a dimensionless rate regulation coefficient whose value dynamically changes according to the concentration intensity of the pulse release event within the window; the higher the peak concentration, the larger the polymerization rate factor. The reference concentration value mentioned here is the same physical quantity as the reference concentration value in step 2, namely the median of the concentration values ​​at all sampling points in the concentration time series, in ng / mL. The product of the baseline polymerization rate and the polymerization rate factor is calculated to obtain the actual polymerization rate of the window. This pathway is established based on the law of mass action: the higher the peak concentration of exosome Aβ42, the stronger the activity of neuronal Aβ42 release, and the higher the instantaneous concentration of free Aβ42 monomers in the synaptic cleft. High monomer concentrations drive faster oligomer nucleation and elongation rates.

[0110] The third rate regulation pathway modulates the synaptic damage rate. This pathway takes the pulse release intensity quantum sequence as input and drives the damage rate, which is exacerbated by the accumulation of Aβ oligomers. The operation process is divided into two stages.

[0111] In the first stage, the current value of the Aβ oligomer accumulation pool at the end of the previous window is taken. This value is then multiplied by the unit oligomer damage coefficient to obtain the basic synaptic damage rate. The unit oligomer damage coefficient is a preset synaptic toxicity conversion constant, whose function is to linearly map the Aβ oligomer accumulation to the basic damage rate.

[0112] In the second stage, the acceleration factor is identified and the accelerated damage rate is calculated. One damage acceleration factor is determined based on the pulse release sharpness characteristics. The product of the baseline synaptic damage rate and this acceleration factor is calculated to obtain the accelerated synaptic damage rate. The larger the acceleration factor, the stronger the additional amplification effect applied to the baseline damage rate. The pathophysiological basis of this pathway is that a sharp concentration spike indicates a large influx of Aβ42 oligomers into the synaptic cleft within a very short time, overwhelming the postsynaptic membrane's clearance and repair mechanisms. The sharper the pulse, the higher the actual damage rate of the synapse by Aβ oligomers. The endocytosis-inducing effect of Aβ42 oligomers on NMDA and AMPA receptors on the postsynaptic membrane has a concentration threshold effect: when the oligomer concentration exceeds a certain threshold within a short time, the receptors are rapidly endocytosed from the membrane surface into the cell, the postsynaptic membrane receptor density decreases sharply, and the excitatory synaptic transmission efficiency exhibits a non-linear accelerated decay. The sharper the pulse release, the shorter the duration for which the Aβ42 oligomer concentration exceeds the threshold, but the higher the instantaneous concentration. Once the receptor is endocytosed in a large quantity at once, it is difficult to replenish rapidly through recycling, resulting in a stepped amplification of the damage effect. In this model, a sharpness threshold grading and two-level acceleration factors are used to characterize the nonlinear effect of this concentration threshold.

[0113] The specific procedures for determining the synaptic damage acceleration factor are as follows.

[0114] In the third rate regulation pathway, the synaptic lesion rate depends not only on the accumulation of Aβ oligomers but also on a graded acceleration regulation of the pulse release sharpness metric. The operation consists of three steps.

[0115] Step 1: For each time window, extract the pulse release intensity metric subsequence for that window. Calculate the arithmetic mean of all pulse release sharpness metrics in that subsequence to obtain the window average sharpness metric. The smaller the window average sharpness metric, the higher the average sharpness of the pulse release events within that window.

[0116] Step 2 compares the average sharpness metric of the window with a preset sharpness threshold. The preset sharpness threshold is 25°. Statistical analysis of the angles of pulse-release events at baseline in preclinical patients shows that approximately 85% of pulse-release events have an angle between 25° and 60°, while pulse-release events with angles less than 25° account for only about 15% of all events and mostly occur during the critical transition period when cognitive function begins to decline. The sharpness threshold refers to the angle boundary value used to distinguish between high-sharpness and low-sharpness pulse-release windows; the smaller the angle, the sharper the concentration peak. The sharpness level judgment result is obtained after comparison. If the average sharpness metric of the window is less than the sharpness threshold, meaning that the pulse release within the window is sharper overall, then the sharpness level is judged as high sharpness, and the time window is determined as a high-sharpness pulse release window.

[0117] If the average sharpness metric of the window is not less than the sharpness threshold, the sharpness level is judged as low sharpness, and the time window is determined as a low sharpness pulse release window.

[0118] Step 3: The corresponding acceleration factor is applied based on the sharpness level assessment. Two synaptic damage acceleration factors are preset: the first acceleration factor corresponds to the high-sharpness pulse release window and is preset to 3.0; the second acceleration factor corresponds to the low-sharpness pulse release window and is preset to 1.0. The difference between the two factors is determined statistically based on the ratio of synaptic damage rates between the high-sharpness window (angle < 25°) and the low-sharpness window (angle ≥ 25°) in preliminary experiments. The value of the first acceleration factor is greater than the value of the second acceleration factor. Both acceleration factors are dimensionless constants greater than or equal to 1. A value of 1 indicates no additional acceleration effect, while a value greater than 1 indicates an additional acceleration amplification applied on top of the baseline damage rate. The first acceleration factor is applied for the high-sharpness window, and the second acceleration factor is applied for the low-sharpness window.

[0119] Multiply the baseline synaptic damage rate calculated in the first stage of the third rate regulation pathway by the acceleration factor determined in the second stage to obtain the accelerated synaptic damage rate, which is the actual rate that drives the current window of synaptic damage to worsen.

[0120] Over consecutive time window sequences, the model drives the dynamic evolution of the disease course in parallel through the three rate-regulating pathways mentioned above. The following six simulation calculations are performed within each time window.

[0121] Step 1: Determine the rates. Based on the three pulse release characteristic subsequences of this window, adjust the pathways according to the aforementioned rates 1 to 3 respectively to determine the actual cutting rate, actual polymerization rate, and accelerated synaptic damage rate of this window.

[0122] Step 2: Determine the new increment in each accumulation pool within this window. Calculate the product of the actual cutting rate and the window width to obtain the Aβ monomer generation amount within this window. Calculate the product of the actual polymerization rate and the window width to obtain the Aβ monomer polymerization consumption amount within this window. Calculate the product of the accelerated synaptic damage rate and the window width to obtain the new synaptic damage amount within this window. The window width refers to the preset window time length, in seconds.

[0123] Step 3: Update the Aβ monomer accumulation pool. Calculate the product of the Aβ monomer accumulation pool value at the end of the previous window, the Aβ monomer clearance rate constant, and the window width to obtain the natural clearance amount of Aβ monomers in the current window. Calculate the sum of the Aβ monomer accumulation pool value at the end of the previous window and the Aβ monomer generation amount in the current window to obtain the first intermediate value. Calculate the difference between the first intermediate value and the Aβ monomer polymerization consumption amount in the current window to obtain the second intermediate value. Calculate the difference between the second intermediate value and the natural clearance amount to obtain the third intermediate value. If the third intermediate value is negative, use 0 as the Aβ monomer accumulation pool value at the end of the current window; otherwise, use the third intermediate value as the Aβ monomer accumulation pool value at the end of the current window. The Aβ monomer accumulation pool is a state variable of the model, and its value is the difference between the Aβ monomer generation amount and the sum of polymerization consumption and clearance amounts since the start of the simulation, reflecting the dynamic equilibrium state of amyloid protein generation and clearance.

[0124] Step 4: Update the Aβ oligomer accumulation pool. Calculate the product of the Aβ oligomer accumulation pool value at the end of the previous window, the Aβ oligomer clearance rate constant, and the window width to obtain the natural clearance amount of Aβ oligomers in the current window. Take the Aβ monomer polymerization consumption in the current window. If this value is greater than the Aβ monomer accumulation pool value at the end of the previous window (subtracted from Step 3), then use this accumulation pool value as the actual polymerization consumption; otherwise, use the Aβ monomer polymerization consumption in the current window as the actual polymerization consumption. Calculate the sum of the Aβ oligomer accumulation pool value at the end of the previous window and the actual polymerization consumption to obtain the fourth intermediate value. Calculate the difference between the fourth intermediate value and the natural clearance amount to obtain the fifth intermediate value. If the fifth intermediate value is negative, use 0 as the Aβ oligomer accumulation pool value at the end of the current window; otherwise, use the fifth intermediate value as the Aβ oligomer accumulation pool value at the end of the current window. The Aβ oligomer accumulation pool is another state variable in the model. Its value is the difference between the amount of polymers generated and the amount of oligomers removed since the start of the simulation, reflecting the accumulation level of toxic oligomers.

[0125] Step 5: Update the cumulative synaptic damage index. Calculate the product of the cumulative synaptic damage index value at the end of the previous window, the synaptic repair rate constant, and the window width to obtain the synaptic repair amount for the current window. Calculate the sum of the cumulative synaptic damage index value at the end of the previous window and the new synaptic damage amount to obtain the sixth intermediate value. Calculate the difference between the sixth intermediate value and the synaptic repair amount to obtain the seventh intermediate value. If the seventh intermediate value is negative, use 0 as the cumulative synaptic damage index value at the end of the current window; otherwise, use the seventh intermediate value as the cumulative synaptic damage index value at the end of the current window. The cumulative synaptic damage index is the third state variable of the model. Its value is the difference between the new damage amount and the repair amount since the start of the simulation, characterizing the cumulative loss of synaptic function and reflecting the progression of neurodegenerative diseases.

[0126] Step 6: Output the current disease progression status. The three values ​​updated in steps 3, 4, and 5—the Aβ monomer accumulation pool value, the Aβ oligomer accumulation pool value, and the synaptic damage accumulation index value—are output as the disease progression status at the end of the current window. These three values ​​together constitute the virtual simulation result of the disease progression at that moment.

[0127] The above 6 steps are repeated in each window. As the window moves forward along the time axis, the three disease progression indicators output by the model evolve continuously over time, forming a dynamic disease progression curve that gradually transitions from the preclinical stage to mild cognitive impairment and even the Alzheimer's disease dementia stage.

[0128] All the above operations were performed independently for the p-Tau181 channel and the Aβ42 oligomer channel, with each channel driving an independent disease progression curve. In the final output, the corresponding indices of the two curves were fused according to the following rules: For the Aβ monomer accumulation pool index, the arithmetic mean of the corresponding values ​​from the two channels was calculated as the fused Aβ monomer accumulation pool output value. Aβ monomers are common upstream products of both cascade reactions; the monomer generation amounts of both the p-Tau181 and Aβ42 oligomer channels originate from the same precursor cleavage process. Therefore, the arithmetic mean was used as a comprehensive estimate to reduce the impact of single-channel measurement noise. For the Aβ oligomer accumulation pool index, the larger of the corresponding values ​​from the two channels was taken as the fused Aβ oligomer accumulation pool output value. The principle of taking the larger value is based on a conservative clinical risk assessment strategy, i.e., using the higher oligomer accumulation amount in the two pathological pathways as a warning to avoid underestimating the disease. For the synaptic damage accumulation index, the weighted sum of the corresponding values ​​from the two channels was taken. In this embodiment, the weight of the Aβ42 oligomer channel is set to 0.7, and the weight of the p-Tau181 channel is set to 0.3. This weight allocation is based on the following: Aβ oligomers are the main direct mediator of synaptic toxicity, and their cumulative level plays a dominant role in driving synaptic damage; p-Tau181 reflects the pathological pathology of tau protein phosphorylation, indirectly affecting synaptic function through the postsynaptic membrane skeletal protein phosphorylation pathway, and its contribution is relatively low. The weight allocation of 0.7 / 0.3 is determined based on the relative proportions of the downregulation of postsynaptic membrane receptors caused by Aβ oligomers and p-Tau181, respectively, in in vitro synaptic toxicity studies. The three fused index values ​​together constitute a comprehensive virtual simulation result of the disease progression, providing clinicians with a reference when judging the disease progression trend and adjusting the dosage of targeted drugs.

[0129] like Figure 2 As shown, embodiments of the present invention also provide a virtual simulation system for the course of Alzheimer's disease based on exosome biomarkers, comprising: The acquisition module is used to acquire the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration output from the cerebrospinal fluid exosome multimodal coupled sensing. The calculation module is used to calculate the ratio of the concentration value of each sampling point to the preset reference concentration value as the normalized concentration value. On the normalized plane formed by the time coordinate and the normalized concentration value, each sampling point is used as the current point. The previous sampling point and the next sampling point are obtained. The current point and the previous sampling point are connected to form a first line segment, and the current point and the next sampling point are connected to form a second line segment. The angle between the first line segment and the second line segment is calculated. The identification module is used to identify the current point as a transient peak point of concentration if the included angle is less than a preset angle threshold and the normalized concentration value of the current point is greater than the normalized concentration values ​​of the previous sampling point and the next sampling point. The extraction module is used to arrange all the identified transient peak points of concentration in chronological order, extract the pulse release frequency sequence based on the time interval between two adjacent transient peak points, extract the pulse release amplitude sequence based on the concentration value of each transient peak point, and retain the included angle corresponding to each transient peak point as a pulse release intensity measurement sequence. The driving module is used to input the pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence as input features into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease course. The dynamic simulation model obtains the virtual simulation results of the disease course based on the dynamic evolution rate of amyloid protein cascade reaction and synaptic damage driven by the pulse release features.

[0130] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0131] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A virtual simulation method for the course of Alzheimer's disease based on exosome biomarkers, characterized in that, include: Acquire the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration output from cerebrospinal fluid exosome multimodal coupled sensing; For the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration, the ratio of the concentration value at each sampling point to the preset reference concentration value is calculated as the normalized concentration value. Taking each sampling point as the current point, the previous and next sampling points are obtained. The current point and the previous sampling point are connected to form the first line segment, and the current point and the next sampling point are connected to form the second line segment. The angle between the first line segment and the second line segment is calculated. If the included angle is less than the preset angle threshold, and the normalized concentration value of the current point is greater than the normalized concentration values ​​of the previous sampling point and the next sampling point, then the current point will be identified as the concentration transient peak point. All identified transient peak points of concentration are arranged in chronological order. The pulse release frequency sequence is extracted based on the time interval between two adjacent transient peak points. The pulse release amplitude sequence is extracted based on the concentration value of each transient peak point. At the same time, the included angle corresponding to each transient peak point of concentration is retained as a pulse release intensity measurement sequence. The pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence are used as input features and input into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease course. The dynamic simulation model obtains the virtual simulation results of the disease course based on the dynamic evolution rate of amyloid protein cascade reaction and synaptic damage driven by the pulse release features.

2. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 1, characterized in that, The time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration were obtained from the multimodal coupled sensing output of cerebrospinal fluid exosomes, including: Cerebrospinal fluid samples are continuously collected at a constant flow rate through a cerebrospinal fluid microdialysis catheter implanted in the lateral ventricle of the patient, and the cerebrospinal fluid samples are then input into a nanoporous exosome capture microfluidic chip. The cerebrospinal fluid sample was enriched with exosomes using a nanoporous exosome capture microfluidic chip to obtain a sample enriched with exosomes. The sample enriched with exosomes was passed through a surface plasmon resonance multichannel biosensor array. The p-Tau181 specific capture antibody in the sensing array captured the p-Tau181 exosomes, and the Aβ42 oligomer specific capture antibody captured the Aβ42 oligomers of the exosomes, thus obtaining a resonance angle shift sensing map. Based on the drift characteristics of the sensing signal when no capture event occurs in the resonance angle shift sensing map, the baseline drift law is determined, and the baseline component is separated from the resonance angle shift sensing map according to the baseline drift law, while retaining the resonance angle shift response components caused by the exosome p-Tau181 capture event and the exosome Aβ42 oligomer capture event. The resonance angle shift response components were mapped to the time series of exosome p-Tau181 concentration and the time series of exosome Aβ42 oligomer concentration, respectively.

3. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 2, characterized in that, The resonance angle shift response components were mapped to time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration, respectively, including: Based on the components of the resonance angle shift response, the p-Tau181 resonance angle shift response caused by the exosome p-Tau181 capture event and the Aβ42 oligomer resonance angle shift response caused by the exosome Aβ42 oligomer capture event were identified; Based on the p-Tau181 resonance angle shift response, and according to the resonance angle shift characteristics caused by the unit concentration binding of the p-Tau181 specific capture antibody on the surface of the sensor chip, the p-Tau181 resonance angle shift at each time point is converted into p-Tau181 concentration values, and the exosome p-Tau181 concentration time series is obtained by arranging them in chronological order. Based on the resonance angle shift response of Aβ42 oligomers, and according to the resonance angle shift characteristics caused by the specific capture of antibody unit concentration binding by Aβ42 oligomers on the sensor chip surface, the resonance angle shift of Aβ42 oligomers at each time point is converted into Aβ42 oligomer concentration values, and arranged in chronological order to obtain the time series of exosome Aβ42 oligomer concentrations.

4. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 3, characterized in that, The calculation of the included angle between the first line segment and the second line segment specifically includes: The location points on the normalization plane are formed by the normalized time coordinates and normalized concentration values ​​of each sampling point; The time series of exosome p-Tau181 concentration and the time series of exosome Aβ42 oligomer concentration were used as the concentration time series to be processed, and a sampling point was selected from each concentration time series to be processed as the current sampling point. Based on the time position of the current sampling point in the concentration time series, determine the previous sampling point corresponding to the previous time point immediately adjacent to the current sampling point and the next sampling point corresponding to the next time point; On the normalized plane, connect the position point corresponding to the current sampling point with the position point corresponding to the previous sampling point to form the first line segment; On the normalized plane, connect the position point corresponding to the current sampling point with the position point corresponding to the next sampling point to form a second line segment; Based on the first line segment and the second line segment, with the location point corresponding to the current sampling point as the intersection point, the included angle between the two line segments formed by the first line segment and the second line segment is formed.

5. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 4, characterized in that, If the included angle is less than a preset angle threshold, and the normalized concentration value of the current point is greater than the normalized concentration values ​​of both the previous and next sampling points, then the current point is identified as a transient concentration peak point, including: The normalized concentration value of the current sampling point is determined based on the normalized concentration value corresponding to the current sampling point in the corresponding concentration time series; the normalized concentration value of the previous sampling point is determined based on the normalized concentration value corresponding to the previous sampling point in the corresponding concentration time series; the normalized concentration value of the next sampling point is determined based on the normalized concentration value corresponding to the next sampling point in the corresponding concentration time series. Based on the angle formed by the first line segment and the second line segment, determine whether the angle is less than a preset angle threshold, and obtain the angle determination result; Based on the current concentration value, the previous concentration value, and the next concentration value, determine whether the current concentration value is greater than both the previous and next concentration values ​​simultaneously, and obtain the concentration comparison result. When the angle judgment result is that the angle is less than the preset angle threshold, and the concentration comparison result is that the concentration value of the current sampling point is greater than the concentration values ​​of the previous sampling point and the next sampling point, the current sampling point is identified as the transient peak point of concentration.

6. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 5, characterized in that, The pulse release frequency sequence is extracted based on the time interval between two adjacent transient peak points, and the pulse release amplitude sequence is extracted based on the concentration value of each transient peak point. Simultaneously, the included angle corresponding to each transient peak point is retained as a pulse release intensity measurement sequence, including: Based on all the identified concentration transient peak points, arrange all the concentration transient peak points in chronological order according to the time position of each concentration transient peak point in its respective concentration time series to obtain the peak point sequence; Based on the time points corresponding to two adjacent transient peak points in the peak point sequence, the time interval between two adjacent transient peak points is determined, and the time interval is used as the pulse release cycle interval. All pulse release cycle intervals constitute the pulse release frequency sequence. Based on the concentration value corresponding to each transient peak point in the peak point sequence in its respective concentration time series, the concentration value of each transient peak point is taken as the peak concentration value, and the pulse release amplitude sequence is composed of all the peak concentration values. Based on the angle between the first line segment and the second line segment calculated during the identification process for each concentration transient peak point in the peak point sequence, this angle is used as the pulse release sharpness metric value corresponding to the concentration transient peak point, and the pulse release intensity metric sequence is composed of all pulse release sharpness metric values.

7. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 6, characterized in that, The pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence are used as input features and fed into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease progression, including: Based on the pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence, the pulse release frequency, pulse release amplitude, and pulse release intensity measurement within the same time window are combined to construct a multidimensional input feature set characterizing the pulse release dynamics of neurons. A multidimensional input feature set is input into a pre-constructed dynamic simulation model of Alzheimer's disease course driven by pulse release features. The dynamic simulation model pre-defines the correspondence between pulse features and disease evolution mechanisms. According to the aforementioned correspondence, the dynamic simulation model regulates the rate at which amyloid precursors are cleaved to produce Aβ using a pulse release frequency sequence, regulates the aggregation rate of Aβ monomers into oligomers using a pulse release amplitude sequence, and regulates the rate at which the degree of synaptic damage intensifies with the accumulation of Aβ oligomers using a pulse release intensity measurement sequence, thereby driving the dynamic evolution of the amyloid cascade reaction and synaptic damage.

8. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 7, characterized in that, The damage rate, which is modulated by pulse release intensity as a measure of the degree of synaptic damage, is exacerbated by the accumulation of Aβ oligomers, including: Based on the pulse release intensity measurement sequence, determine the magnitude of the pulse release sharpness measurement value within each time window; Based on the magnitude of the pulse release sharpness measurement value within each time window, the synaptic damage acceleration factor corresponding to each time window is determined. Based on the cumulative amount of Aβ oligomers simulated by the dynamic simulation model in the current time window, combined with the synaptic damage acceleration factor, the degree of synaptic damage is driven to accelerate and intensify as the cumulative amount of Aβ oligomers increases.

9. The virtual simulation method for Alzheimer's disease progression based on exosome biomarkers according to claim 8, characterized in that, Based on the magnitude of the pulse release sharpness metric within each time window, the synaptic damage acceleration factor corresponding to each time window is determined, including: Based on the pulse release sharpness measurement value within each time window, determine whether the pulse release sharpness measurement value is less than the preset sharpness threshold, and obtain the sharpness level judgment result; When the sharpness level judgment result is that the pulse release sharpness measurement value is less than the sharpness threshold, the time window is determined as a high-intensity pulse release window, and the preset high-intensity corresponding first acceleration factor is called as the synaptic damage acceleration factor for the time window. When the sharpness level judgment result is that the sharpness measurement value of the pulse release is not less than the sharpness threshold, the time window is determined as a low-intensity pulse release window, and the preset second acceleration factor corresponding to low intensity is called as the synaptic damage acceleration factor of the time window, wherein the first acceleration factor is greater than the second acceleration factor.

10. A virtual simulation system for the course of Alzheimer's disease based on exosome biomarkers, wherein the system implements the method as described in any one of claims 1 to 9, characterized in that, include: The acquisition module is used to acquire the time series of exosome p-Tau181 concentration and exosome Aβ42 oligomer concentration output from the cerebrospinal fluid exosome multimodal coupled sensing. The calculation module is used to calculate the ratio of the concentration value of each sampling point to the preset reference concentration value as the normalized concentration value. On the normalized plane formed by the time coordinate and the normalized concentration value, each sampling point is used as the current point. The previous sampling point and the next sampling point are obtained. The current point and the previous sampling point are connected to form a first line segment, and the current point and the next sampling point are connected to form a second line segment. The angle between the first line segment and the second line segment is calculated. The identification module is used to identify the current point as a transient peak point of concentration if the included angle is less than a preset angle threshold and the normalized concentration value of the current point is greater than the normalized concentration values ​​of the previous sampling point and the next sampling point. The extraction module is used to arrange all the identified transient peak points of concentration in chronological order, extract the pulse release frequency sequence based on the time interval between two adjacent transient peak points, extract the pulse release amplitude sequence based on the concentration value of each transient peak point, and retain the included angle corresponding to each transient peak point as a pulse release intensity measurement sequence. The driving module is used to input the pulse release frequency sequence, pulse release amplitude sequence, and pulse release intensity measurement sequence as input features into a pre-constructed pulse release feature-driven dynamic simulation model of Alzheimer's disease course. The dynamic simulation model obtains the virtual simulation results of the disease course based on the dynamic evolution rate of amyloid protein cascade reaction and synaptic damage driven by the pulse release features.

Images