Sparse data reconstruction method and related device

Abstract

The invention discloses a sparse data reconstruction method and a related device, and relates to the field of data processing, and the method comprises the steps: obtaining an intracranial pressure sparse data flow collected at a sampling rate in a preset low sampling rate range in a sliding time window, calculating an intracranial pressure mean value, taking the intracranial pressure mean value as an initial value, and calculating the initial value; the method comprises the following steps: carrying out dynamic characteristic analysis on an intermediate intracranial pressure data sequence obtained by preprocessing an intracranial pressure sparse data stream, adding a step mark corresponding to a pathological step to obtain a target intracranial pressure data sequence, carrying out segmented fitting on the target intracranial pressure data sequence to obtain a segmented fitting result, and carrying out signal reconstruction to obtain a multi-component mathematical model set; the sampling rate is set to be within a preset high sampling rate range, and a reconstructed intracranial pressure change waveform is generated. According to the method, the intracranial pressure sparse data flow is preprocessed, and then step detection and model reconstruction are carried out, so that an intracranial pressure signal equivalent to an intracranial pressure signal acquired at a high sampling rate and a change trend waveform of the intracranial pressure signal are reconstructed from the intracranial pressure sparse data flow at the low sampling rate.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a sparse data reconstruction method and related apparatus. Background Technology

[0002] Intracranial pressure (ICP) is a key monitoring indicator in the treatment of critical illnesses such as traumatic brain injury, cerebral hemorrhage, and brain tumors. ICP fluctuates dynamically with heartbeat and respiratory rhythm. When abnormalities such as intracranial edema or hemorrhage occur, the pressure can rise suddenly, endangering the patient's life.

[0003] Traditional wired intracranial pressure (ICP) monitoring devices use a sampling rate several times higher than the heart rate (typically ≥100Hz) to collect pressure data, followed by a moving average over a time window containing several fluctuation cycles (e.g., 6 seconds) to obtain a stable mean ICP value. However, because the implanted portion of the wired sensor needs to be connected to an external monitoring device via cable, the risk of infection is high during long-term monitoring; therefore, routine clinical monitoring time generally does not exceed 7 days. However, clinical studies have shown that patients with traumatic brain injury face a life-threatening risk due to elevated ICP within 30 days, highlighting the significant clinical importance of long-term ICP monitoring.

[0004] To achieve long-term intracranial pressure monitoring, wireless intracranial pressure monitoring devices have been developed to realize the goal of wireless implantable intracranial pressure monitoring. These devices are tiny, with limited built-in battery capacity, and require long-term continuous pressure monitoring within a very small space, thus posing extremely high power consumption requirements. To minimize power consumption and extend device operating time, wireless intracranial pressure monitoring devices typically use low sampling rates (such as 1Hz or 0.5Hz). However, low sampling rates often result in sparse intracranial pressure data streams. To obtain a stable mean intracranial pressure from these sparse data streams, a moving average method is generally applied directly to the low sampling rate data for calculation.

[0005] However, obtaining a stable mean requires a very long moving average window (e.g., data spanning tens of seconds or even minutes). When intracranial pressure experiences a pathological, sudden, and rapid change (step increase), this method cannot reflect the true pressure changes in a timely and accurate manner, leading to a significant lag in early warning, which clearly does not meet the clinical needs of real-time monitoring of critically ill patients. Furthermore, simple moving average processing smooths out key dynamic features in the signal, especially physiological fluctuations with frequencies higher than the Nyquist frequency (half the sampling rate) (such as cardiac pulsation harmonics), resulting in incomplete reconstructed trend information that cannot be fully equated to the trend acquired by high-sampling-rate devices.

[0006] Therefore, how to provide a sparse data reconstruction method to accurately reconstruct the real intracranial pressure signal and its changing trend from a low-sampling-rate sparse intracranial pressure data stream, so as to provide clinicians with reliable and accurate diagnostic basis, has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention discloses a sparse data reconstruction method and related apparatus to accurately reconstruct the real intracranial pressure signal and its changing trend from a low-sampling-rate sparse intracranial pressure data stream, providing clinicians with reliable and accurate diagnostic basis.

[0008] A sparse data reconstruction method includes:

[0009] Acquire sparse intracranial pressure data streams within a sliding time window at a sampling rate within a preset low sampling rate range, and calculate the mean intracranial pressure of the sparse intracranial pressure data streams;

[0010] The sparse intracranial pressure data stream is preprocessed to obtain a continuous and effective intermediate intracranial pressure data sequence.

[0011] Using the mean intracranial pressure as the initial value, dynamic feature analysis is performed on the intermediate intracranial pressure data sequence, and step markers are added to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence.

[0012] The target intracranial pressure data sequence is segmented and fitted to obtain segmented fitting results, wherein the segmented fitting results include: multiple segmented data, and the optimal fitting model corresponding to each segmented data;

[0013] The segmented fitting results are reconstructed to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure.

[0014] The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform.

[0015] Optionally, the mean intracranial pressure is used as the initial value to perform dynamic feature analysis on the intermediate intracranial pressure data sequence, and step markers are added to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence, including:

[0016] Using the mean intracranial pressure as the initial value, a change point detection algorithm based on statistical significance test is used to identify pressure step change segments in the intermediate intracranial pressure data sequence.

[0017] A step marker is added for each pressure step change segment, and the step marker is recorded in the step marker candidate list. The step marker includes: the time point of step occurrence and the step change amplitude.

[0018] Each step marker in the candidate list of step markers is identified based on an adaptive threshold, and each step marker is identified as representing a physiological fluctuation or a pathological step.

[0019] Delete the step markers representing physiological fluctuations from the candidate list of step markers, and add the remaining step markers to the corresponding pressure step change segments to obtain the target intracranial pressure data sequence.

[0020] Optionally, the target intracranial pressure data sequence is segmented and fitted to obtain segmented fitting results, including:

[0021] Using step markers as boundaries, the target intracranial pressure data sequence is divided into multiple segments;

[0022] For each segment of data, multiple different types of fitting models are used to perform dynamic curve fitting, and the fitting error corresponding to each fitting model is obtained.

[0023] The fitting model with the smallest fitting error is selected as the optimal fitting model for this segment of data.

[0024] The segmented fitting result is composed of all the segmented data and the optimal fitting model corresponding to each segment.

[0025] Optionally, signal reconstruction is performed on the piecewise fitting results to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure, including:

[0026] For each segment of data, the optimal fitting model corresponding to the segment of data is determined as the trend term;

[0027] The segmented data is processed using a constrained nonlinear least squares algorithm to obtain periodic coefficients;

[0028] The primary and secondary periodic fluctuation terms are determined based on the periodic coefficients.

[0029] Based on the trend term, the main periodic fluctuation term, and the secondary periodic fluctuation term corresponding to each segment of data, a multi-component mathematical model describing the dynamic changes in intracranial pressure is constructed.

[0030] The multi-component mathematical models corresponding to all segmented data are processed into a set to obtain the multi-component mathematical model set.

[0031] Optionally, the sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform, including:

[0032] The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate the reconstructed original intracranial pressure change waveform;

[0033] The original intracranial pressure change waveform is subjected to moving average filtering within the sliding time window to obtain a smooth and continuous intracranial pressure change waveform.

[0034] Optionally, acquiring a sparse intracranial pressure data stream within a sliding time window at a sampling rate within a preset low sampling rate range includes:

[0035] A sliding time window mechanism is used to acquire sparse intracranial pressure data streams collected at a sampling rate within a preset low sampling rate range within the sliding time window;

[0036] The sparse intracranial pressure data stream is divided into multiple data segments of equal length.

[0037] Calculate the mean of each of the data segments;

[0038] The differences between the various means were tested using a t-test.

[0039] When the test results show that there are differences, the average of the means of all data segments is taken to obtain the overall mean;

[0040] When the total mean is less than a preset mean threshold, the variance of the mean of all data segments is calculated.

[0041] When the variance is less than a preset variance threshold, the intracranial pressure sparse data stream is determined to be in a stable state, and the initialization operation of the intracranial pressure sparse data stream is completed.

[0042] Optionally, the sparse intracranial pressure data stream is preprocessed to obtain a continuous and effective intermediate intracranial pressure data sequence, including:

[0043] The Unet neural network is used to classify each intracranial pressure data in the sparse intracranial pressure data stream into usable data or unusable data, wherein the input of the Unet neural network is a one-dimensional array;

[0044] For each unavailable data point, calculate the mean of two adjacent available data points, and replace the unavailable data point with the mean.

[0045] Once all unusable data has been replaced, a continuous and valid intermediate intracranial pressure data sequence is obtained.

[0046] A sparse data reconstruction apparatus, comprising:

[0047] The data acquisition unit is used to acquire the intracranial pressure sparse data stream collected within a sliding time window at a sampling rate within a preset low sampling rate range, and to calculate the mean intracranial pressure of the intracranial pressure sparse data stream.

[0048] The preprocessing unit is used to preprocess the sparse intracranial pressure data stream to obtain a continuous and effective intermediate intracranial pressure data sequence.

[0049] The labeling unit is used to take the mean intracranial pressure as the initial value, perform dynamic feature analysis on the intermediate intracranial pressure data sequence, and add step markers to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence.

[0050] The fitting unit is used to perform segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results, wherein the segmented fitting results include: multiple segmented data and the optimal fitting model corresponding to each segmented data.

[0051] The model reconstruction unit is used to reconstruct the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure.

[0052] The reconstruction unit is used to set the sampling rate of the multi-component mathematical model set to a preset high sampling rate range in order to generate a reconstructed intracranial pressure change waveform.

[0053] A computer storage medium storing at least one instruction that, when executed by a processor, implements the sparse data reconstruction method described above.

[0054] An electronic device, comprising: a memory and a processor;

[0055] The memory is used to store at least one instruction;

[0056] The processor is used to execute the at least one instruction to implement the sparse data reconstruction method described above.

[0057] As can be seen from the above technical solution, the present invention discloses a sparse data reconstruction method and related apparatus. The method involves acquiring a sparse intracranial pressure data stream within a sliding time window at a sampling rate within a preset low sampling rate range, calculating the mean intracranial pressure of the sparse intracranial pressure data stream, preprocessing the sparse intracranial pressure data stream to obtain a continuous and effective intermediate intracranial pressure data sequence, using the mean intracranial pressure as the initial value, performing dynamic feature analysis on the intermediate intracranial pressure data sequence, adding step markers to pressure step change segments corresponding to pathological step changes to obtain a target intracranial pressure data sequence, performing segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results, reconstructing the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes of intracranial pressure, and setting the sampling rate of the multi-component mathematical model set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform. This invention eliminates noise and missing value interference by preprocessing the sparse intracranial pressure data stream, obtaining a continuous and effective intermediate intracranial pressure data sequence. Step detection is achieved through dynamic feature analysis of the intermediate intracranial pressure data sequence to accurately capture pathological steps in intracranial pressure and add step markers, resulting in the target intracranial pressure data sequence. By segmenting and fitting the target intracranial pressure data sequence and reconstructing a multi-component mathematical model set, and then combining this multi-component mathematical model set with high sampling rate operation, the multi-scale dynamic features of intracranial pressure can be accurately decoupled and high-resolution waveforms can be reconstructed. This allows for the reconstruction of intracranial pressure signals and their changing trends from a low-sampling-rate sparse intracranial pressure data stream, achieving results comparable to those obtained with high sampling rate acquisition. This provides clinicians with reliable and accurate diagnostic information while preserving necessary dynamic features, aiding in the analysis of the patient's condition. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the published drawings without creative effort.

[0059] Figure 1 This is a flowchart of a sparse data reconstruction method disclosed in an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of the structure of a sparse data reconstruction device disclosed in an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] This invention discloses a sparse data reconstruction method and related apparatus. By preprocessing the sparse intracranial pressure data stream, noise and missing value interference can be eliminated to obtain a continuous and effective intermediate intracranial pressure data sequence. By performing dynamic feature analysis on the intermediate intracranial pressure data sequence to achieve step detection, the pathological steps of intracranial pressure can be accurately captured and step markers can be added to obtain the target intracranial pressure data sequence. By segmenting and fitting the target intracranial pressure data sequence and reconstructing a multi-component mathematical model set, and combining the multi-component mathematical model set with high sampling rate operation, the multi-scale dynamic features of intracranial pressure can be accurately decoupled and high-resolution waveforms can be reconstructed. This enables the reconstruction of intracranial pressure signals and their changing trend waveforms from low-sampling-rate sparse intracranial pressure data streams, which is comparable to those obtained from high-sampling-rate acquisition. This provides clinicians with reliable and accurate diagnostic basis while preserving necessary dynamic features, which is helpful for the analysis of the condition.

[0064] See Figure 1 The present invention discloses a flowchart of a sparse data reconstruction method, which includes the following steps:

[0065] Step S101: Obtain the intracranial pressure sparse data stream collected within the sliding time window at a sampling rate within a preset low sampling rate range, and calculate the intracranial pressure mean of the intracranial pressure sparse data stream.

[0066] The value of the preset low sampling rate range is determined according to actual needs. For example, the preset low sampling rate range is 0.5 Hz to 1.5 Hz.

[0067] Within a sliding time window, the system continuously receives sparse intracranial pressure data streams from a wirelessly implanted sensor. These sparse intracranial pressure data streams are acquired at a preset low sampling rate (e.g., 1Hz or 0.5Hz), and the raw intracranial pressure data acquired each time is denoted as RawData(k) (where k is the number of samples). The system stores the RawData(k) of each sampling point into a dynamic data queue in chronological order, thereby forming a complete intracranial pressure sparse data stream RawData_Stream within each sliding time window.

[0068] In the process of storing the raw intracranial pressure data RawData(k) collected each time into the intracranial pressure sparse data stream RawData_Stream, a sliding time window mechanism is adopted. Specifically, new data is continuously added to the end of the window, and historical data that exceeds the window capacity is automatically discarded.

[0069] After obtaining the sparse intracranial pressure data stream, the average intracranial pressure within the sliding time window is obtained by calculating the average value of all the original intracranial pressure data in the sparse intracranial pressure data stream.

[0070] Step S102: Preprocess the sparse intracranial pressure data stream to obtain a continuous and effective intermediate intracranial pressure data sequence.

[0071] Unet neural network is a convolutional neural network architecture designed specifically for image segmentation tasks. Its core features are reflected in symmetrical structure, skip connections, and multi-scale feature fusion.

[0072] In this embodiment, the Unet neural network is used to classify each intracranial pressure data in the sparse intracranial pressure data stream into usable data or unusable data (i.e., abnormal data).

[0073] For each unavailable data point, calculate the mean of the two adjacent available data points, and replace the unavailable data point with the mean.

[0074] Once all unusable data has been replaced, a continuous and valid intermediate intracranial pressure data sequence is obtained.

[0075] It should be noted that the input of the Unet neural network can be a one-dimensional array (tensor) or a multi-dimensional array (tensor). In this embodiment, the input of the Unet neural network is a one-dimensional array.

[0076] Step S103: Using the mean intracranial pressure as the initial value, perform dynamic feature analysis on the intermediate intracranial pressure data sequence, and add step markers to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence.

[0077] The physiological characteristics of intracranial pressure exhibit multi-scale physiological rhythm features: high-frequency heart rate fluctuations originate from the periodic changes in intracranial volume caused by cardiac pulsation; low-frequency respiratory fluctuations are caused by thoracic pressure regulating venous return. The superposition of heart rate fluctuations and respiratory fluctuations together reflects the slow variation trend of long-term intracranial pressure regulation.

[0078] The pathological characteristics of intracranial pressure, mainly pathological steps, disrupt this balance, manifesting as a rapid, non-periodic increase.

[0079] Based on the differences between physiological and pathological characteristics, this invention uses the mean intracranial pressure as the initial value, performs dynamic feature analysis on the intermediate intracranial pressure data sequence, distinguishes between physiological fluctuations and pathological steps, and adds step markers to the pressure step change segments corresponding to pathological steps, thereby obtaining the target intracranial pressure data sequence.

[0080] Step S104: Perform segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results.

[0081] In practical applications, multiple different types of fitting models are used to fit each segment of the target intracranial pressure data sequence to obtain the final segmented fitting result.

[0082] The piecewise fitting results in this application include: multiple piecewise data, and the optimal fitting model corresponding to each piecewise data.

[0083] Step S105: Reconstruct the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure.

[0084] In this embodiment, a multi-component mathematical model P_model describing the dynamic changes of intracranial pressure is established for each segment of data in the segmented fitting results, thereby obtaining a set of multi-component mathematical models describing the dynamic changes of intracranial pressure.

[0085] Step S106: Set the sampling rate of the multi-component mathematical model set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform.

[0086] The value of the preset high sampling rate range is determined according to actual needs, and this invention does not limit it.

[0087] In this embodiment, the sampling rate of the multi-component mathematical model set is set to a high sampling rate within a preset high sampling rate range, and then the multi-component mathematical model set is run to generate a reconstructed intracranial pressure change waveform.

[0088] In summary, this invention discloses a sparse data reconstruction method. It acquires a sparse intracranial pressure data stream within a sliding time window at a sampling rate within a preset low sampling rate range, calculates the mean intracranial pressure of the sparse data stream, preprocesses the sparse data stream to obtain a continuous and effective intermediate intracranial pressure data sequence, uses the mean intracranial pressure as the initial value, performs dynamic feature analysis on the intermediate intracranial pressure data sequence, adds step markers to pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence, performs segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results, reconstructs the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes of intracranial pressure, and sets the sampling rate of the multi-component mathematical model set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform. This invention eliminates noise and missing value interference by preprocessing the sparse intracranial pressure data stream, obtaining a continuous and effective intermediate intracranial pressure data sequence. Step detection is achieved through dynamic feature analysis of the intermediate intracranial pressure data sequence to accurately capture pathological steps in intracranial pressure and add step markers, resulting in the target intracranial pressure data sequence. By segmenting and fitting the target intracranial pressure data sequence and reconstructing a multi-component mathematical model set, and then combining this multi-component mathematical model set with high sampling rate operation, the multi-scale dynamic features of intracranial pressure can be accurately decoupled and high-resolution waveforms can be reconstructed. This allows for the reconstruction of intracranial pressure signals and their changing trends from a low-sampling-rate sparse intracranial pressure data stream, achieving results comparable to those obtained with high sampling rate acquisition. This provides clinicians with reliable and accurate diagnostic information while preserving necessary dynamic features, aiding in the analysis of the patient's condition.

[0089] Furthermore, this invention supports wireless implantable devices operating at extremely low sampling rates. Compared to traditional high sampling rate solutions, this solution reduces device power consumption by two orders of magnitude, significantly extending the continuous monitoring time and meeting long-term monitoring requirements of up to 30 days. Simultaneously, in terms of response speed, this invention reduces the response time of traditional low sampling rate methods from several minutes to seconds, comparable to the response speed of high sampling rate wired devices.

[0090] In one embodiment, step S101 may specifically include:

[0091] (1) The intracranial pressure sparse data stream is acquired by using a sliding time window mechanism within the sliding time window at a sampling rate within a preset low sampling rate range.

[0092] (2) The intracranial pressure sparse data stream is divided into equal-length segments to obtain multiple data segments.

[0093] (3) Calculate the mean of each data segment.

[0094] The mean of each data segment is obtained by calculating the average of all raw intracranial pressure data within each data segment.

[0095] (4) Use t-test to test the differences between the means.

[0096] The t-test (Student's t-test) is a statistical inference method based on the t-distribution, used to compare whether there is a significant difference between the means of two groups of data, or to test whether the mean of a single group of data is significantly different from the known population mean. Its core principle is to calculate the t-statistic, combined with the degrees of freedom and the significance level (α), to determine whether the observed difference is likely caused by random error rather than a true difference. Based on this, this embodiment uses the t-test to examine the difference in the means of all data segments and obtains the test results.

[0097] (5) When the test results show that there is a difference, the average of the mean of all data segments is taken to obtain the overall mean.

[0098] (6) When the total mean is less than the preset mean threshold, calculate the variance of the mean of all data segments.

[0099] The value of the preset average threshold is determined according to actual needs, such as 0.5 mmHg.

[0100] (7) When the variance is less than the preset variance threshold, the intracranial pressure sparse data stream is determined to be stable data, and the initialization operation of the intracranial pressure sparse data stream is completed.

[0101] The value of the preset variance threshold is determined according to actual needs, for example, 2.5.

[0102] (8) Calculate the mean intracranial pressure of the sparse intracranial pressure data stream.

[0103] In one embodiment, step S103 may specifically include:

[0104] (1) Using the mean intracranial pressure as the initial value, a change point detection algorithm based on statistical significance test is used to identify the pressure step change segment in the intermediate intracranial pressure data sequence.

[0105] This embodiment performs dynamic feature analysis on the preprocessed intermediate intracranial pressure data sequence based on the physiological characteristics of intracranial pressure.

[0106] Statistical significance testing quantifies the degree of discrepancy between sample data and the null hypothesis to determine whether the difference is caused by random error.

[0107] Change point detection aims to identify points in a data sequence where statistical properties (such as mean and variance) change significantly.

[0108] In this embodiment, the mean intracranial pressure is used as the initial value, which is the baseline reference value representing the normal intracranial pressure level. The statistical significance test can quantify whether the difference between the intermediate intracranial pressure data sequence and the initial value is statistically significant, that is, to determine whether the difference is caused by random fluctuations.

[0109] The change point detection algorithm can identify the start and end times and change characteristics of pressure steps in the intracranial pressure data sequence, thereby obtaining the pressure step change segment.

[0110] (2) Add a step marker for each pressure step change segment and record the step marker in the step marker candidate list.

[0111] The step indicator includes: the time point t_step when the step occurs and the step change magnitude ΔP_step. The step change magnitude ΔP_step is the difference in value between the stable point and the point where the step occurs.

[0112] (3) Identify each step marker in the candidate list of step markers based on an adaptive threshold, and identify each step marker as representing a physiological fluctuation or a pathological step.

[0113] The adaptive threshold is not a fixed value, but is dynamically adjusted in real time based on changes in the physiological baseline of intracranial pressure data.

[0114] Physiological fluctuations are characterized by small amplitudes (e.g., <baseline ±10% or <3 mmHg) and are usually within the adaptive threshold range.

[0115] Pathological steps are characterized by amplitudes that significantly exceed adaptive thresholds (e.g., > baseline ± 20% or > 8 mmHg) and durations that are long (e.g., > 10 minutes).

[0116] Since each step marker contains the step occurrence time t_step and the step change amplitude ΔP_step, each step marker can be identified as representing a physiological fluctuation or a pathological step by using an adaptive threshold.

[0117] (4) Delete the step markers that represent physiological fluctuations from the candidate list of step markers, and add the remaining step markers to the corresponding pressure step change segment to obtain the target intracranial pressure data sequence.

[0118] The target intracranial pressure data sequence is actually an intermediate intracranial pressure data sequence with added step markers corresponding to pathological step jumps.

[0119] In one embodiment, step S104 may specifically include:

[0120] (1) Using step markers as boundaries, the target intracranial pressure data sequence is divided into multiple segments.

[0121] For the target intracranial pressure data sequence AnalyzedData with added step markers representing pathological steps, the target intracranial pressure data sequence is divided into multiple segments using the step markers as boundaries.

[0122] (2) For each segment of data, multiple different types of fitting models are used to perform dynamic curve fitting, and the fitting error corresponding to each fitting model is obtained.

[0123] For example, for each segment of data, four fitting models—linear, power-law, exponential, and logarithmic—are used for dynamic curve fitting to obtain the fitting parameters corresponding to each fitting model. By fitting the fitting parameters with the data of this segment, the fitting error corresponding to each fitting model is obtained, such as the root mean square error (RMSE).

[0124] (3) Select the fitting model with the smallest fitting error as the optimal fitting model for this segment of data.

[0125] For each data segment, the model with the smallest fitting error is selected as the optimal fitting model for that data segment.

[0126] (4) The segmented fitting results are formed by combining all the segmented data and the optimal fitting model corresponding to each segmented data.

[0127] In practical applications, the optimal fitting parameters of all segmented data can be used as the structured data object FittedModelSet obtained by segmented fitting.

[0128] In one embodiment, step S105 may specifically include:

[0129] (1) For each segment of data, the optimal fitting model corresponding to the segment of data is determined as the trend term.

[0130] The trend term describes long-term changes in intracranial pressure.

[0131] (2) The segmented data is processed by a constrained nonlinear least squares algorithm to obtain the periodic coefficients.

[0132] Nonlinear least squares is a mathematical method for estimating the parameters of a nonlinear model by minimizing the sum of squared errors; it is a variant of least squares. Its model is represented as y = f(x, θ), where θ is the parameter to be estimated, and the nonlinear characteristic is reflected in the model's nonlinear dependence on θ.

[0133] This embodiment uses a constrained nonlinear least squares algorithm to process the segmented data and obtain the periodic coefficient θ.

[0134] (3) Determine the primary periodic fluctuation term and the secondary periodic fluctuation term based on the periodic coefficient.

[0135] The primary periodic fluctuation term sinθ and the secondary periodic fluctuation term cosθ are determined based on the period coefficient θ.

[0136] The primary and secondary periodic fluctuation terms are used to simulate the effects of physiological rhythms such as heartbeat and respiration.

[0137] (4) Based on the trend term, the main periodic fluctuation term and the secondary periodic fluctuation term corresponding to each segment of data, construct a multi-component mathematical model describing the dynamic changes of intracranial pressure.

[0138] The expression for the multi-component mathematical model is as follows:

[0139] P_model(t) = Trend(t) + Wave_primary(t) + Wave_secondary(t);

[0140] In the formula, P_model(t) represents a multi-component mathematical model, Trend(t) represents the trend term, Wave_primary(t) represents the primary periodic fluctuation term, Wave_secondary(t) represents the secondary periodic fluctuation term, and t represents the data acquisition time.

[0141] (5) The multi-component mathematical models corresponding to all segmented data are processed into a set to obtain the multi-component mathematical model set.

[0142] The ReconstructedSet of multi-component mathematical models includes the multi-component mathematical model P_model(t) corresponding to all segmented data.

[0143] In one embodiment, step S106 may specifically include:

[0144] The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate the reconstructed original intracranial pressure change waveform;

[0145] The original intracranial pressure change waveform is subjected to moving average filtering within the sliding time window to obtain a smooth and continuous intracranial pressure change waveform.

[0146] This embodiment combines a multi-component mathematical model set with high sampling rate operation, which can accurately decouple the multi-scale dynamic characteristics of intracranial pressure and reconstruct the original intracranial pressure change waveform, i.e., the high-resolution waveform. By performing moving average filtering on the reconstructed original intracranial pressure change waveform within a sliding time window, it is ensured that the reconstructed intracranial pressure change waveform is highly consistent with the intracranial pressure change waveform acquired at the high sampling rate. This provides clinicians with reliable and accurate diagnostic information while preserving necessary dynamic features, which is beneficial for disease analysis and enhances clinical applicability.

[0147] Retrospective validation using clinical data demonstrates that the sparse data reconstruction method provided by this invention has the following effectiveness:

[0148] (1) High reconstruction accuracy: The root mean square error between the reconstructed signal and the original data sampled at 100Hz is less than 0.5mmHg.

[0149] (2) Fast step response: For simulated sudden increase in intracranial pressure (>5 mmHg / min), the warning time of the method of the present invention is 2-3 minutes earlier than the low sampling rate direct moving average method.

[0150] (3) Excellent trend consistency: The correlation coefficient (R) between the finally extracted intracranial pressure trend and the gold standard method is high. 2 () greater than 0.98.

[0151] (4) Significant power consumption advantage: Under the premise of ensuring the same monitoring accuracy, the power consumption of the equipment can be reduced to less than 1 / 100 of the original solution.

[0152] Therefore, this invention effectively solves the core contradiction between "low power consumption" and "high accuracy / fast response" in the field of wireless implantable intracranial pressure monitoring, and has significant technological advancements and clinical application value.

[0153] Corresponding to the above method embodiments, the present invention also discloses a sparse data reconstruction apparatus.

[0154] See Figure 2 The present invention discloses a schematic diagram of a sparse data reconstruction device, which may include:

[0155] The data acquisition unit 201 is used to acquire intracranial pressure sparse data streams collected within a sliding time window at a sampling rate within a preset low sampling rate range, and to calculate the intracranial pressure mean of the intracranial pressure sparse data streams.

[0156] The value of the preset low sampling rate range is determined according to actual needs. For example, the preset low sampling rate range is 0.5 Hz to 1.5 Hz.

[0157] Within a sliding time window, the system continuously receives sparse intracranial pressure data streams from a wirelessly implanted sensor. These sparse intracranial pressure data streams are acquired at a preset low sampling rate (e.g., 1Hz or 0.5Hz), and the raw intracranial pressure data acquired each time is denoted as RawData(k) (where k is the number of samples). The system stores the RawData(k) of each sampling point into a dynamic data queue in chronological order, thereby forming a complete intracranial pressure sparse data stream RawData_Stream within each sliding time window.

[0158] In the process of storing the raw intracranial pressure data RawData(k) collected each time into the intracranial pressure sparse data stream RawData_Stream, a sliding time window mechanism is adopted. Specifically, new data is continuously added to the end of the window, and historical data that exceeds the window capacity is automatically discarded.

[0159] After obtaining the sparse intracranial pressure data stream, the average intracranial pressure within the sliding time window is obtained by calculating the average value of all the original intracranial pressure data in the sparse intracranial pressure data stream.

[0160] The preprocessing unit 202 is used to preprocess the sparse intracranial pressure data stream to obtain a continuous and effective intermediate intracranial pressure data sequence.

[0161] The labeling unit 203 is used to take the mean intracranial pressure as the initial value, perform dynamic feature analysis on the intermediate intracranial pressure data sequence, and add step markers to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence.

[0162] The physiological characteristics of intracranial pressure exhibit multi-scale physiological rhythm features: high-frequency heart rate fluctuations originate from the periodic changes in intracranial volume caused by cardiac pulsation; low-frequency respiratory fluctuations are caused by thoracic pressure regulating venous return. The superposition of heart rate fluctuations and respiratory fluctuations together reflects the slow variation trend of long-term intracranial pressure regulation.

[0163] The pathological characteristics of intracranial pressure, mainly pathological steps, disrupt this balance, manifesting as a rapid, non-periodic increase.

[0164] Based on the differences between physiological and pathological characteristics, this invention uses the mean intracranial pressure as the initial value, performs dynamic feature analysis on the intermediate intracranial pressure data sequence, distinguishes between physiological fluctuations and pathological steps, and adds step markers to the pressure step change segments corresponding to pathological steps, thereby obtaining the target intracranial pressure data sequence.

[0165] Fitting unit 204 is used to perform segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results, wherein the segmented fitting results include: multiple segmented data and the optimal fitting model corresponding to each segmented data.

[0166] In practical applications, multiple different types of fitting models are used to fit each segment of the target intracranial pressure data sequence to obtain the final segmented fitting result.

[0167] The piecewise fitting results in this application include: multiple piecewise data, and the optimal fitting model corresponding to each piecewise data.

[0168] The reconstruction model determination unit 205 is used to reconstruct the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure.

[0169] In this embodiment, a multi-component mathematical model P_model describing the dynamic changes of intracranial pressure is established for each segment of data in the segmented fitting results, thereby obtaining a set of multi-component mathematical models describing the dynamic changes of intracranial pressure.

[0170] The reconstruction unit 206 is used to set the sampling rate of the multi-component mathematical model set to a preset high sampling rate range in order to generate a reconstructed intracranial pressure change waveform.

[0171] The value of the preset high sampling rate range is determined according to actual needs, and this invention does not limit it.

[0172] In this embodiment, the sampling rate of the multi-component mathematical model set is set to a high sampling rate within a preset high sampling rate range, and then the multi-component mathematical model set is run to generate a reconstructed intracranial pressure change waveform.

[0173] In summary, this invention discloses a sparse data reconstruction device that acquires a sparse intracranial pressure data stream within a sliding time window at a sampling rate within a preset low sampling rate range, calculates the mean intracranial pressure of the sparse intracranial pressure data stream, preprocesses the sparse intracranial pressure data stream to obtain a continuous and effective intermediate intracranial pressure data sequence, uses the mean intracranial pressure as the initial value, performs dynamic feature analysis on the intermediate intracranial pressure data sequence, adds step markers to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence, performs segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results, reconstructs the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes of intracranial pressure, and sets the sampling rate of the multi-component mathematical model set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform. This invention eliminates noise and missing value interference by preprocessing the sparse intracranial pressure data stream, obtaining a continuous and effective intermediate intracranial pressure data sequence. Step detection is achieved through dynamic feature analysis of the intermediate intracranial pressure data sequence to accurately capture pathological steps in intracranial pressure and add step markers, resulting in the target intracranial pressure data sequence. By segmenting and fitting the target intracranial pressure data sequence and reconstructing a multi-component mathematical model set, and then combining this multi-component mathematical model set with high sampling rate operation, the multi-scale dynamic features of intracranial pressure can be accurately decoupled and high-resolution waveforms can be reconstructed. This allows for the reconstruction of intracranial pressure signals and their changing trends from a low-sampling-rate sparse intracranial pressure data stream, achieving results comparable to those obtained with high sampling rate acquisition. This provides clinicians with reliable and accurate diagnostic information while preserving necessary dynamic features, aiding in the analysis of the patient's condition.

[0174] Furthermore, this invention supports wireless implantable devices operating at extremely low sampling rates. Compared to traditional high sampling rate solutions, this solution reduces device power consumption by two orders of magnitude, significantly extending the continuous monitoring time and meeting long-term monitoring requirements of up to 30 days. Simultaneously, in terms of response speed, this invention reduces the response time of traditional low sampling rate methods from several minutes to seconds, comparable to the response speed of high sampling rate wired devices.

[0175] In one embodiment, the data acquisition unit 201 can be specifically used for:

[0176] A sliding time window mechanism is used to acquire sparse intracranial pressure data streams collected at a sampling rate within a preset low sampling rate range within the sliding time window;

[0177] The sparse intracranial pressure data stream is divided into multiple data segments of equal length.

[0178] Calculate the mean of each of the data segments;

[0179] The differences between the various means were tested using a t-test.

[0180] When the test results show that there are differences, the average of the means of all data segments is taken to obtain the overall mean;

[0181] When the total mean is less than a preset mean threshold, the variance of the mean of all data segments is calculated.

[0182] When the variance is less than a preset variance threshold, the intracranial pressure sparse data stream is determined to be in a stable state, and the initialization operation of the intracranial pressure sparse data stream is completed.

[0183] In one embodiment, the preprocessing unit 202 may specifically be used for:

[0184] The Unet neural network is used to classify each intracranial pressure data in the sparse intracranial pressure data stream into usable data or unusable data, wherein the input of the Unet neural network is a one-dimensional array;

[0185] For each unavailable data point, calculate the mean of two adjacent available data points, and replace the unavailable data point with the mean.

[0186] Once all unusable data has been replaced, a continuous and valid intermediate intracranial pressure data sequence is obtained.

[0187] In one embodiment, the marking unit 203 can be specifically used for:

[0188] Using the mean intracranial pressure as the initial value, a change point detection algorithm based on statistical significance test is used to identify pressure step change segments in the intermediate intracranial pressure data sequence.

[0189] A step marker is added for each pressure step change segment, and the step marker is recorded in the step marker candidate list. The step marker includes: the time point of step occurrence and the step change amplitude.

[0190] Each step marker in the candidate list of step markers is identified based on an adaptive threshold, and each step marker is identified as representing a physiological fluctuation or a pathological step.

[0191] Delete the step markers representing physiological fluctuations from the candidate list of step markers, and add the remaining step markers to the corresponding pressure step change segments to obtain the target intracranial pressure data sequence.

[0192] In one embodiment, the fitting unit 204 can be specifically used for:

[0193] Using step markers as boundaries, the target intracranial pressure data sequence is divided into multiple segments;

[0194] For each segment of data, multiple different types of fitting models are used to perform dynamic curve fitting, and the fitting error corresponding to each fitting model is obtained.

[0195] The fitting model with the smallest fitting error is selected as the optimal fitting model for this segment of data.

[0196] The segmented fitting result is composed of all the segmented data and the optimal fitting model corresponding to each segment.

[0197] In one embodiment, the reconstructed model determination unit 205 can be specifically used for:

[0198] For each segment of data, the optimal fitting model corresponding to the segment of data is determined as the trend term;

[0199] The segmented data is processed using a constrained nonlinear least squares algorithm to obtain periodic coefficients;

[0200] The primary and secondary periodic fluctuation terms are determined based on the periodic coefficients.

[0201] Based on the trend term, the main periodic fluctuation term, and the secondary periodic fluctuation term corresponding to each segment of data, a multi-component mathematical model describing the dynamic changes in intracranial pressure is constructed.

[0202] The multi-component mathematical models corresponding to all segmented data are processed into a set to obtain the multi-component mathematical model set.

[0203] In one embodiment, the reconstruction unit 206 can be specifically used for:

[0204] The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate the reconstructed original intracranial pressure change waveform;

[0205] The original intracranial pressure change waveform is subjected to moving average filtering within the sliding time window to obtain a smooth and continuous intracranial pressure change waveform.

[0206] It should be noted that for the specific working principles of each component in the device embodiment, please refer to the corresponding section of the method embodiment, which will not be repeated here.

[0207] Corresponding to the above embodiments, the present invention also discloses a computer storage medium that stores at least one instruction, which, when executed by a processor, implements the steps shown in the embodiments of the sparse data reconstruction method.

[0208] Computer storage media can be tangible media that may contain or store programs for use by or in conjunction with an instruction execution system, apparatus, or device. Computer storage media can be machine-readable signal media or machine-readable storage media. Computer storage media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0209] Corresponding to the above embodiments, such as Figure 3 As shown, the present invention also provides a schematic diagram of the structure of an electronic device, which may include: a processor 1 and a memory 2;

[0210] The processor 1 and memory 2 communicate with each other via communication bus 3.

[0211] Processor 1, for executing at least one instruction;

[0212] Memory 2 is used to store at least one instruction;

[0213] Processor 1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0214] Memory 2 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0215] In this embodiment, the processor executes at least one instruction to implement the steps shown in the sparse data reconstruction method.

[0216] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0217] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0218] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A sparse data reconstruction method, characterized in that, include: Acquire sparse intracranial pressure data streams within a sliding time window at a sampling rate within a preset low sampling rate range, and calculate the mean intracranial pressure of the sparse intracranial pressure data streams; The sparse intracranial pressure data stream is preprocessed to obtain a continuous and effective intermediate intracranial pressure data sequence. Using the mean intracranial pressure as the initial value, dynamic feature analysis is performed on the intermediate intracranial pressure data sequence, and step markers are added to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence. The target intracranial pressure data sequence is segmented and fitted to obtain segmented fitting results, wherein the segmented fitting results include: multiple segmented data, and the optimal fitting model corresponding to each segmented data; The segmented fitting results are reconstructed to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure. The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate a reconstructed intracranial pressure change waveform.

2. The sparse data reconstruction method according to claim 1, characterized in that, Using the mean intracranial pressure as the initial value, dynamic feature analysis is performed on the intermediate intracranial pressure data sequence, and step markers are added to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence, including: Using the mean intracranial pressure as the initial value, a change point detection algorithm based on statistical significance test is used to identify pressure step change segments in the intermediate intracranial pressure data sequence. A step marker is added for each pressure step change segment, and the step marker is recorded in the step marker candidate list. The step marker includes: the time point of step occurrence and the step change amplitude. Each step marker in the candidate list of step markers is identified based on an adaptive threshold, and each step marker is identified as representing a physiological fluctuation or a pathological step. Delete the step markers representing physiological fluctuations from the candidate list of step markers, and add the remaining step markers to the corresponding pressure step change segments to obtain the target intracranial pressure data sequence.

3. The sparse data reconstruction method according to claim 1, characterized in that, The target intracranial pressure data sequence is segmented and fitted to obtain segmented fitting results, including: Using step markers as boundaries, the target intracranial pressure data sequence is divided into multiple segments; For each segment of data, multiple different types of fitting models are used to perform dynamic curve fitting, and the fitting error corresponding to each fitting model is obtained. The fitting model with the smallest fitting error is selected as the optimal fitting model for this segment of data. The segmented fitting result is composed of all the segmented data and the optimal fitting model corresponding to each segment.

4. The sparse data reconstruction method according to any one of claims 1 to 3, characterized in that, Signal reconstruction is performed on the piecewise fitting results to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure, including: For each segment of data, the optimal fitting model corresponding to the segment of data is determined as the trend term; The segmented data is processed using a constrained nonlinear least squares algorithm to obtain periodic coefficients; The primary and secondary periodic fluctuation terms are determined based on the periodic coefficients. Based on the trend term, the main periodic fluctuation term, and the secondary periodic fluctuation term corresponding to each segment of data, a multi-component mathematical model describing the dynamic changes in intracranial pressure is constructed. The multi-component mathematical models corresponding to all segmented data are processed into a set to obtain the multi-component mathematical model set.

5. The sparse data reconstruction method according to any one of claims 1 to 3, characterized in that, The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate reconstructed intracranial pressure change waveforms, including: The sampling rate of the multi-component mathematical model set is set to a preset high sampling rate range to generate the reconstructed original intracranial pressure change waveform; The original intracranial pressure change waveform is subjected to moving average filtering within the sliding time window to obtain a smooth and continuous intracranial pressure change waveform.

6. The sparse data reconstruction method according to claim 1, characterized in that, Acquire sparse intracranial pressure data streams within a sliding time window, collected at sampling rates within a preset low sampling rate range, including: A sliding time window mechanism is used to acquire sparse intracranial pressure data streams collected at a sampling rate within a preset low sampling rate range within the sliding time window; The sparse intracranial pressure data stream is divided into multiple data segments of equal length. Calculate the mean of each of the data segments; The differences between the various means were tested using a t-test. When the test results show that there are differences, the average of the means of all data segments is taken to obtain the overall mean; When the total mean is less than a preset mean threshold, the variance of the mean of all data segments is calculated. When the variance is less than a preset variance threshold, the intracranial pressure sparse data stream is determined to be in a stable state, and the initialization operation of the intracranial pressure sparse data stream is completed.

7. The sparse data reconstruction method according to claim 1 or 6, characterized in that, The sparse intracranial pressure data stream is preprocessed to obtain a continuous and effective intermediate intracranial pressure data sequence, including: The Unet neural network is used to classify each intracranial pressure data in the sparse intracranial pressure data stream into usable data or unusable data, wherein the input of the Unet neural network is a one-dimensional array; For each unavailable data point, calculate the mean of two adjacent available data points, and replace the unavailable data point with the mean. Once all unusable data has been replaced, a continuous and valid intermediate intracranial pressure data sequence is obtained.

8. A sparse data reconstruction device, characterized in that, include: The data acquisition unit is used to acquire the intracranial pressure sparse data stream collected within a sliding time window at a sampling rate within a preset low sampling rate range, and to calculate the mean intracranial pressure of the intracranial pressure sparse data stream. The preprocessing unit is used to preprocess the sparse intracranial pressure data stream to obtain a continuous and effective intermediate intracranial pressure data sequence. The labeling unit is used to take the mean intracranial pressure as the initial value, perform dynamic feature analysis on the intermediate intracranial pressure data sequence, and add step markers to the pressure step change segments corresponding to pathological step changes to obtain the target intracranial pressure data sequence. The fitting unit is used to perform segmented fitting on the target intracranial pressure data sequence to obtain segmented fitting results, wherein the segmented fitting results include: multiple segmented data and the optimal fitting model corresponding to each segmented data. The model reconstruction unit is used to reconstruct the signal based on the segmented fitting results to obtain a set of multi-component mathematical models describing the dynamic changes in intracranial pressure. The reconstruction unit is used to set the sampling rate of the multi-component mathematical model set to a preset high sampling rate range in order to generate a reconstructed intracranial pressure change waveform.

9. A computer storage medium, characterized in that, The computer storage medium stores at least one instruction, which, when executed by a processor, implements the sparse data reconstruction method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The electronic device includes: a memory and a processor; The memory is used to store at least one instruction; The processor is used to execute the at least one instruction to implement the sparse data reconstruction method as described in any one of claims 1 to 7.

Images