IoT-based biosensor-based platelet-rich plasma concentration detection system

Abstract

This invention relates to the interdisciplinary field of biomedical detection and Internet of Things (IoT) sensing technology, specifically disclosing a platelet-rich plasma concentration detection system based on IoT biosensors. The system includes a microfluidic sensor chip, a multi-frequency electrochemical impedance spectroscopy (MES) acquisition unit, a cell morphology identification module, an IoT communication unit, and a remote data processing server. It monitors platelet dynamic aggregation in real time through a three-electrode system, and combines high-dimensional features of broadband impedance spectroscopy with a pre-trained convolutional neural network model to accurately distinguish between simple platelet aggregates, platelet-leukocyte complexes, and microthrombus structures. Furthermore, it leverages the IoT architecture to complete encrypted data upload, remote model optimization, and intelligent coagulation risk assessment. Through the above technical solutions, this invention improves the sensitivity, specificity, and clinical applicability of the detection, constructing a closed-loop blood monitoring system from bedside detection to cloud-based decision support.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of biomedical detection and Internet of Things (IoT) sensing technology, specifically relating to a platelet-rich plasma concentration detection system based on IoT biosensors. Background Technology

[0002] With the deep integration of biomedical testing and IoT technology, sensor-based blood component analysis has become a key technology for clinical diagnosis and personalized treatment. Platelet-rich plasma (PRP), a biological agent rich in growth factors, is crucial for accurate concentration detection in orthopedic repair, skin regeneration, and coagulation function assessment. IoT biosensors, by integrating microfluidics and electrochemical sensing technologies, can achieve real-time dynamic monitoring of key cellular components in blood samples, providing high-frequency, objective data support for clinical decision-making.

[0003] Electrochemical impedance spectroscopy (EIS) is a core approach for assessing platelet function and concentration. It analyzes charge transfer and substance exchange processes at the electrode interface by applying multi-frequency alternating electrical signals and capturing the response current. This technique aims to characterize platelet aggregation under specific inducers by analyzing fluctuations in impedance modulus, thereby retrieving the effective cell concentration in the sample. Within an IoT framework, the raw electrical signals acquired by the sensors are processed and converted into digital features to enable remote monitoring and intelligent early warning of coagulation risks under physiological conditions.

[0004] Current biosensors primarily rely on single-dimensional changes in impedance amplitude for total quantification during detection, making it difficult to finely identify complex cellular structures. This method cannot effectively distinguish between simple platelet aggregates and heterogeneous complexes rich in leukocytes, nor can it identify physical interference from fibrinogen in microthrombi, resulting in insufficient specificity of the test results. Furthermore, under low platelet concentration conditions, environmental noise and baseline drift often lead to a reduced signal-to-noise ratio, making it difficult for traditional linear analysis indicators to capture subtle physical features, easily causing missed diagnoses or misjudgments in clinical practice. Due to the lack of analytical capabilities for high-dimensional spatial features such as impedance phase angle and relaxation time distribution, the system cannot achieve dynamic mapping of electrode surface deposit morphology, failing to meet the stringent requirements for high-precision blood analysis in complex clinical environments.

[0005] Therefore, a platelet-rich plasma concentration detection system based on Internet of Things biosensors is desired. Summary of the Invention

[0006] The purpose of this invention is to provide a platelet-rich plasma concentration detection system based on Internet of Things biosensors, which can solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The IoT-based biosensor-based platelet-rich plasma concentration detection system includes a microfluidic sensor chip, a multi-frequency electrochemical impedance spectroscopy acquisition unit, a cell morphology identification module, an IoT communication unit, and a remote data processing server, wherein: The microfluidic sensing chip is configured to hold the platelet-rich plasma sample to be tested, and integrates a working electrode, a counter electrode and a reference electrode inside it to form a three-electrode system, so as to monitor the dynamic aggregation process of platelets on the electrode surface in real time after the application of an inducing agent. The multi-frequency electrochemical impedance spectroscopy acquisition unit is connected to the microfluidic sensing chip and is configured to apply an alternating electrical signal covering a wide frequency range to the three-electrode system and simultaneously acquire the corresponding current response signal, thereby calculating a set of high-dimensional impedance spectrum data including impedance amplitude, phase angle, Nyquist plot characteristic parameters and relaxation time distribution. The cell morphology identification module is connected to the multi-frequency electrochemical impedance spectroscopy acquisition unit and is configured to receive the high-dimensional impedance spectroscopy data. It also performs feature extraction and pattern recognition on the data using a built-in pre-trained convolutional neural network model, mapping the electrochemical response of the electrode surface deposits to the corresponding cell morphology categories to distinguish between simple platelet aggregates, platelet-leukocyte complex aggregates, and microthrombus structures containing fibrin networks. The IoT communication unit is connected to the multi-frequency electrochemical impedance spectroscopy acquisition unit and the cell morphology identification module, respectively. It is configured to encrypt the raw impedance data, morphological identification results and system status information and upload them to the remote data processing server, and receive instruction update or model parameter adjustment signals from the server. The remote data processing server establishes a two-way communication link with the IoT communication unit through a wide area network, and is configured to store historical test data, perform cross-sample trend analysis, optimize convolutional neural network model parameters, and provide coagulation risk assessment reports and clinical decision support information to authorized terminals.

[0008] Preferably, the alternating electrical signal applied by the multi-frequency electrochemical impedance spectroscopy acquisition unit covers a continuous frequency band from low frequency to high frequency, and the Nyquist plot feature parameters acquired by it include the semicircle diameter in the high frequency region, the slope in the low frequency region, and the inflection point position in the mid frequency region, which are used to characterize the charge transfer and diffusion behavior at different time scales of the electrode interface.

[0009] Furthermore, the convolutional neural network model built into the cell morphology identification module uses a large number of labeled impedance spectrum-morphology paired datasets for supervised learning during the training phase. The paired datasets cover typical aggregation patterns under different platelet concentrations, white blood cell ratios, and fibrinogen contents, enabling the model to identify weak white blood cell-mediated aggregation features in low signal-to-noise ratio environments.

[0010] Furthermore, the working electrode surface of the microfluidic sensing chip is modified with a specific biorecognition layer. The biorecognition layer contains molecular probes that can selectively bind to platelet membrane glycoproteins to enhance the capture efficiency of target cells and inhibit non-specific adsorption, thereby improving the specificity and repeatability of impedance signals.

[0011] Preferably, the IoT communication unit adopts a low-power wide-area network communication protocol, which supports stable data transmission in various network environments such as hospital LAN, mobile cellular network or satellite link, and has the functions of interrupted transmission resume and local caching to ensure the integrity of detection data under network fluctuation conditions.

[0012] Furthermore, the remote data processing server is equipped with an anomaly detection engine, which can automatically identify outlier impedance spectra in a single test based on historical baseline data and trigger retest instructions or manual review processes to prevent misjudgments caused by sample contamination or operational errors.

[0013] Furthermore, the cell morphology identification module is also equipped with a visualization output interface, which can overlay the identification results onto the Nyquist or Bode map in the form of a pseudo-color heatmap, intuitively displaying the cell aggregate types corresponding to different regions, making it easier for clinicians to quickly interpret the compositional features of complex samples.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The platelet-rich plasma concentration detection system based on IoT biosensors provided by this invention breaks through the limitation of traditional impedance methods that rely solely on amplitude changes for total quantity estimation. By deeply integrating the high-dimensional features of multi-frequency impedance spectra with machine learning-driven morphological identification capabilities, it achieves accurate analysis of the microstructure of deposits on the electrode surface.

[0015] 2. The system can not only distinguish between simple platelet aggregation and complex structures containing leukocytes or fibrin, but also maintain effective coagulation risk assessment capabilities under pathological conditions of low platelet concentration by leveraging leukocyte-mediated aggregation characteristics, thus improving the sensitivity and specificity of the detection. Based on an IoT architecture, the system enables real-time uploading of detection data, remote model iteration, and intelligent early warning, constructing a closed-loop diagnostic support system from bedside detection to cloud-based analysis. This provides a high-precision, traceable, and intelligent blood monitoring solution for patients with thrombocytopenia, postoperative recovery groups, and individuals with coagulation disorders. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall technical solution architecture according to the present invention; Figure 2This is a schematic diagram of the core principle framework of cell morphology identification based on multi-frequency electrochemical impedance spectroscopy and convolutional neural network according to the present invention. Figure 3 This is a flowchart illustrating the logical flow of multi-frequency electrochemical impedance spectroscopy data acquisition and high-dimensional feature construction according to the present invention. Figure 4 This is a schematic diagram illustrating the multi-level interaction relationship and data flow between the IoT communication unit and the remote data processing server according to the present invention. Figure 5 This is a schematic diagram illustrating the principle framework of the target cell capture and signal enhancement of the microfluidic sensor chip surface specific bio-recognition layer according to the present invention. Detailed Implementation

[0017] Example 1: Please refer to the appendix Figure 1 To be continued Figure 5 To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0018] The IoT-based biosensor-based platelet-rich plasma concentration detection system includes a microfluidic sensor chip, a multi-frequency electrochemical impedance spectroscopy acquisition unit, a cell morphology identification module, an IoT communication unit, and a remote data processing server. The microfluidic sensing chip is configured to serve as a physical carrier and chemical reaction microenvironment for platelet-rich plasma samples to be tested. It integrates a microfluidic structure and a three-electrode electrochemical sensing system consisting of a working electrode, a counter electrode, and a reference electrode. The microfluidic sensing chip is configured to capture the dynamic aggregation process of platelets on the surface of the working electrode using the three-electrode electrochemical sensing system after receiving an inducer injection signal, and convert the biochemical signal into a measurable electrical response signal. The multi-frequency electrochemical impedance spectroscopy acquisition unit is physically connected to the electrode interface of the microfluidic sensing chip. It is configured to generate and apply a set of controlled alternating electrical signals covering a continuous frequency band from 0.1 Hz to 1 MHz to the three-electrode electrochemical sensing system, and to capture the corresponding response current signal in real time. The multi-frequency electrochemical impedance spectroscopy acquisition unit calculates a set of high-dimensional impedance spectrum data through a built-in digital signal processor. The high-dimensional impedance spectrum data includes impedance amplitude, phase angle, characteristic parameters in the complex impedance plane diagram, and relaxation time distribution characteristics obtained based on distribution function inversion. The cell morphology identification module establishes a data communication link with the multi-frequency electrochemical impedance spectroscopy acquisition unit, is configured to receive the high-dimensional impedance spectroscopy data, and performs multi-dimensional feature extraction and pattern recognition on the data through a pre-trained convolutional neural network model deployed within it. The cell morphology identification module is configured to map the complex electrochemical response of the electrode surface deposits to specific cell morphology categories, and distinguish simple platelet aggregates, complex aggregates containing leukocytes, and microthrombus structures containing fibrin filament networks at the physical level. The IoT communication unit establishes electrical connections with the multi-frequency electrochemical impedance spectroscopy acquisition unit and the cell morphology identification module, respectively. It is configured to encapsulate the original impedance data, morphology identification results and sensor hardware status information using an industrial-grade encryption protocol, and upload the encapsulated data packets to the remote data processing server via a low-power wide area network protocol. At the same time, it receives logical update instructions or weight adjustment parameters of the convolutional neural network model from the server. The remote data processing server establishes a two-way transparent transmission channel with the IoT communication unit through a wide area network, is configured to build a large-scale historical detection database of platelet-rich plasma, and perform cross-sample statistical trend analysis. The remote data processing server has an embedded model optimization engine, which is used to continuously update the convolutional neural network model based on the feedback detection results, and finally generate a comprehensive assessment report including coagulation risk level, cell composition ratio and clinical treatment recommendations.

[0019] The microfluidic sensing chip includes a microchannel layer made of polydimethylsiloxane and an electrode array layer supported by a glass substrate. The microchannel layer has a serpentine mixing region and a central detection cavity etched within it. The serpentine mixing region ensures that the platelet-rich plasma and the inducer are fully mixed before reaching the detection area. The surface of the working electrode is covalently bonded to a specific biorecognition layer, which specifically contains monoclonal antibodies or specific nucleic acid aptamer molecular probes targeting the platelet membrane surface glycoprotein IIb / IIIa receptors. This modified structure significantly improves the specific capture efficiency of the working electrode for target platelets and forms a dense biosensing film at the electrode interface. By hindering changes in charge transfer resistance, it greatly enhances the signal-to-noise ratio of the impedance signal and reduces interference from non-target proteins in the plasma.

[0020] The multi-frequency electrochemical impedance spectroscopy acquisition unit includes a high-precision waveform generator, a potentiostat control circuit, and a differential signal conditioning circuit. The high-precision waveform generator is configured to generate weak sinusoidal perturbation signals with amplitudes between 5 mV and 15 mV to ensure the detection process remains within the electrochemical linear kinetic range, avoiding physical damage to platelet aggregation caused by large-amplitude voltage signals. The differential signal conditioning circuit amplifies the response current transimpedance using a built-in high-gain operational amplifier and performs multi-stage active filtering to remove 50 Hz power frequency interference and high-frequency radio frequency noise. After acquiring the digitized voltage and current sequence, the digital signal processor uses a fast Fourier transform algorithm to extract the real and imaginary parts of the impedance at each frequency point, thereby constructing a Nyquist plot. In the Nyquist plot, the acquisition unit is configured to accurately extract the semicircular diameter in the high-frequency region, which physically corresponds to the charge transfer resistance at the electrode interface; simultaneously, it extracts the slope of the straight line in the low-frequency region to characterize the diffusion impedance features of ions in the sample.

[0021] The cell morphology identification module incorporates a pre-trained convolutional neural network model with a deep residual network architecture. This model receives a normalized one-dimensional impedance vector or a two-dimensional reconstructed impedance spectrum at its input layer. The convolutional neural network model extracts local spatial correlation features from the impedance data through multiple convolutional kernels, such as phase peak fluctuations near specific frequency points. Since platelet aggregates and leukocyte-platelet complexes exhibit subtle differences in capacitance and conductivity, these differences manifest as different feature peak broadenings and shifts in the relaxation time distribution spectrum. The model fuses these nonlinear features through fully connected layers and outputs a three-dimensional probability vector, corresponding to the confidence scores of simple aggregates, complex aggregates, and fibrin networks, respectively. The cell morphology identification module is further configured with a visualization output interface, which overlays the identified cell morphology information as pseudo-color pixel blocks onto the corresponding frequency region of the Nyquist image, achieving a semantic fusion display of electrochemical spectra and microscopic morphological features.

[0022] The IoT communication unit integrates a hardware-level security encryption chip, supporting real-time encryption of data streams based on the Advanced Encryption Standard (AES-256). This unit is configured to be adaptive to different network environments, automatically and seamlessly switching between 5G and Narrowband Internet of Things (NB-IoT) links based on signal strength. To address potential network outages in clinical settings, the IoT communication unit is equipped with local non-volatile memory as a buffer, supporting a resume transmission protocol to ensure that backlogged historical test data is completely retransmitted to the server in timestamp order as soon as the network link is restored.

[0023] The remote data processing server is deployed in a distributed architecture. Its internal anomaly detection engine utilizes unsupervised learning algorithms based on isolated forests or support vector machines to perform real-time integrity verification of the uploaded impedance curves. If the Nyquist graph generated by a single detection exhibits distortion, such as an inductively coupled arc that does not conform to electrochemical principles or a sudden jump in data points, the anomaly detection engine will automatically mark the sample as "suspected contamination" or "operational anomaly" and issue a retest instruction to the front-end operator. Furthermore, the server also uses an integrated learning framework to perform federated learning on detection data from multiple sensors at different geographical locations. Without compromising patient privacy, it continuously iterates and optimizes the neural network weight parameters in the cell morphology identification module, improving the system's flood identification accuracy when processing extremely low concentration platelet samples.

[0024] Example 2: To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0025] This embodiment describes an IoT biosensor-based platelet-rich plasma concentration detection system based on an edge computing architecture. Compared to Embodiment 1, this embodiment focuses on enhancing the local data processing capabilities and real-time feedback mechanism, making it suitable for bedside monitoring scenarios in operating rooms where real-time performance is extremely critical.

[0026] The IoT-based biosensor-based platelet-rich plasma concentration detection system includes a microfluidic sensor chip, a multi-frequency electrochemical impedance spectroscopy acquisition unit, an edge-side morphological processing gateway, an IoT secure transmission unit, and a cloud-based medical big data platform. The microfluidic sensing chip, as a precision disposable biological consumable, internally comprises a sample inlet, a pretreatment filter membrane, a micro-electrochemical detection chamber, and a waste liquid recovery tank. The inner wall of the micro-electrochemical detection chamber is fitted with a working electrode made of gold using a chemical vapor deposition process. To further enhance the ability to identify extremely low concentrations of platelets, the surface of the working electrode is modified with a three-dimensional gold nanoparticle array to significantly increase the effective electrochemical active area. The chip is pre-loaded with lyophilized platelet-inducing reagent. When a platelet-rich plasma sample enters the detection chamber, the reagent rapidly rehydrates and induces platelet aggregation. The resulting interfacial impedance evolution is continuously monitored by the multi-frequency electrochemical impedance spectroscopy acquisition unit.

[0027] The multi-frequency electrochemical impedance spectroscopy acquisition unit is based on a high-performance field-programmable gate array (FPGA). This unit is equipped with a wideband transimpedance amplifier and a high-speed analog-to-digital converter with dual synchronous sampling, enabling it to simultaneously capture excitation voltage and response current signals at megahertz sampling rates. By performing a discrete Fourier transform, the acquisition unit can not only obtain the conventional impedance magnitude but also accurately analyze the imaginary and real parts of the impedance. The acquisition unit is configured to execute a parameter fitting logic based on a double-layer capacitance model, deconstructing the complex impedance response into physical parameters such as electrolyte resistance, double-layer capacitance, and charge transfer resistance, providing structured input features for subsequent morphological identification.

[0028] The edge-side morphological processing gateway, as the core processing node in this embodiment, relies on an embedded computing platform integrating a neural processing unit (NPU). This gateway is connected to the multi-frequency electrochemical impedance spectroscopy acquisition unit and is configured to receive a high-dimensional data stream containing impedance amplitude, phase angle, Nyquist plot feature parameters, and relaxation time distribution. A lightweight deep learning identification model is deployed within the gateway. This model, optimized through pruning and quantization techniques, can directly perform real-time cell morphology classification at the edge. The lightweight model is configured to focus on analyzing the characteristic peak broadening phenomenon in the relaxation time distribution function. Since the polarization response time constant of platelet-leukocyte complexes at the electrode interface is typically greater than that of simple platelet aggregates, the gateway can utilize this kinetic difference to locally and in real-time determine whether the electrode surface deposits contain heterogeneous cell complexes.

[0029] The IoT secure transmission unit is configured to establish a logically isolated secure communication link between the edge gateway and the cloud platform. This unit employs a hardware authentication mechanism based on Physically Unclonable Functions (PUFs) to ensure that each biosensor accessing the system has a unique and legitimate identity. During data transmission, the IoT secure transmission unit is configured to prioritize the preliminary identification results generated at the edge, sending critical coagulation warning information first via a low-latency channel, while asynchronously uploading the massive original high-dimensional impedance sequence when the system is idle.

[0030] The cloud-based medical big data platform receives aggregated data from multiple edge gateways via the internet. The platform includes a deep learning model iteration module that utilizes the collected massive amounts of clinical impedance data to periodically and automatically retrain and validate models. The platform is configured to use Long Short-Term Memory (LSTM) networks to perform time-series analysis on multiple test records of a single patient, providing doctors with dynamic assessment trend graphs of the patient's coagulation function by observing the evolution of platelet aggregation kinetic parameters over several days or weeks. Furthermore, the platform has a cross-institutional data sharing interface, supporting the synchronization of test results to the hospital's electronic medical record system within authorized scope.

[0031] In the specific workflow, the alternating electrical signal applied by the multi-frequency electrochemical impedance spectroscopy acquisition unit employs logarithmically evenly spaced frequency sampling points. This sampling method ensures sufficient spectral resolution in both the low-frequency diffusion region and the high-frequency charge transfer region. The Nyquist plot characteristic parameters calculated by the multi-frequency electrochemical impedance spectroscopy acquisition unit also include the inflection point frequency in the mid-frequency region, which is configured as an auxiliary indicator for assessing platelet membrane integrity.

[0032] The edge-side morphological processing gateway employs an adaptive feature weighting mechanism when performing pattern recognition. When the system detects that the impedance magnitude of the original signal is in a low range (indicating a low platelet concentration scenario), the gateway automatically increases the weight ratio of phase angle features and relaxation time distribution features in the model prediction. This logic is based on the following technical discovery: when the total cell count is insufficient, the absolute change in impedance amplitude is easily masked by environmental thermal noise, but the heterogeneous aggregation process mediated by leukocytes generates unique feature fingerprints at specific phase frequency points. By enhancing the sensitivity of these high-dimensional features, the system can maintain accurate discrimination of coagulation risk even under extreme conditions of extremely low cell concentration.

[0033] Furthermore, the biorecognition layer modified on the working electrode surface of the microfluidic sensing chip also integrates anti-nonspecific adsorption components, such as polyethylene glycol derivatives. This effectively prevents the formation of disordered protein crowns on the electrode surface by large amounts of albumin or globulin in plasma, ensuring that the measured impedance changes originate entirely from the physical occupancy and electrochemical shielding effect of the target cell aggregates.

[0034] The remote cloud platform is also equipped with an environmental compensation algorithm module, which can perform real-time calibration of the original impedance value returned by the edge end based on the ambient temperature, humidity and chip production batch information transmitted back in real time by the IoT communication unit, so as to offset the electrolyte resistance fluctuation caused by environmental temperature drift and ensure the consistency of cross-regional and cross-seasonal test results.

[0035] Example 3: To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0036] This embodiment describes an Internet of Things (IoT) biosensor platelet-rich plasma concentration detection system with multi-channel parallel detection capability and integrated visual human-computer interaction interface.

[0037] The IoT-based biosensor-based platelet-rich plasma concentration detection system includes a multi-channel microfluidic analysis terminal, a real-time impedance signal parsing unit, a deep vision and morphology fusion identification module, a multimodal IoT communication engine, and a central collaborative server. The multi-channel microfluidic analysis terminal is characterized by integrating multiple independent detection stations, each configured to independently accommodate a microfluidic sensor chip. This parallel architecture supports multiple control experiments on the same patient sample (e.g., adding different inducer concentrations), or simultaneous detection of samples from multiple patients. Each microfluidic sensor chip contains the aforementioned working electrode, counter electrode, and reference electrode system, and is equipped with an independent fluid-driven pump for precisely controlling the flow rate of the plasma sample within the microchannel.

[0038] The real-time impedance signal analysis unit is physically distributed below each detection station and includes a set of parallel analog front-end acquisition circuits. Each acquisition circuit is configured to independently perform impedance spectrum scanning over a wide frequency range. The analysis unit has a built-in hardware-accelerated calculation module for converting the acquired current and voltage signals into complex impedance parameters in real time. In particular, the analysis unit is configured to calculate the trajectory curvature of the impedance vector in the complex plane. The trajectory curvature characteristics can effectively reflect the nonlinear changes in the thickness and density of the deposition layer during platelet aggregation, providing a deeper physical basis for subsequent morphological classification.

[0039] The depth vision and morphology fusion recognition module, connected to the real-time impedance signal analysis unit, is based on a multimodal perception framework. This module is configured to receive not only high-dimensional impedance spectrum data (such as amplitude, phase, and Nyquist features) but also low-magnification raw image features acquired by a miniature optical sensor. The module's built-in convolutional neural network model employs a dual-branch structure: the first branch extracts frequency domain features of the impedance spectrum, and the second branch extracts macroscopic optical signal fluctuation features from the electrode surface. Through feature-level fusion technology, this module can distinguish with higher confidence the uniform impedance distribution generated by simple platelet aggregation from the non-uniform, highly discrete impedance features generated when a fibrin filament network is present.

[0040] The multimodal IoT communication engine is configured to support multi-source access capabilities including Wi-Fi, Bluetooth Low Energy (BLE), and cellular networks. This engine supports short-range communication with nearby mobile terminals (such as a doctor's tablet), enabling local real-time monitoring of the detection process. Simultaneously, the engine incorporates a data compression algorithm that utilizes the frequency domain redundancy of impedance spectrum data to losslessly compress high-dimensional feature data before uploading it via a remote wide area network, reducing power consumption and bandwidth usage while ensuring information integrity.

[0041] The central collaborative server, deployed in the central computer room of the medical institution or a third-party cloud service center, is configured to schedule and manage the testing tasks of the entire system and trace data. This server has a built-in large-scale convolutional neural network model training center, capable of periodically aggregating "difficult-to-identify sample" data (i.e., samples with a confidence level below a preset threshold) reported by all networked terminals. Experts then remotely annotate these samples, retrain the model, and issue update patches. The server is also equipped with an intelligent alarm module that automatically triggers epidemiological warnings or clinical follow-up reminders for specific populations when an abnormal concentration of coagulation risk assessment reports is detected in a certain area.

[0042] In specific implementation details, the Nyquist plot characteristic parameters acquired by the multi-frequency electrochemical impedance spectroscopy acquisition unit further include the low-frequency slope. The magnitude of this slope is configured to reflect the degree of obstruction to material diffusion near the electrode interface. When platelets form dense monolayer or multilayer aggregates on the electrode surface, the ion diffusion path becomes more tortuous, resulting in a decrease in the low-frequency slope.

[0043] The cell morphology identification module is also equipped with a pseudo-color heatmap mapping logic. This logic is configured to assign different color depths to impedance points corresponding to different frequency bands in the Nyquist plot based on the output probability distribution of the last layer Softmax of the convolutional neural network model. For example, the high-frequency semicircular region corresponding to charge transfer resistance is marked in red (representing the main contribution area of ​​platelets), and the low-frequency linear region corresponding to diffusion impedance is marked in blue (representing the contribution area of ​​fibrin network). In this way, clinicians can intuitively determine whether there are abnormal complex aggregate structures in the sample by the distribution of color patches, improving the readability of complex blood sample analysis.

[0044] Furthermore, the multi-channel analysis terminal integrates a temperature control unit, which can precisely maintain the temperature within the microfluidic chip at a physiological level of 37 degrees Celsius, with fluctuations of less than 0.1 degrees Celsius. This design is based on the high temperature sensitivity of platelet aggregation reactions, and by eliminating the interference of ambient temperature fluctuations on the impedance baseline, it further improves the repeatability of the system in different application scenarios.

[0045] The IoT communication unit is also equipped with a local caching strategy. When the detection system is in an environment with unstable signals, such as a mobile ambulance, the system automatically enters "offline mode" and stores all raw impedance spectrum data, morphological identification conclusions, and timestamps in a local high-speed Flash memory. Once the system detects a valid network connection, it will perform breakpoint resumption via the encrypted MQTT protocol and automatically complete the timeline alignment and data entry on the server side.

[0046] Example 4: To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0047] This embodiment describes an Internet of Things (IoT) biosensor system for detecting platelet-rich plasma concentration, which integrates adaptive signal enhancement logic and multi-dimensional clinical decision support functions.

[0048] The IoT-based biosensor-based platelet-rich plasma concentration detection system includes an intelligent microfluidic sensing terminal, a high-dimensional impedance feature extraction unit, a neural network morphology deconstruction module, a secure IoT relay station, and a clinical decision support server. The intelligent microfluidic sensing terminal, in addition to basic electrochemical acquisition functions, is also equipped with a mechanical oscillation submodule. This submodule is configured to apply micro-vibrations of a specific frequency to the sample within the microfluidic chip before sampling, eliminating gravitational interference caused by erythrocyte sedimentation and ensuring that the electrode surface primarily captures the effective components from platelet-rich plasma. The chip's working electrode utilizes a composite material of porous carbon nanotubes and gold nanoparticles. This hierarchical porous structure not only significantly improves the sensitivity of the impedance response but also provides better anchoring points for platelets, enabling stable electrical signal fluctuations even in pathological samples with extremely low platelet counts.

[0049] The high-dimensional impedance feature extraction unit is internally configured to execute an advanced analytical algorithm called "Distributed Relaxation Time (DRT)". This algorithm transforms the traditional impedance spectrum from the frequency domain to the time constant domain. The extraction unit is configured to identify peaks in the DRT spectrum representing different physical processes, such as short-time constant peaks representing internal electrolyte polarization and long-time constant peaks representing biofilm formation on the electrode surface. By extracting the peak height, full width at half maximum (FWHM), and center position of these peaks, the high-dimensional impedance feature extraction unit can construct feature vectors exceeding 128 dimensions, capturing weak morphological fingerprints that are completely masked in ordinary amplitude analysis.

[0050] The neural network morphology deconstruction module is configured within a high-end microcontroller inside the terminal. This module employs a convolutional neural network architecture based on an attention mechanism. This attention mechanism is configured to automatically identify the dimension in the feature vector that contributes most to cell classification. For example, when distinguishing between "platelet-leukocyte complexes," the neural network morphology deconstruction module automatically increases the weight of mid-to-low frequency phase angle features. The module's output is directly connected to an embedded display screen, capable of displaying the currently detected platelet concentration estimate and the proportion of the main morphological components on the electrode surface in real time.

[0051] The secure IoT relay station can physically be a smart gateway deployed within the ward. This relay station is configured to receive data from multiple sensor terminals using Bluetooth Low Energy and perform data cleaning and preliminary anomaly screening locally. The relay station is equipped with a blockchain-based identity traceability module, ensuring that every test report has an immutable timestamp and digital signature throughout its generation, transmission, and storage, meeting the stringent data compliance requirements of clinical diagnosis and treatment.

[0052] The clinical decision support server, serving as the intelligent central hub of the entire system, is equipped with a large database of experts in coagulation risk assessment. This server is configured to receive morphological recognition results forwarded by relay stations and perform comprehensive reasoning by combining them with the patient's medical history (such as surgical history and medication history). If the system detects a significant fibrin network structure on the electrode surface and the patient is in the postoperative thrombotic phase, the server will send a red alert for the highest level of coagulation risk to the attending physician's terminal via an instant messaging protocol, along with a detailed impedance evolution analysis report.

[0053] Furthermore, the high-dimensional impedance feature extraction unit is also equipped with a baseline self-calibration logic. During the dry state stage before sample injection and the buffer filling stage, the high-dimensional impedance feature extraction unit performs two impedance scans to obtain the initial impedance noise floor and distribution parameters of the sensor hardware. In subsequent plasma testing, the system automatically subtracts these background components from the real-time data to offset measurement errors caused by fluctuations in electrode manufacturing processes or parasitic inductance in wiring.

[0054] During training, the neural network morphology deconstruction module employed a Generative Adversarial Network (GAN) to augment the training set. Due to the high cost of obtaining specialized clinical samples containing leukocytes or microthrombi, the server utilized the GAN model to generate a large amount of simulated impedance spectrum data, covering various extreme interference conditions. Through learning on this augmented dataset, the model demonstrated generalization ability, effectively identifying weak leukocyte-mediated aggregation features that would be easily judged as noise in traditional methods. This has significant clinical value for the accurate assessment of coagulation function in patients with thrombocytopenia.

[0055] Finally, the central server is also equipped with a remote firmware upgrade (OTA) module. When there are significant improvements to the electrochemical impedance spectroscopy algorithm, or when new morphological feature points are discovered, the system can push the new algorithm weights and logic instructions to terminal devices around the world through the Internet of Things communication unit, achieving synchronous online evolution of detection accuracy.

[0056] This invention is not limited to the specific details of the above embodiments. Various improvements and modifications can be made by those skilled in the art without departing from the spirit and scope of this invention. For example, the electrode material of the microfluidic chip can be replaced with other conductive materials such as carbon fiber or graphene, or the convolutional neural network model can be replaced with other advanced deep learning architectures such as a transformer. These equivalent variations based on the core concept of this invention should all be included within the scope of protection of this invention. The various functional modules described in the system can be implemented using independent hardware chips or through software logic integrated into a system-on-a-chip (SoC). Furthermore, the connection method between the components in the system, provided that data transmission and power supply are satisfied, can be wired, wireless, or a combination of both.

Claims

1. A platelet-rich plasma concentration detection system based on Internet of Things biosensors, characterized in that, include: The microfluidic sensing chip is configured to serve as a carrier for platelet-rich plasma samples to be tested. It integrates a three-electrode system consisting of a working electrode, a counter electrode, and a reference electrode. This system is used to capture the dynamic aggregation process of platelets on the surface of the working electrode after receiving an inducer and to convert the biochemical signal into an electrical response signal. The multi-frequency electrochemical impedance spectroscopy acquisition unit is electrically connected to the microfluidic sensing chip and is configured to apply a controlled alternating electrical signal covering the frequency band from 0.1 Hz to 1 MHz to the three-electrode system and simultaneously capture the current response signal to calculate high-dimensional impedance spectral data including impedance amplitude, phase angle, complex impedance plane characteristic parameters and relaxation time distribution. The cell morphology identification module is communicatively connected to the multi-frequency electrochemical impedance spectroscopy acquisition unit. It is configured to receive the high-dimensional impedance spectroscopy data and extract impedance features through a built-in pre-trained convolutional neural network model. The electrochemical response is mapped to the corresponding cell morphology category to distinguish simple platelet aggregates, platelet-leukocyte complex aggregates, and microthrombus structures containing fibrin networks. The Internet of Things (IoT) communication unit is connected to the multi-frequency electrochemical impedance spectroscopy acquisition unit and the cell morphology identification module, respectively. It is configured to encrypt and encapsulate the raw impedance data, morphological identification results and system status information, upload them to the remote server, and receive model update instructions. The remote data processing server communicates bidirectionally with the IoT communication unit via a wide area network, and is configured to store historical test data, perform cross-sample trend analysis and optimize neural network model parameters, and provide coagulation risk assessment reports to the terminal.

2. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The microfluidic sensing chip includes a microchannel layer made of polydimethylsiloxane and an electrode array layer supported by a glass substrate. The microfluidic layer is etched with a serpentine mixing region and a central detection cavity. The serpentine mixing region is used to ensure that the platelet-rich plasma to be tested and the inducing agent are fully mixed before reaching the central detection cavity. The inner wall of the central detection chamber is prepared with a working electrode made of gold through a chemical vapor deposition process. The surface of the working electrode is modified with a three-dimensional gold nanoarray to increase the effective electrochemical activity area. The surface of the working electrode is modified with a specific biorecognition layer by a covalent bonding process. The specific biorecognition layer contains monoclonal antibodies or specific nucleic acid aptamer molecular probes targeting platelet membrane glycoprotein IIb / IIIa receptors. It is configured to improve the specific capture efficiency of target platelets and form a biosensing membrane at the electrode interface, which enhances the signal-to-noise ratio of impedance signal by hindering changes in charge transfer resistance. The specific biorecognition layer also integrates a non-specific adsorption component composed of polyethylene glycol derivatives to prevent albumin or globulin in plasma from forming a protein crown on the electrode surface.

3. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The multi-frequency electrochemical impedance spectroscopy acquisition unit includes a high-precision waveform generator, a potentiostat control circuit, and a differential signal conditioning circuit. The high-precision waveform generator is configured to generate sinusoidal perturbation signals with amplitudes between 5 mV and 15 mV. The differential signal conditioning circuit amplifies the response current through a built-in high-gain operational amplifier and performs multi-stage active filtering to filter out 50 Hz power frequency interference and high-frequency radio frequency noise. The multi-frequency electrochemical impedance spectroscopy acquisition unit also has a built-in digital signal processor, configured to extract the real and imaginary parts of the impedance at each frequency point using the fast Fourier transform algorithm, and to construct a complex impedance plane diagram. The digital signal processor is also configured to execute parameter fitting logic based on a double-layer capacitance model, decomposing the impedance response into physical parameters of electrolyte resistance, double-layer capacitance, and charge transfer resistance. In the complex impedance plane diagram, the acquisition unit is configured to extract the semicircular diameter in the high-frequency region, the slope of the straight line in the low-frequency region, and the inflection frequency in the mid-frequency region. The semicircular diameter in the high-frequency region is used to characterize the charge transfer resistance of the electrode interface, and the slope of the straight line in the low-frequency region is used to characterize the diffusion impedance characteristics of ions in the sample.

4. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The pre-trained convolutional neural network model built into the cell morphology identification module adopts a deep residual network architecture and receives a normalized one-dimensional impedance vector or a two-dimensional reconstructed impedance spectrum at the input layer. The convolutional neural network model extracts the phase peak fluctuation features that appear near the frequency point in the impedance data through multiple convolutional kernels. The identification module is equipped with an adaptive feature weighting submodule. When the impedance modulus of the original signal is detected to be in a preset low range, the adaptive feature weighting submodule automatically increases the weight ratio of phase angle feature and relaxation time distribution feature in the model prediction to enhance the identification sensitivity of leukocyte-mediated aggregation features in a low cell concentration environment. The convolutional neural network model also includes a fully connected layer, which is used to fuse the extracted nonlinear features and output a three-dimensional probability vector. The values ​​in the three-dimensional probability vector correspond to the confidence scores of simple aggregates, complex aggregates and fibrin networks, respectively. The identification module is also equipped with a visualization output interface, which is used to overlay the identified cell morphology information as pseudo-color pixel blocks on the corresponding frequency point area of ​​the complex impedance plane diagram, so as to realize the semantic fusion display of electrochemical spectrum and microscopic morphological features.

5. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The IoT communication unit integrates a hardware-level security encryption chip, which supports real-time encryption of data streams based on the Advanced Encryption Standard 256-bit algorithm. The IoT communication unit has a network environment adaptive switching circuit, which is used to switch between the fifth-generation mobile communication technology link and the narrowband IoT link according to the signal strength in the current environment. The IoT communication unit is also equipped with a local non-volatile memory as a data buffer and has built-in breakpoint resume protocol logic, which is used to store the original impedance sequence and morphological identification results locally when the network link is interrupted, and to resend them to the remote data processing server in complete order according to the timestamp after the network is restored. The IoT communication unit adopts a hardware identity authentication mechanism based on physically unclonable functions to ensure that sensors accessing the system have a unique and legitimate identity. During data transmission, the IoT communication unit is configured to prioritize data, sending coagulation warning information via a low-latency channel first, while uploading high-dimensional impedance sequences asynchronously when the system is idle.

6. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The remote data processing server is deployed in a distributed architecture and is equipped with an anomaly detection engine. The anomaly detection engine uses unsupervised learning algorithms based on isolated forests or support vector machines to perform integrity verification on the impedance curves uploaded in real time. When the spectrum shape generated by a single detection exhibits an inductive arc that does not conform to the principles of electrochemicals or a sudden jump in data points, the anomaly detection engine is used to mark the sample as an operational anomaly and send a retest instruction to the front end. The remote data processing server also has a built-in model optimization engine, which is used to perform federated learning on detection data from multiple different geographical locations through an integrated learning framework, so as to periodically update the neural network weight parameters in the cell morphology identification module without disclosing patient privacy. The server is also equipped with a long short-term memory network analysis module, which is used to perform time-series analysis on multiple test records of a single patient. By observing the evolution trajectory of platelet aggregation kinetic parameters within a preset time period, a dynamic assessment trend chart of the patient's coagulation function is generated. The server also has a remote firmware upgrade module, which is used to push algorithm update patches to terminals around the world via an Internet of Things communication unit.

7. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The system also includes a multi-channel analysis terminal, which integrates multiple independent detection stations. Each detection station is configured to independently insert a microfluidic sensor chip to support multiple control experiments on the same patient sample or simultaneous detection of samples from multiple different patients. Each of the aforementioned testing stations is equipped with an independent fluid drive pump for precisely controlling the flow rate of plasma samples within the microfluidic sensor chip; The multi-channel analysis terminal also integrates a high-precision temperature control unit, which includes a heating element and a feedback sensor to precisely maintain the ambient temperature inside the microfluidic sensor chip at 37 degrees Celsius, with a temperature fluctuation range of less than 0.1 degrees Celsius, so as to eliminate the interference of ambient temperature changes on platelet aggregation reaction and impedance baseline. The multi-channel analysis terminal is also equipped with a mechanical oscillation submodule, which is used to apply micro-vibration at a specific frequency to the microfluidic sensor chip before sampling to eliminate gravity interference caused by erythrocyte sedimentation and ensure that the components captured on the electrode surface are the effective components in platelet-rich plasma.

8. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The internal logic of the multi-frequency electrochemical impedance spectroscopy acquisition unit is configured to execute a relaxation time distribution analytical algorithm to convert the impedance spectrum from the frequency domain to the time constant domain. The multi-frequency electrochemical impedance spectroscopy acquisition unit is equipped with a feature extraction subunit, which is used to identify the characteristic peaks in the relaxation time distribution spectrum that represent different physical processes. The characteristic peaks include short time constant peaks that represent the internal polarization of the electrolyte and long time constant peaks that represent the formation of biofilm on the electrode surface. The feature extraction subunit constructs a high-dimensional feature vector with more than 128 dimensions by calculating the peak height, full width at half maximum (FWHM), and center position of the feature peak, which is used to capture weak morphological features masked by impedance amplitude. The multi-frequency electrochemical impedance spectroscopy acquisition unit is also equipped with a baseline self-calibration logic module, which is used to perform two impedance scans in the dry state stage before sample injection and in the buffer filling stage, respectively, to obtain the initial impedance noise floor and distribution parameters of the sensor hardware, and automatically subtract the initial impedance noise floor and distribution parameters from the real-time data in the subsequent plasma detection process to offset the measurement error caused by the fluctuation of electrode manufacturing process or the parasitic inductance of wiring.

9. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 4, characterized in that, The cell morphology identification module adopts a convolutional neural network architecture based on an attention mechanism, which is configured to automatically identify the dimension that contributes the most to cell classification in the high-dimensional feature vector. When distinguishing between platelet and leukocyte complexes, the cell morphology identification module automatically increases the weight of low- and mid-frequency phase angle features in the calculation process; The identification module is also equipped with a pseudo-color heatmap mapping logic circuit. According to the value of the probability distribution function in the output layer of the convolutional neural network model, the logic circuit assigns different color depths to the impedance points corresponding to different frequency bands in the complex impedance plane diagram. The high-frequency region corresponding to the charge transfer resistance is marked as the first color to represent the platelet contribution area, and the low-frequency region corresponding to the diffusion impedance is marked as the second color to represent the fibrin network contribution area. The identification module also integrates a deep visual fusion submodule, which is connected to a miniature optical sensor to collect low-magnification raw image features of the electrode surface. It also achieves feature-level fusion of impedance frequency domain features and macroscopic optical signal fluctuation features through a dual-branch neural network structure to distinguish between simple platelet aggregation with uniform impedance distribution and fibrin filament network with non-uniform impedance distribution.

10. The platelet-rich plasma concentration detection system based on IoT biosensors according to claim 1, characterized in that, The remote data processing server is equipped with a clinical decision support engine, which is connected to a coagulation risk assessment expert database. It is used to receive morphological recognition results and combine them with patient medical record information obtained from external medical systems for comprehensive reasoning. When the clinical decision support engine identifies a fibrin network structure on the electrode surface and the patient is in a preset thrombotic period, the server sends a red warning signal for coagulation risk to the associated mobile terminal through the Internet of Things communication unit. The server is also equipped with an environmental compensation algorithm module, which is used to calibrate the original impedance value in real time based on the ambient temperature, humidity and chip production batch information returned by the IoT communication unit, so as to offset the electrolyte resistance fluctuation caused by environmental temperature drift. The system also includes a distributed storage module, which uses blockchain technology to generate an immutable timestamp and digital signature for each test report. The remote data processing server is also equipped with an integrated learning training center, which is used to aggregate the difficult-to-distinguish sample data reported by each terminal with a recognition confidence level lower than a preset threshold, and retrain the model after remote annotation by experts.

Images