Intelligent cleaning monitoring system and method for ultrasonic image probe

Abstract

The invention relates to the technical field of medical instruments, in particular to an intelligent cleaning monitoring system and method for an ultrasonic image probe, and aims to solve the problems that existing cleaning is lack of quantitative monitoring, subjective in effect, difficult in data tracing and the like. A probe identifier is obtained before cleaning, and a reference map is generated through multispectral pre-scanning; cavitation energy indexes and cleaning fluid parameters are monitored in real time in the cleaning process; performing multispectral scanning after cleaning, and quantifying a residual pollutant index; calculating a cleaning mass fraction by fusing the indexes; and generating a digital cleaning certificate according to the score, and encrypting, storing and tracing. According to the scheme, objective quantification of the effect, real-time process monitoring and whole-process data tracing are realized, and the problems of opaque cleaning, unreliable result, difficulty in responsibility tracing and the like are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical devices, and in particular, relates to an intelligent cleaning and monitoring system and method for an ultrasonic image probe. BACKGROUND

[0002] The technical field of medical devices covers a variety of products such as instruments, equipment, and appliances used for disease diagnosis, prevention, monitoring, treatment, or alleviation. The core content of this technical field is to combine engineering, biology, and medical knowledge to develop high-tech equipment for protecting human health and life safety. As an important part of the modern medical system, the technical level of medical devices is directly related to the accuracy of clinical diagnosis and the effectiveness of treatment. With the continuous progress of material science, electronic technology, and information science, medical devices are developing towards intelligence, miniaturization, and precision, which has a profound impact on many aspects such as clinical medicine, rehabilitation nursing, and public health.

[0003] Among them, the cleaning and disinfection method of the ultrasonic image probe is a key link in the infection control of medical devices. This method aims to completely remove biological tissues, coupling agents, and pathogenic microorganisms remaining on the surface of the probe through physical or chemical means to prevent cross-infection between different patients. In clinical applications, the cleanliness of the ultrasonic probe directly affects medical safety as it directly contacts the patient's skin or mucosa. A complete cleaning and disinfection process usually includes pretreatment, cleaning, disinfection, rinsing, and drying steps to ensure that the probe meets the specified hygiene standards after each use, thereby ensuring the safety and accuracy of subsequent diagnostic work.

[0004] The existing technology adopts a standardized fixed process in the probe cleaning process, lacks real-time evaluation of the initial contamination level of the probe, and cannot dynamically adjust the cleaning parameters according to the actual situation. The verification of the cleaning effect mainly relies on the naked eye observation of the operator or regular sampling detection, lacks objective and quantitative real-time monitoring means, and it is difficult to ensure the uniformity of the quality of each cleaning. Key data such as cleaning liquid concentration, action temperature, and time during the cleaning process are often not systematically recorded and analyzed, leading to the difficulty in tracing the cleaning process and providing data support for process optimization. In addition, the cleaning system and the monitoring system are usually independent of each other, and cannot form a closed-loop feedback control, resulting in the coexistence of resource waste and the risk of incomplete cleaning. These problems are particularly prominent in medical institutions with heavy diagnostic and treatment tasks, directly restricting the improvement of infection control level and equipment turnover efficiency. SUMMARY

[0005] The purpose of the present application is to provide an intelligent cleaning and monitoring system and method for an ultrasonic image probe, which aims to solve the technical problems of lack of objective and quantitative monitoring in the cleaning process, reliance on subjective judgment of cleaning effect, and inability to effectively trace the data of the whole process in the prior art.

[0006] In order to achieve the above object, the present application adopts the following technical scheme: an intelligent cleaning and monitoring method for an ultrasonic imaging probe, comprising the following steps:

[0007] S1: when the cleaning process is started, the unique identity code of the ultrasonic imaging probe to be cleaned is obtained, and the historical cleaning data and the physical model parameters of the probe are called based on the unique identity code; then the multispectral imaging unit is driven to perform pre-cleaning surface scanning on the probe, a baseline image data set covering a preset spectral range is collected, and the initial contamination distribution characteristics of the surface of the probe are extracted through an image processing algorithm to generate a baseline contamination map.

[0008] S2: based on the baseline contamination map, the cleaning operation execution unit is activated, which performs multi-stage physical and chemical cleaning operations including high-pressure spraying, ultrasonic immersion cleaning and high-temperature disinfection in a sealed cleaning cavity according to a preset cleaning protocol; in the ultrasonic immersion cleaning stage, the acoustic sensing unit is used to collect the sound field signals of the liquid medium in the cleaning cavity in real time, and the sound cavity cavitation energy index representing the cavitation effect intensity is calculated through frequency spectrum analysis; at the same time, the chemical sensing unit is used to monitor the key chemical parameters of the cleaning liquid online to generate a real-time fluid state vector.

[0009] S3: after the cleaning operation is completed, the multispectral imaging unit is driven to perform post-cleaning surface scanning on the probe, a finished image data set under the same spectral range and imaging parameters as the baseline image data set is collected; the finished image data set and the baseline image data set are pixel-level registered, and the difference between the two registered data sets is calculated to generate a residual contamination difference image.

[0010] S4: the residual contamination difference image is subjected to image segmentation and morphological analysis to identify and quantify the patch area, gray integral value and shape complexity of the residual contamination, and the surface residual contamination index is calculated according to a preset contamination quantification model; the surface residual contamination index, the sound cavity cavitation energy index and the real-time fluid state vector are fused, and a comprehensive cleaning quality evaluation score is calculated and generated through a multi-dimensional weighted fusion algorithm.

[0011] S5: the comprehensive cleaning quality evaluation score is compared with a preset qualified threshold value, if the score is higher than the threshold value, it is determined that the cleaning is qualified, a digital cleaning certificate containing the unique identity code, time stamp, monitoring data of each stage and final evaluation score is generated by the system, the certificate is encrypted and bound with the unique identity code of the probe, and stored in a tamper-proof distributed database; if the score is lower than the threshold value, it is determined that the cleaning is unqualified, and the system automatically triggers a new round of cleaning operation.

[0012] As an embodiment of the present application, the baseline contamination map includes an initial contamination coordinate matrix, a contamination spectral response curve, and an initial contamination load index; the acoustic cavity cavitation energy index includes an ultrasonic wave main frequency energy value, a harmonic energy distribution, and a broadband noise integral energy value; the real-time fluid state vector includes a conductivity, a pH value, a temperature, and a light absorption at a specific wavelength of the cleaning liquid; the surface residual contamination index includes a contamination coverage, an average residual thickness estimate, and a contamination form dispersion; and the digital cleaning certificate includes a probe RFID / NFC serial number, a cleaning start and end timestamp, a pre-cleaning and post-cleaning image digest, an acoustic cavity cavitation energy index time series, a fluid state vector history record, a final cleaning quality evaluation score, and an operator digital signature.

[0013] As a further embodiment of the present application, the specific process of generating the baseline contamination map in step S1 is as follows:

[0014] S111: A radio frequency identification (RFID) reader or near field communication (NFC) module is used to read an electronic tag attached to the probe handle or cable connector to obtain a globally unique probe serial number (US-ID); the US-ID is used as an index to search the model, geometric size data, material information, and past cleaning and use records of the probe in the device history database.

[0015] S112: A two-dimensional rotary translation mechanism controlled by a stepper motor built into the cleaning cavity is driven to accurately control the movement of the probe in the field of view of the multi-spectral imaging unit; the multi-spectral imaging unit includes a complementary metal-oxide-semiconductor (CMOS) image sensor with a resolution of no less than 3840x2160 pixels, and a ring-shaped light source system arranged around the probe array; the light source system is composed of multiple groups of independently controllable light-emitting diodes (LEDs) with emission wavelengths covering 365 nanometer ultraviolet light (UV-A), 450 nanometer blue light, 525 nanometer green light, 630 nanometer red light, and 940 nanometer infrared light.

[0016] S113: The system sequentially illuminates the LED light sources of different wavelengths and collects a high-resolution image at each wavelength to form a baseline image dataset; then, a semantic segmentation model based on a convolutional neural network (CNN) is applied to process the image dataset, identify and segment potential contamination regions attached to the probe acoustic window and shell surface with specific spectral response characteristics, such as the fluorescence reaction of protein residues under ultraviolet light or the phosphorescence effect of specific coupling agents under blue light excitation; finally, the coordinates, area, and average pixel intensity under each spectrum of all segmented contamination regions are encoded to generate a structured baseline contamination map.

[0017] As a further embodiment of the present application, the specific process of calculating the cavitation energy index of the acoustic cavity and generating the real-time fluid state vector in step S2 is as follows:

[0018] S211: A broadband hydrophone with a center frequency of 2.5 megahertz (MHz) is symmetrically installed on the inner wall of the cleaning cavity; during the ultrasonic immersion cleaning stage, the hydrophone continuously collects the acoustic field pressure signal at a sampling rate of no less than 5 megasamples per second (MS / s); the collected time-domain signal is sent to a field programmable gate array (FPGA) implemented signal processing unit.

[0019] S212: The signal processing unit performs a short-time Fourier transform (STFT) with a length of 4096 points on the collected time-domain signal, continuously generating acoustic field spectra with a time resolution of 10 milliseconds; for each spectral frame, the system calculates the cavitation energy index (CEI) of the acoustic cavity, which is mathematically defined as:

[0020]

[0021] where P(f) is the power spectral density function, the integral interval , is a broadband noise evaluation band excluding the fundamental frequency and its low-order harmonics of the ultrasonic transducer (for example, from 100 kilohertz to 800 kilohertz), and the denominator is the total energy of the full frequency band; the system monitors the CEI value in real time, and if the average value in the last 1 second is lower than the preset cavitation efficiency threshold, the output power or frequency of the ultrasonic generator is automatically adjusted to maintain the optimal cleaning state.

[0022] S213: A microfluidic sensing chip is integrated in the cleaning liquid circulation loop, which has a four-electrode sensor for measuring conductivity, an ion-selective field effect transistor (ISFET) based pH sensor, a platinum resistance temperature sensor (PT100), and a small optical absorption spectrometer etched on the chip; the system collects readings from these sensors every 5 seconds to form a four-dimensional real-time fluid state vector , which is compared with the initial state vector to calculate the fluid degradation index, ensuring that the chemical effectiveness of the cleaning liquid is within the effective range, where is the electrical conductivity measured at time , which reflects the change in ion concentration in the cleaning liquid. A decrease in conductivity may indicate the consumption or dilution of active ingredients; an increase may indicate the accumulation of impurities or the generation of by-products; is the pH value measured at time The pH value (negative logarithm of hydrogen ion concentration) measured at any given time reflects changes in the acidity or alkalinity of the cleaning solution. Many cleaning solutions are pH sensitive, and pH shifts may indicate decreased chemical stability or the formation of reaction byproducts. In time The temperature measured at any given time reflects how temperature affects chemical reaction rates and sensor responses. Simultaneously, temperature drift may affect other sensor readings (such as conductivity and pH), therefore it needs to be compensated for or considered as a state variable. In time The absorbance value measured at specific times is a signal from a small optical absorption spectrometer, used to monitor the concentration of specific chemicals in the cleaning solution (such as dyes, oxidants, decomposition products, etc.). Absorption at different wavelengths can reflect the presence and changes of specific components; It is in time The four-dimensional real-time fluid state vector comprehensively reflects the current electrochemical and optical properties of the cleaning fluid.

[0023] As a further embodiment of the present invention, the specific process for calculating the surface residual contaminant index in steps S3 and S4 is as follows:

[0024] S311: After cleaning, the same scanning path and imaging parameters as in step S112 are used to acquire the completed image dataset; an image registration algorithm based on scale-invariant feature transform (SIFT) is applied to calculate the geometric transformation matrix between the reference image dataset and the completed image dataset to achieve precise alignment at the sub-pixel level.

[0025] S411: For each spectral band, the registered completed image is subtracted pixel by pixel from the reference image, and the absolute value is taken to generate a series of difference images; an adaptive threshold segmentation algorithm (such as the Otsu method) is applied to the difference images to extract the regions with pixel values ​​higher than the background noise level, and these regions are identified as residual pollutant patches.

[0026] S412: Calculate the total area of ​​all identified residual contaminant patches. and the area of ​​each patch. ,perimeter and average gray value Surface residual contaminant index ( It is calculated using the following formula:

[0027]

[0028] in It is the total surface area of ​​the probe. It is the total number of plaques. , , are the weighting coefficients corresponding to the contamination coverage, the average residual thickness, and the morphology dispersion, respectively, and the sum of them is 1.

[0029] As a further embodiment of the present application, the specific process of calculating the comprehensive cleaning quality score in step S4 is as follows:

[0030] S421: The calculated surface residual contamination index (RCI), the time series integral value of the acoustic cavity cavitation energy index (CEI) in the ultrasonic immersion cleaning stage, and the norm of the fluid state vector deviating from the initial state are normalized so that their numerical ranges are mapped to the [0, 1] interval. The normalized indicators are respectively denoted as , , and .

[0031] S422: The comprehensive cleaning quality score (CQS) is calculated by a weighted summation model, and its mathematical expression is:

[0032]

[0033] wherein, is the normalized surface residual contamination index, full name: Residual Contamination Index (RCI), source: the surface residual contamination concentration (such as particles, organic matter, ions, etc.) measured by sensors or detection means, time series integral value: indicating the cumulative amount of residual contamination (may be the integral of continuous measurement data) in the entire cleaning process, normalization: mapping the original RCI value to the [0, 1] interval, 0→ completely no residual (ideal cleanliness), 1→ severe residual (unwashed or cleaning failure); the (1- ) term converts the contamination index into a cleanliness score, that is, the lower the RCI, the higher the cleanliness score; is the normalized cavitation energy index, full name: Cavitation Energy Index (CEI), source: the cavitation effect intensity (i.e. the energy of bubble generation and collapse) generated by the acoustic field in the ultrasonic cleaning stage, time series integral value: reflecting the total accumulation of cavitation energy in the entire ultrasonic cleaning stage, normalization: mapping to [0, 1], 0→ no cavitation effect (ineffective cleaning), 1→ maximum cavitation intensity (ideal physical cleaning condition), the higher the value directly representing the intensity and efficiency of the physical cleaning process, the stronger the physical action and the greater the cleaning potential; Fluid Degradation Index (FDI) for normalized fluid state deviation index, full name: Fluid Degradation Index (Fluid Degradation Index), source: state vector constructed based on multi-parameter (conductivity, pH, temperature, light absorption) collected by microfluidic chip , and the "deviation degree" obtained by comparing with the initial state ; the calculation method is, for example, using Euclidean distance, Manhattan distance or weighted distance:

[0034]

[0035] Then normalized to [0, 1], 0→fluid state unchanged (fresh), 1→seriously degraded (failure), , the effectiveness score of the cleaning fluid, the fresher the fluid→the smaller the FDI→the higher the effectiveness score, if the cleaning fluid is exhausted or deteriorated, even if the process is strong, it cannot be effectively cleaned; the weight coefficients α, β, γ are preset parameters, reflecting the relative importance of surface cleanliness, cleaning process intensity and cleaning medium quality in the final evaluation, and satisfying α+β+γ=1. For example, in an application scenario with the highest priority of final cleanliness, α can be set to 0.7, β to 0.2, and γ to 0.1.

[0036] The present application also provides an intelligent cleaning and monitoring system for an ultrasonic imaging probe, which is used to execute the above method, and the system comprises:

[0037] A probe identity and state recognition module, which is integrated with a radio frequency identification reader and a near field communication module, is used to read the unique identity code of the probe and communicate with the background database to retrieve the historical data and physical model parameters of the probe.

[0038] A multispectral imaging analysis module, which comprises a high-resolution CMOS image sensor, a program-controlled light source system composed of a multi-wavelength LED array, and a precise two-dimensional rotary translation mechanism, is used to scan the surface of the probe before and after cleaning, collect the reference image data set and the finished image data set, and calculate and generate the reference contaminant spectrum and the surface residual contaminant index through the built-in image processing unit.

[0039] A cleaning operation and process monitoring module, which comprises a sealed cleaning cavity, a high-pressure spraying system, an ultrasonic generator and a transducer, a heating and disinfecting assembly; the module is further integrated with a wideband hydrophone and a microfluidic chemical sensing chip, which are used to monitor the acoustic cavity cavitation energy index and the real-time fluid state vector of the cleaning fluid during the cleaning process.

[0040] Data fusion and decision authentication module, which is an embedded computing unit, runs a multi-dimensional data fusion algorithm inside, integrates various quantitative indicators from the multi-spectral imaging analysis module and the cleaning operation and process monitoring module, calculates the final cleaning quality evaluation score, and makes a pass or fail judgment according to the comparison result of the score and the preset threshold, and finally generates or rejects to generate a digital cleaning certificate.

[0041] Data recording and tracing interface module, which is responsible for encrypting and time-stamping the generated digital cleaning certificate together with all the original data and analysis results of the process monitoring, and writing it into a distributed ledger database based on blockchain technology, ensuring the integrity and tamper resistance of the data; the module also provides an interface conforming to the HL7 or DICOM standard for secure data exchange with the hospital information system (HIS) or the medical image archive and communication system (PACS).

[0042] Compared with the prior art, the present application has the following advantages:

[0043] The present application introduces multi-spectral imaging technology and quantitative image analysis algorithms, transforming the detection of probe surface contaminants from subjective human eye judgment to objective and repeatable machine vision quantitative evaluation, greatly improving the accuracy and reliability of cleaning effect verification. Further, the present application not only evaluates the cleaning result, but also monitors and controls the cleaning process itself in real time through online monitoring of the cavitation effect of the acoustic cavity and the chemical composition of the cleaning liquid, ensuring that the cleaning operation is always performed under optimal conditions. The system integrates probe identity recognition, pre-cleaning state evaluation, cleaning process dynamic monitoring, cleaning result quantitative verification, and data encryption and evidence tracing into a fully automatic closed-loop process, generating a digital certificate containing physical, chemical, and acoustic evidence for each cleaning operation, thereby fundamentally solving the core problems of process opacity, unreliable results, and difficult responsibility tracing in the field of medical instrument cleaning and disinfection. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 The overall flowchart of the intelligent cleaning and monitoring method for ultrasonic imaging probes provided by the embodiments of the present application is shown in the figure.

[0045] Figure 2 The functional module diagram of the intelligent cleaning and monitoring system for ultrasonic imaging probes provided by the embodiments of the present application is shown in the figure.

[0046] Figure 3 The detailed flowchart of the pre-cleaning surface scanning and analysis step for the probe in the embodiments of the present application is shown in the figure.

[0047] Figure 4A flowchart of the process of on-line monitoring and data collection in the cleaning process in the embodiment of the present application is shown in the figure.

[0048] Figure 5 A flowchart of the process of quantitative evaluation of cleaning effect and decision authentication in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0049] In order to make the objectives, technical solutions and advantages of the present application clearer and more comprehensible, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application. In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection" and "linking" should be understood in a broad sense, for example, they can be fixed connection, or detachable connection, or integrally connected; they can be mechanical connection, or electrical connection; they can be directly connected, or indirectly connected through an intermediate medium. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0050] Please refer to Figure 1 The intelligent cleaning monitoring method for ultrasonic image probe provided by the embodiment of the present application comprises the following steps:

[0051] Step S1, when the cleaning process is started, the unique identity code of the ultrasonic image probe to be cleaned is acquired, and the historical cleaning data and the physical model parameters of the probe are called based on the unique identity code; then the multi-spectral imaging unit is driven to perform pre-cleaning surface scanning on the probe, the baseline image data set covering the preset spectral range is collected, and the initial pollutant distribution characteristics of the probe surface are extracted through image processing algorithm to generate a baseline pollutant map.

[0052] Step S2, based on the baseline pollutant map, the cleaning operation execution unit is activated, which performs multi-stage physical and chemical cleaning operations including high-pressure spraying, ultrasonic immersion cleaning and high-temperature disinfection in a sealed cleaning cavity according to a preset cleaning protocol; in the ultrasonic immersion cleaning stage, the acoustic sensing unit is used to collect the sound field signal of the liquid medium in the cleaning cavity in real time, and perform frequency spectrum analysis thereon to calculate the sound cavity cavitation energy index representing the cavitation effect intensity; at the same time, the chemical sensing unit is used to monitor the key chemical parameters of the cleaning liquid on-line to generate a real-time fluid state vector.

[0053] Step S3, after the cleaning operation is completed, driving the multi-spectral imaging unit to perform a post-cleaning surface scan of the probe, collecting a finished image dataset under the same spectral range and imaging parameters as the reference image dataset; performing pixel-level registration between the finished image dataset and the reference image dataset, and calculating the difference between the two registered datasets to generate a residual contaminant difference image.

[0054] Step S4, performing image segmentation and morphological analysis on the residual contaminant difference image, identifying and quantifying the patch area, gray integral value and shape complexity of the residual contaminant, and calculating the surface residual contaminant index according to a pre-set contaminant quantification model; fusing the surface residual contaminant index, the acoustic cavity cavitation energy index, and the real-time fluid state vector, and calculating a comprehensive cleaning quality evaluation score through a multi-dimensional weighted fusion algorithm.

[0055] Step S5, comparing the comprehensive cleaning quality evaluation score with a pre-set qualified threshold, determining the cleaning result, and generating a digital cleaning certificate or triggering a new round of cleaning operation according to the determination result.

[0056] In one specific embodiment, the reference contaminant map is a structured data object containing fields: initial contaminant coordinate matrix, contaminant spectral response curve, and initial contaminant load index. The acoustic cavity cavitation energy index is a scalar value calculated based on the ultrasonic wave main frequency energy value, harmonic energy distribution, and broadband noise integral energy value. The real-time fluid state vector is a multi-dimensional vector whose dimensions include the conductivity, pH value, temperature, and light absorption at a specific wavelength of the cleaning fluid. The surface residual contaminant index is a comprehensive index composed of contaminant coverage, average residual thickness estimate, and contaminant morphology dispersion. The digital cleaning certificate is an encrypted data packet containing the probe RFID / NFC serial number, cleaning start and end time stamp, pre-cleaning and post-cleaning image digest, acoustic cavity cavitation energy index time series, fluid state vector history record, final cleaning quality evaluation score, and operator digital signature.

[0057] Please refer to Figure 3 which shows the specific process of generating the reference contaminant map in step S1. In one embodiment, the process can be further divided into the following sub-steps:

[0058] Step S111, read the electronic tag attached to the probe handle or cable connector through a radio frequency identification (RFID) reader or near field communication (NFC) module, and obtain its globally unique probe serial number (US-ID). The US-ID is used as the primary key for index query in the device history database. The device history database is a relational database, and its table structure includes but is not limited to the following fields: us_id (VARCHAR, PRIMARYKEY), probe_model (VARCHAR), geometry_data (JSONB), material_spec (TEXT), last_cleaned_timestamp (TIMESTAMP), total_usage_cycles (INT), last_cleaning_certificate_hash (VARCHAR).

[0059] Specifically, when a C5-2 abdominal probe (US-ID: PROBE-C52-SN789012) is placed in the cleaning device, the RFID reader works at a frequency of 13.56 MHz to read the US-ID stored in the electronic tag memory. The system queries the database with this ID to retrieve the geometry_data field. This field stores a JSON object that describes the physical dimensions and scan path planning parameters of the probe, for example: {"model":"C5-2","type":"convex","face_radius_mm":60,"body_length_mm":180,"scan_profile":{"rotation_axis":"longitudinal","rotation_total_angle_deg":360,"rotation_step_deg":5,"translation_axis":"longitudinal","translation_distance_mm":180,"translation_speed_mm_s":15}}. At the same time, the system retrieves the historical cleaning records of the probe for subsequent trend analysis.

[0060] As an alternative, the probe identification can be achieved by an industrial grade 2D barcode reader integrated inside the cleaning chamber. The probe handle is engraved with a DataMatrix 2D barcode, which encodes the US-ID. At the start of the process, an auxiliary light source illuminates the 2D barcode area, a camera captures the image, and a software module based on an open source decoding library (e.g. libdmtx) resolves the US-ID. The constraint of this solution is the requirement of a clean and unobstructed 2D barcode surface, with high positioning accuracy.

[0061] As another alternative, in the exceptional case of RFID or 2D barcode reading failure, the system can switch to a manual input mode. In this case, the touchscreen interface prompts the operator to select from a pre-defined list of probe models and manually enter the serial number. To reduce input errors, the system is equipped with checksum algorithms for specific manufacturer serial number formats. For example, after entering PROBE-C52-SN789012, the system verifies the format and checksum correctness according to pre-defined rules. If the US-ID does not exist in the database, the system triggers the "New Probe Registration" process, guiding the operator to confirm the probe model and create a new database entry.

[0062] Step S112, drive a two-dimensional rotary translation mechanism built-in the cleaning chamber, controlled by a stepper motor, to precisely control the movement of the probe within the field of view of the multi-spectral imaging unit. The two-dimensional rotary translation mechanism includes a rotary shaft and a linear translation shaft. The rotary shaft is controlled by a stepper motor with a 1.8-degree step angle, configured with a 16-fold micro-step driver, achieving an angular resolution of 360 / (1.8 / 16)=3200 micro-steps / revolution. The translation shaft is driven by another stepper motor through a ball screw, achieving a linear resolution of 0.01 millimeter / step. The multi-spectral imaging unit includes a back-illuminated CMOS image sensor with a resolution of 3840x2160 pixels, and a ring-shaped light source system arranged around the probe array. The light source system is composed of five groups of independently controllable high-power LEDs, with central emission wavelengths of 365 nanometers (UV-A), 450 nanometers (blue light), 525 nanometers (green light), 630 nanometers (red light), and 940 nanometers (near-infrared light), respectively.

[0063] Specifically, based on the scan_profile data obtained in step S111, the system generates a motion instruction sequence for the probe of model C5-2. The instruction sequence drives the rotary motor to rotate the probe with an increment of 5 degrees per step, a total of 72 steps to complete 360 degrees of rotation. Between every two rotary steps, the translation shaft uniformly translates the probe at a speed of 15 millimeters / second for 2.5 millimeters (180 mm / 72 steps). This coordinated motion ensures that the entire cylindrical surface of the probe can be completely covered by the imaging unit in a spiral scanning trajectory.

[0064] Step S113, the system sequentially turns on the LED light sources of different wavelengths at each scanning position point, and collects a frame of high-resolution image under each wavelength to form a reference image dataset. The parameters of the collection sequence, such as the driving current of each wavelength LED and the exposure time of the CMOS sensor, are retrieved from a pre-set “imaging profile” according to the probe model and the common types of contaminants. For example, a profile for detecting biofilm and protein residues may be set as follows: the 365 nm LED is driven at 100% duty cycle, and the exposure time is 200 ms to maximize the signal-to-noise ratio of the fluorescent signal; while the 630 nm red light LED is driven at 50% duty cycle, and the exposure time is 20 ms to obtain the reference image of the surface topography. In a complete scanning cycle, if there are 72 scanning position points, the system will collect 72*5=360 frames of original images to form the reference image dataset.

[0065] Subsequently, a semantic segmentation model based on convolutional neural network (CNN) is applied to process the image dataset. The CNN model adopts the U-Net architecture, and its input layer receives a 5-channel image tensor (each channel corresponds to a spectral band), and the spatial resolution is uniformly pre-processed to 1024x1024 pixels. The network passes through an encoder-decoder structure, fuses multi-scale features through skip connection, and finally outputs a single-channel probability map. The value of each pixel in the map (between 0 and 1) represents the probability of the pixel belonging to the contaminant. The model has been trained offline on a probe image dataset containing tens of thousands of images manually labeled by experts with various types of contaminants (such as coupling agent residues, bloodstains, gels, protein films, etc.). The loss function used in the training process is the weighted sum of the Dice loss and the binary cross-entropy loss to balance the requirements for small target detection and pixel-level classification accuracy.

[0066] Specifically, when processing a frame of five-channel image, the probability map output by the CNN model is binarized by a fixed threshold (e.g. 0.5) to generate a contaminant mask image. The system stitches all the mask images generated at the scanning position points into a complete contaminant spread map of the probe surface through coordinate transformation. Finally, each independent connected domain (i.e. contaminant patch) in the spread map is analyzed to extract its centroid coordinates, bounding rectangle, pixel area, and average pixel intensity value in the five original spectral channels. These structured information is encoded to generate a reference contaminant map.

[0067] An example of the data structure of a reference contaminant map is as follows:

[0068] JSON

[0069] 1{

[0070] 2 "us_id": "PROBE-C52-SN789012",

[0071] 3 "timestamp": "2024-05-21T09:30:15Z",

[0072] 4 "initial_contaminant_load_index": 0.76,

[0073] 5 "contaminant_blobs": [

[0074] 6 {

[0075] 7 "blob_id": 1,

[0076] 8 "bounding_box": [1024, 512, 150, 80],

[0077] 9 "pixel_area": 9870,

[0078] 10 "spectral_signature": {

[0079] 11 "uv_365nm_intensity": 225.4,

[0080] 12 "blue_450nm_intensity": 98.1,

[0081] 13 "green_525nm_intensity": 110.5,

[0082] 14 "red_630nm_intensity": 105.2,

[0083] 15 "ir_940nm_intensity": 70.8

[0084] 16},

[0085] 17 "classification_probability": {

[0086] 18 "protein": 0.92,

[0087] 19 "gel": 0.05,

[0088] 20 "other": 0.03

[0089] 21}

[0090] 22} 23 ]

[0092] 24}

[0093] where initial_contaminant_load_index is a comprehensive contaminant load index, which is calculated by the total contaminant coverage area and the spectral characteristics of each patch weighted.

[0094] Referring to Figure 4 , which shows the specific process of online monitoring and data collection for the cleaning process in step S2. In an embodiment, the process can be further divided into the following sub-steps:

[0095] Step S211, one or more broadband hydrophones are symmetrically installed on the inner wall of the cleaning cavity. In a specific embodiment, two piezoelectric ceramic hydrophones with a center frequency of 2.5 megahertz (MHz) and a bandwidth of 3 megahertz are used, respectively installed on the opposite two side walls of the cavity to obtain more comprehensive sound field information. In the ultrasonic immersion cleaning stage, the ultrasonic wave generator drives a plurality of piezoelectric transducer arrays attached to the bottom of the cavity to work at a fundamental frequency of 40 kilohertz (kHz). The hydrophone continuously collects sound field pressure signals at a sampling rate of no less than 5 megasamples per second (MS / s). The choice of this sampling rate is based on the Nyquist sampling theorem, which ensures that all signal components within the working bandwidth of the hydrophone can be collected without distortion. The collected digital signal stream is sent to a special signal processing unit based on a field programmable gate array (FPGA). FPGA is chosen because of its inherent parallel processing capability and low delay characteristics, which can meet the real-time and high-throughput spectral analysis requirements of the sound field signal. The internal logic of the FPGA is divided into a data acquisition interface module, a dual-port RAM buffer module, and a hardware-implemented fast Fourier transform (FFT) core.

[0096] Step S212, the signal processing unit performs continuous short-time Fourier transform (STFT) on the collected time-domain signal stream. Specifically, the FPGA frames the signal stream, with each frame containing 4096 sampling points and a 50% overlap (i.e. 2048 sampling points) between frames. A Hamming window function is applied to each frame of signal to suppress spectral leakage, and then sent to the 4096-point FFT core for transformation, continuously generating sound field power spectrum with a time resolution of about 10 milliseconds. For each calculated power spectrum density function P(f), the system calculates the cavitation energy index (CEI) of the sound cavity. This index is defined as the ratio of the energy in a specific wideband noise frequency band to the total energy of the full frequency band, and its mathematical expression is:

[0097]

[0098] The source and selection of the parameters have clear physical meaning. The integration interval , The choice of the band [150kHz, 800kHz] is intended to capture the broadband noise generated by the collapse of cavitation bubbles, while avoiding direct interference with the fundamental frequency of the ultrasonic transducer (40kHz) and its higher harmonics (80kHz, 120kHz,...). In a typical configuration, this band is set to [150kHz, 800kHz]. The denominator in the fraction corresponds to the upper limit of the effective bandwidth of the hydrophone, e.g. 1.5 MHz.

[0099] The calculation of the CEI is illustrated by a simulated example: at a certain time t, the integrated energy of the power spectrum computed by the FPGA over the full band [0, 1.5MHz] is 1.52 Watt. The integrated energy over the broadband noise evaluation band [150kHz, 800kHz] is 0.58 Watt. The cavitation energy index at this moment is CEI = 0.58 / 1.52 ~ 0.382. The system monitors the CEI value in real time. If the sliding average of the CEI over the last 1 second is below a pre-set cavitation efficiency threshold (e.g. 0.30), a closed-loop feedback control is triggered. The threshold is determined experimentally during the development phase of the system, and is associated with the optimal cleaning effect for a specific combination of cleaning agent and contaminants. The feedback control logic first attempts to increase the output power of the ultrasonic generator by 5%, and monitors the change in CEI. If the CEI value does not significantly increase, the system enters a frequency sweep mode, sweeping the frequency over the range 40kHz ± 2kHz with a step of 0.1kHz, aiming to break the standing waves in the acoustic field and activate the cavitation effect in the previously inert regions. If the CEI remains below the threshold after three consecutive adjustments, the system records a "low cavitation efficiency" warning and notifies the upper control system.

[0100] Step S213, integrate a multi-parameter online monitoring module in the circulation loop of the cleaning liquid. In one embodiment, the module is a flow cell integrated with a microfluidic sensing chip. The chip is etched with multiple sensors through micro-electro-mechanical system (MEMS) technology: a four-electrode platinum sensor for measuring conductivity, a pH sensor based on ion-selective field effect transistor (ISFET), a high-precision platinum resistance temperature sensor (PT100), and a small optical absorption spectrometer integrated with a miniature grating and a linear photodiode array. The system polls the readings of these sensors every 5 seconds, forming a four-dimensional real-time fluid state vector where is the optical absorbance at a specific wavelength (e.g. 280nm, for detecting protein dissolution).

[0101] Before the start of the cleaning cycle, the system collects an initial state vector V_0 of the cleaning liquid. For example, the initial vector of a freshly prepared alkaline cleaning liquid can be [ = 12.5 mS / cm, pH = 11.2, T = 25.1 °C, = 0.02 AU]. At the 10th minute of the cleaning, the real-time vector collected was [ = 16.8 mS / cm, pH = 10.6, T = 55.3 °C, = 0.45 AU]. The changes in the vector reflect the chemical dynamics of the cleaning process: the rise in conductivity and the significant increase in optical absorbance indicate the dissolution of contaminants (e.g., salts and proteins) from the probe surface into the cleaning fluid; the decrease in pH value indicates the consumption of the alkaline active ingredient.

[0102] Based on the difference between the real-time vector and the initial vector, the system calculates the Fluid Degradation Index (FDI) to quantify the performance decay and contamination level of the cleaning fluid. The FDI can be defined as the normalized vector variation norm:

[0103]

[0104] where is the normalized weight vector of the parameters, whose values are set according to the importance of different chemical parameters on the cleaning performance. If the FDI exceeds a pre-set degradation threshold (e.g., 0.85), the system will determine that the current cleaning fluid has failed and will be forcibly locked after the completion of the current cleaning task, prohibiting its use in new cleaning cycles, while generating a maintenance alert of “Please replace the cleaning fluid”.

[0105] As an alternative, the online monitoring module can be composed of discrete industrial-grade sensors. The cleaning fluid is pumped into a serial measurement pipeline, in which a loop inductance conductivity meter, a glass electrode pH meter with temperature compensation, an armored PT100 temperature meter, and an external ultraviolet-visible spectrophotometer with a quartz flow cell are installed in sequence. The sensors of this scheme are more robust and durable, easy to calibrate and replace individually, but the disadvantage is that the system integration is more complex, and the pipeline dead volume is larger, resulting in a relatively long response time.

[0106] Please refer to Figure 5 , which shows the detailed process of performing quantitative evaluation and decision authentication of cleaning effect in the embodiments of the present application, which covers the core operations of steps S3, S4 and S5. In a specific embodiment, the specific process of calculating the surface residual contaminant index in steps S3 and S4 can be further divided into the following sub-steps:

[0107] Step S311, after the cleaning operation is completed and the probe is preliminarily dried, the same scanning path, motion command sequence, light source parameters and imaging parameters as in step S112 are used to drive the two-dimensional rotary translation mechanism and the multi-spectral imaging unit to perform a post-cleaning surface scan of the probe, and a complete set of finished image data is acquired. This strict parameter reproducibility is the basis for ensuring the effectiveness of subsequent differential analysis. Subsequently, the system applies a scale-invariant feature transform (SIFT)-based image registration algorithm to accurately align each frame of the finished image data set with its corresponding frame in the reference image data set.

[0108] Specifically, for each spectral band image pair (e.g., the pre-cleaning 365 nm UV image and the post-cleaning 365 nm UV image), the SIFT algorithm first detects scale space extrema in both images independently as keypoints. Then, a 128-dimensional local image descriptor is computed for each keypoint, which is invariant to illumination changes, rotations and scale changes. The system finds the initial matching point pairs using the nearest neighbor matching strategy by computing the Euclidean distance between the keypoint descriptors of the two images. To eliminate false matches due to surface changes (contaminants being removed) or noise, the system uses the random sample consensus (RANSAC) algorithm to estimate a robust geometric transformation model. The RANSAC algorithm iteratively selects a minimum subset (e.g., 4 pairs of points) from the matching point pairs, computes a perspective transformation matrix, and then uses this matrix to verify all other matching point pairs, counting the number of inliers. After several hundred iterations, the transformation matrix with the most inliers is selected as the final transformation model. Finally, the system applies this transformation matrix to the finished image to generate a new image that is resampled and accurately aligned with the reference image at the sub-pixel level using bilinear or bicubic interpolation methods. This process is independently performed for the five spectral band images.

[0109] Step S411, for each spectral band, the pixel-by-pixel gray value difference between the registered finished image and the corresponding reference image in step S311 is calculated and the absolute value is taken, generating a series of (five in this embodiment) difference images. The pixel values in the difference images directly reflect the amount of change in the surface reflection or fluorescence characteristics before and after cleaning, and high value areas correspond to successfully removed contaminants. Subsequently, in order to accurately separate the signal of residual contaminants from background noise, the system applies an adaptive threshold segmentation algorithm to each difference image. In a preferred embodiment, Otsu's method is used. This algorithm automatically finds a best global threshold by maximizing the inter-class variance, dividing the image pixels into foreground (residual contaminants) and background.

[0110] Specifically, for the UV band difference images, their gray scale histograms can exhibit a bimodal distribution, one peak corresponding to the background noise and unchanged regions, and the other peak corresponding to the regions where the fluorescent signal is greatly reduced after the contaminants are removed. The Otsu algorithm computes a gray scale value that maximizes the variance between these two classes, e.g., T = 42. All regions with pixel values above this threshold T are preliminarily identified as the removed contaminant regions. Conversely, if the original post-cleaning image is analyzed, the identification logic for the residual contaminant regions is reversed. To further improve the segmentation accuracy, the system can perform a morphological post-processing operation, e.g., applying a 3x3 pixel structuring element for morphological opening (erosion followed by dilation) to remove isolated noisy pixels, and then morphological closing (dilation followed by erosion) to fill possible small holes inside the residual contaminant patches.

[0111] Step S412, after obtaining the residual contaminant binary mask images for all spectral bands, the system performs a logical OR operation on these mask images to fuse them into a final residual contaminant distribution map. Subsequently, all independent connected regions (i.e., residual contaminant patches) in this distribution map are quantitatively analyzed. The system computes the geometric and photometric properties of each patch, including its pixel area , perimeter , and the corresponding average gray scale value in each difference image. Based on these measurements, the surface residual contaminant index (RCI) is computed. This index is a dimensionless comprehensive score, whose formula is defined as:

[0112]

[0113] Each of the three components of this formula has a clear physical meaning. The first term,

[0114]

[0115] represents the contaminant coverage, where is the total area of all residual patches, and is the total surface area of the probe (obtained from the physical model parameters in step S111). This term reflects the extent of the contamination.

[0116] The second term,

[0117]

[0118] represents the area-weighted average gray scale of the contaminant patches, which can be viewed as an estimate of the average thickness or density of the residual contaminants. The higher the gray scale value of a patch in the difference image, the more severe the original contamination before it was removed, and the more risky its residual part can be. This term reflects the depth of the contamination.

[0119] Third term,

[0120]

[0121] representing the dispersion or irregularity of the shape of the contaminant. Among them,

[0122]

[0123] is the isoperimetric quotient of the individual patch, which takes a value in the interval (0, 1], 1 representing a perfect circle. Irregular, edge-complex patches (such as biofilm colonies) have lower isoperimetric quotient values. This term, in the form of 1-..., assigns higher penalty values to patches with more irregular shapes, as such residues are more difficult to clean and more likely to harbor microorganisms. N is the total number of patches.

[0124] 、 、 are weighting coefficients corresponding to coverage, average residue thickness, and shape dispersion, respectively, whose sum is 1. These weights are set according to risk assessment, for example, for scenarios where biofilm residues are of particular concern, a higher value (such as 0.5) can be assigned.

[0125] In one specific calculation example, suppose the analysis yields: the total surface area of the probe = 12000 mm2, and a total of N = 3 residual patches are found.

[0126] Patch 1: = 10 mm2, = 15 mm, I_avg1 = 150.

[0127] Patch 2: = 5 mm2, = 10 mm, I_avg2 = 120.

[0128] Patch 3: = 2 mm2, = 8 mm, I_avg3 = 200.

[0129] Let the weights be = 0.4, = 0.4, = 0.2.

[0130] The total residual area = 10 + 5 + 2 = 17 mm2.

[0131] The coverage term = 0.4 * (17 / 12000) ≈ 0.00057.

[0132] Thickness term = 0.4*((10150+5120+2200) / 17) = 0.4*(2500 / 17)≈58.82. For ease of fusion, this term needs to be normalized, assuming its maximum possible value is 1000, then the normalized value is 58.82 / 1000 = 0.05882.

[0133] Circularity calculation for the morphology term:

[0134] = 4π10 / 15²≈0.558

[0135] = 4π5 / 10²≈0.628

[0136] = 4π2 / 8²≈0.393

[0137] Average circularity = (0.558+0.628+0.393) / 3≈0.526

[0138] Morphology term = 0.2*(1-0.526) = 0.0948.

[0139] Final RCI (before normalization, sum of all components) = 0.00057+0.05882+0.0948≈0.154. This value will enter the next step of normalization and fusion calculation.

[0140] As a subsequent part of step S4, the specific process of calculating the comprehensive cleaning quality assessment score can be further decomposed into the following sub-steps:

[0141] Step S421, normalize the surface residual contamination index (RCI) calculated in the last step, the time series integral value of the cavitation energy index (CEI) of the entire ultrasonic immersion cleaning phase recorded in step S212, and the final fluid degradation index (FDI) calculated in step S213. The purpose of normalization is to map these indicators of different sources and dimensions to the unified [0,1] interval, so as to perform fair weighted fusion.

[0142] Specifically, the normalization adopts the Min-Max Scaling method. For RCI, its normalization formula is = (RCI-RCI_min) / (RCI_max-RCI_min). Where RCI_min is theoretically 0 (completely clean), and RCI_max is a preset upper limit value representing an unacceptable pollution level, which is set by experimental data statistics or according to clinical risk standards. For the time series integral value of CEI (denoted as CEI_int), representing the total input of acoustic energy throughout the cleaning process, its normalization formula is = (CEI int - CEI min int) / (CEI max int - CEI min int). CEI min int and CEI max int represent the integral value range corresponding to invalid cleaning and optimal cleaning process respectively. For fluid degradation index FDI, which is itself partially normalized, it can be used directly or scaled again to obtain . For example, if the valid range of FDI is [0, 1.2], then = FDI / 1.2.

[0143] Step S422, the final comprehensive cleaning quality score (CQS) is calculated by a multi-dimensional weighted summation model. This model integrates all information of result evaluation (surface cleanliness) and process monitoring (physical action intensity, chemical medium quality). Its mathematical expression is:

[0144]

[0145] In this expression, the term (1- ) converts the contamination index into a cleanliness score, i.e. the lower the RCI, the higher the cleanliness score. directly represents the quality of the physical cleaning process. The term (1- ) converts the degradation degree of the cleaning fluid into its effectiveness score, i.e. the higher the FDI, the lower the effectiveness score. The weight coefficients a, b, g are preset parameters that can be adjusted through the system configuration interface, and their sum is 1 (a + b + g = 1). These weights reflect the relative importance of different dimensions in the final quality assessment.

[0146] For example, in a cleaning scenario for high-risk invasive ultrasound probes (such as transesophageal echocardiography probes), the final surface cleanliness is the primary concern, so the weights can be set as a = 0.7, b = 0.2, g = 0.1. Assuming that after normalization, we get = 0.08, = 0.85, = 0.20. Then the final CQS is calculated as follows:

[0147] CQS = 0.7 * (1 - 0.08) + 0.2 * 0.85 + 0.1 * (1 - 0.20)

[0148] CQS = 0.7 * 0.92 + 0.2 * 0.85 + 0.1 * 0.80

[0149] CQS = 0.644 + 0.170 + 0.080 = 0.894

[0150] This score (89.4%) will be used for the next step of qualification.

[0151] Step S5, the calculated comprehensive quality score (CQS) is compared with a pre-set qualification threshold (QualificationThreshold). The threshold is set according to the relevant medical device cleaning and disinfection guidelines, manufacturer's instructions, and the requirements of the hospital infection control department, for example, it can be set to 0.90.

[0152] If the CQS is higher than or equal to the qualification threshold (for example, the calculated CQS is 0.91≥0.90), the system determines that this cleaning is qualified. Then, the system automatically generates a digital cleaning certificate containing comprehensive traceability information. The certificate is a structured data file in JSON or XML format, which contains at least: the unique identity code of the probe (US-ID), the precise time stamp of the start and end of cleaning (consistent with ISO8601 standard), the summary of the pre-cleaning and post-cleaning multi-spectral image sets (for example, thumbnails of key areas or SHA-256 hash values of the entire data set), the complete time series data of the acoustic cavity cavitation energy index (CEI), the history record of the fluid state vector (V_t), the sub-item scores constituting the CQS (, the final CQS total score, and the system ID and operator's digital signature (if any) performing this cleaning.

[0153] To ensure the non-tamperability and credibility of the certificate, the system encrypts the generated certificate data and binds its hash value with the US-ID of the probe as a new entry written into a private blockchain or distributed ledger technology (DLT) based database. Each cleaning event constitutes a block, which contains the certificate hash and is linked to the previous block (the last cleaning record of the probe), forming an unalterable and auditable traceability chain.

[0154] ​​​If the CQS is below the pass threshold (e.g., the calculated CQS is 0.894 < 0.90), the system determines that this cleaning attempt is a failure. At this point, the system will automatically trigger a warning and display the reason analysis on the operation interface, e.g., pointing out whether it is due to “excessive surface residual contaminants (low RCI score)” or “insufficient ultrasonic cavitation process effectiveness (low CEI score).” Based on this analysis, the system automatically enters the re-cleaning (Re-cleaning) mode. In the re-cleaning mode, the system can intelligently adjust the cleaning protocol, e.g., if the failure reason is due to excessive RCI, the system can extend the time of the high-pressure spraying stage or increase the concentration of the cleaning agent; if the failure reason is due to insufficient CEI, the system can use stronger ultrasonic power or a wider sweep range. After the re-cleaning is completed, the system will repeat steps S3 to S5 to perform a new round of evaluation. All failed cleaning attempts and related monitoring data will also be recorded in detail in the log for long-term quality control and process optimization analysis.

[0155] Referring to Figure 2 The present application also provides an intelligent cleaning and monitoring system for ultrasonic imaging probes. The system is the physical carrier for implementing the above method, which includes the following functional modules:

[0156] Probe identity and state recognition module. This module physically integrates a 13.56 MHz RFID / NFC reader and a high-resolution industrial camera for two-dimensional code reading, installed at the entrance of the cleaning chamber. The embedded controller of this module is responsible for executing the logic of step S111, achieving automatic acquisition of probe identity and instant calling of its historical data, physical model parameters (such as geometry_data and

[0157] Multi-spectral imaging analysis module, which is the “eyes” of the system. The core of this module is a precision two-dimensional rotary translation mechanism, a high-resolution CMOS image sensor, and a multi-wavelength LED ring light source system. The module has a high-performance embedded computing unit built-in, such as a System on Chip (SoC) with GPU. This computing unit is responsible for all tasks related to image processing, including controlling scanning motion (step S112), image acquisition sequence (step S113), CNN-based contaminant segmentation (step S113), SIFT-based image registration (step S311), differential image calculation, and finally the calculation of the surface residual contaminant index (RCI) (steps S411 and S412).

[0158] ​Cleaning operation and process monitoring module, which is the "actuator" and "nervous system" of the system. This module includes a sealed cleaning cavity, a water pump for high-pressure spraying and a precision nozzle array, an ultrasonic generator and multiple piezoelectric transducers, and a heating assembly for high-temperature disinfection. Further, this module integrates process monitoring sensors, i.e. broadband hydrophone and microfluidic chemical sensing chip (or discrete sensor group). A dedicated FPGA board is connected to the hydrophone, responsible for real-time FFT and calculation of cavitation energy index (CEI) of the acoustic cavity (steps S211 and S212). A microcontroller (MCU) is responsible for polling the readings of the chemical sensor, generating a real-time fluid state vector (V_t) and calculating the fluid degradation index (FDI) (step S213). This module also performs closed-loop feedback control, such as adjusting the power and frequency of the ultrasonic generator according to the CEI value.

[0159] Data fusion and decision authentication module, which is the "brain" of the system. This module is usually an industrial-grade computer running a real-time operating system (RTOS). It acts as a central coordinator, receiving RCI values from the multi-spectral imaging analysis module and final values of CEI and FDI from the cleaning operation and process monitoring module. The software core of this module is the data fusion and decision algorithm, responsible for performing the normalization process in step S421, the CQS weighted calculation in step S422, and the eligibility determination in step S5. According to the determination result, it issues instructions to other modules, such as the command to generate a certificate or the command to start a re-cleaning cycle.

[0160] Data recording and traceability interface module, which is the "archive" and "diplomat" of the system. This module is responsible for implementing the digital certificate generation, encryption and secure storage described in step S5. It runs a lightweight blockchain node software internally, responsible for packaging the hash value of the cleaning certificate into a transaction and submitting it to the hospital's private distributed ledger network. In addition, this module provides standardized data exchange interfaces, such as an interface compliant with the HL7 (Health Level Seven) standard, which can push the cleaning eligible event message (containing probe ID, timestamp and certificate hash) to the hospital information system (HIS), or an interface compliant with the DICOM (Digital Imaging and Communications in Medicine) standard, which can associate the cleaning certificate with the probe as a non-image object (Non-image Object) and store it in the medical image archive and communication system (PACS). This allows clinicians to view the latest and trusted cleaning and disinfection records of the probe used when reviewing patient images, forming a complete closed-loop traceability.

[0161] The above merely describes the preferred embodiments of the present application, and is not intended to limit the present application in other forms. Any skilled person in the art can make changes or modifications to the above disclosed technical content into equivalent embodiments with equivalent changes, and apply to other fields, but as long as it does not deviate from the technical solution of the present application, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application still belongs to the protection scope of the technical solution of the present application.

Claims

1. An intelligent cleaning and monitoring method for an ultrasound imaging probe, characterized in that, The method comprises the following steps: S1: Obtain the unique identity code of the ultrasonic imaging probe to be cleaned, call the historical cleaning data and physical model parameters, drive the multi-spectral imaging unit to perform pre-cleaning surface scanning on the probe, collect the baseline image data set, extract the initial contamination distribution characteristics of the probe surface, and generate the baseline contamination atlas; S2: Based on the baseline contamination atlas, activate the cleaning operation execution unit to execute multi-stage cleaning operation, collect and analyze the acoustic field signals of the liquid medium through the acoustic sensing unit during the multi-stage cleaning operation, calculate the cavitation energy index of the acoustic cavity, and monitor the key chemical parameters of the cleaning liquid through the chemical sensing unit to generate the cleaning process state data set; S3: After the cleaning operation is completed, drive the multi-spectral imaging unit to perform post-cleaning surface scanning on the probe, collect the finished image data set consistent with the baseline image data set parameters, perform pixel-level registration and calculate the difference, and generate the residual contamination difference image; S4: Based on the residual contamination difference image, perform image segmentation and morphological analysis, quantify the residual contamination characteristics to calculate the surface residual contamination index, and fuse the surface residual contamination index and the cleaning process state data set to obtain the comprehensive cleaning quality evaluation score through multi-dimensional weighted algorithm.

2. The intelligent cleaning monitoring method for an ultrasonic imaging probe according to claim 1, characterized in that, The baseline contamination atlas includes initial contamination coordinate matrix, contamination spectral response curve, and initial contamination load index, the cleaning process state data set includes the cavitation energy index of the acoustic cavity and the real-time fluid state vector, and the surface residual contamination index includes contamination coverage, average residual thickness estimate, and contamination morphology dispersion.

3. The method of claim 2, wherein, The generation step of the baseline contamination atlas is specifically: S111: Read the electronic tag of the probe through the radio frequency identification reader or near field communication module, obtain its unique identity code, and use it as an index to retrieve the model, geometric size data and historical records of the probe in the device historical database; S112: Drive the two-dimensional rotary translation mechanism controlled by the stepper motor to accurately control the motion of the probe in the field of view of the multi-spectral imaging unit according to the preset trajectory, and the multi-spectral imaging unit includes a high-resolution image sensor and a ring-shaped light source system composed of multi-wavelength light emitting diodes; S113: Turn on the light sources of different wavelengths in turn and collect high-resolution images under each wavelength to form the baseline image data set, apply the semantic segmentation model based on convolutional neural network to process the data set, identify and segment the contamination area, encode its coordinates, area and spectral intensity, and generate the structured baseline contamination atlas.

4. The method of claim 3, wherein, The generation step of the cleaning process state data set is specifically: S211: Collect the acoustic field pressure time domain signal of the ultrasonic immersion cleaning stage at a preset sampling rate through the broadband hydrophone installed in the cleaning cavity, and send it to the signal processing unit realized by the field programmable gate array; S212: The signal processing unit performs a short-time Fourier transform on the time-domain signal to generate a sound field spectrum, calculates the ratio of the energy in the broadband noise evaluation band to the total energy of the full band to obtain the cavitation energy index of the acoustic cavity, and collects conductivity, pH value, temperature, and light absorption readings through a microfluidic sensing chip integrated in the cleaning fluid circulation loop to form the real-time fluid state vector.

5. The method of claim 4, wherein, The generating step of the residual contaminant differential image is specifically: S311: The same scanning path and imaging parameters as the pre-cleaning scan are used to collect the finished image dataset, and a scale-invariant feature transform-based image registration algorithm is applied to calculate the geometric transformation matrix between the reference image dataset and the finished image dataset, achieving sub-pixel level alignment; S312: For each spectral band, the gray value of each pixel of the registered finished image and reference image is subtracted and the absolute value is taken, an adaptive threshold segmentation algorithm is applied to extract regions above the background noise level, and the residual contaminant differential image is generated.

6. The method of claim 5, wherein, The calculating step of the surface residual contaminant index is specifically: S411: For all residual contaminant patches identified in the residual contaminant differential image, the total area and the area, perimeter, and average gray value of each patch are calculated; S412: According to the contaminant coverage, the average residual thickness estimated from the average gray value, and the morphological dispersion calculated from the area-to-perimeter ratio, the surface residual contaminant index is obtained through a weighted summation formula.

7. The method of claim 6, wherein the method further comprises: The calculating step of the comprehensive cleaning quality evaluation score is specifically: S421: The surface residual contaminant index, the time series integral value of the cavitation energy index of the acoustic cavity, and the norm of the deviation of the real-time fluid state vector from the initial state are normalized to map their numerical ranges to a unified interval; S422: The normalized surface residual contaminant index, cavitation energy index integral value, and fluid state vector deviation norm are weighted and fused through a weighted summation model to calculate the comprehensive cleaning quality evaluation score.

8. The intelligent cleaning monitoring method for an ultrasonic imaging probe according to claim 1, characterized in that, The method further includes the S5 step: S5: Compare the comprehensive cleaning quality evaluation score with the preset qualified threshold value. If the score is higher than the threshold value, the cleaning is determined to be qualified, a digital cleaning certificate is generated and bound to the probe unique identity code for storage; If the score is lower than the threshold value, the cleaning is determined to be unqualified, triggering a new round of cleaning operation; The digital cleaning certificate includes the probe unique identity code, cleaning start and end time stamps, pre-cleaning and post-cleaning image summaries, cavitation energy index time series, fluid state vector history, final cleaning quality evaluation score, and operator digital signature.

9. The intelligent cleaning monitoring method for an ultrasonic imaging probe according to claim 8, characterized in that, The S5 step is specifically: S511: Call the comprehensive cleaning quality evaluation score and compare its numerical value with the qualified score threshold value stored in the system configuration; S512: When the evaluation score is greater than or equal to the qualified threshold, it is determined that the cleaning is qualified, the digital cleaning certificate containing the whole-process monitoring data is generated, the certificate is encrypted and signed, and then associated with the unique identity code and written into the tamper-proof distributed database; S513: When the evaluation score is less than the qualified threshold, it is determined that the cleaning is unqualified, and an instruction is sent to the cleaning operation execution unit to automatically start a new round of cleaning operation process.

10. An intelligent cleaning monitoring system for an ultrasound imaging probe, characterized in that, The system is used to implement the intelligent cleaning monitoring method of the ultrasonic image probe according to any one of claims 1-9, and the system comprises: An initial state evaluation module is configured to acquire a unique identity code of an ultrasonic image probe to be cleaned, drive a multi-spectral imaging unit to perform pre-cleaning scanning, collect and analyze a reference image data set, and generate a reference contaminant map; A cleaning process monitoring module is configured to perform multi-stage physical and chemical cleaning operations, and collect and analyze sound field signals and cleaning liquid parameters online through acoustic and chemical sensing units to generate a cleaning process state data set; A cleaning effect imaging module is configured to drive the multi-spectral imaging unit to perform post-cleaning scanning after the cleaning operation is completed, collect a finished image data set, and perform registration and difference calculation with the reference image data set to generate a residual contaminant difference image; A cleaning quality evaluation module is configured to analyze the residual contaminant difference image to calculate a surface residual contaminant index, and fuse the index with the cleaning process state data set to calculate and generate a comprehensive cleaning quality evaluation score; A decision authentication and evidence module is configured to compare the comprehensive cleaning quality evaluation score with a qualified threshold, and determine the qualification to generate and evidence a digital cleaning certificate according to the comparison result, or determine the unqualification to trigger a new round of cleaning operation.

Images