AI-driven medical laboratory internal examination defect prediction and closed-loop management system

Abstract

The invention discloses an AI-driven medical laboratory internal examination defect prediction and closed-loop management system, and particularly relates to the technical field of defect prediction and closed-loop management. The method comprises the following steps: continuously acquiring operation fine granularity characteristics of large automatic detection equipment in an optical granularity signal drift coefficient, a single-channel metering deviation degree and an electrode response time abnormal coefficient through a data collection module, and synchronously acquiring batch-level difference characteristics of a reagent batch storage temperature fluctuation coefficient and a historical use exposure accumulation coefficient; a pressure critical risk assessment model fusing machine-material double-side recessive degradation characteristics is constructed, a pressure critical risk state label that alarm is not triggered but a systematic deviation trend exists is output, risk identification and specific intervention actions are tightly coupled through a closed-loop rectification module, and potential deviation sources are synchronously reduced from an equipment side and a reagent side.

Description

Technical Field

[0001] This invention relates to the field of defect prediction and closed-loop management technology, and more specifically, to an AI-driven medical laboratory internal audit defect prediction and closed-loop management system. Background Technology

[0002] In actual medical laboratory operations, medium and large-scale automated testing equipment is not always in a binary state of "fault / normal," but rather oscillates within a latent degradation range that falls within the manufacturer's threshold. For example, minor optical alignment drift in laser particle size analyzers, gradual aging of colorimetric channel light sources before their brightness falls below the alarm threshold, minute leaks in individual channels of multi-channel pipette arms, and sluggish response of ISE electrode films before complete failure—these conditions are insufficient to trigger equipment self-tests or alarms, and they can introduce low-intensity, slowly changing systematic biases into the test results over a longer timescale.

[0003] The problem is that such deviations are not isolated equipment issues, but rather have a cumulative and amplified effect with reagent systems, quality control strategies, and internal audit mechanisms. On the one hand, batch differences in reagents and fluctuations in storage and transportation conditions can create a complex coupling with the aforementioned hidden hardware degradation, resulting in "same-direction deviations" or "mutual compensations," making routine internal quality control appear "under control" in the short term, thus masking the continuous deterioration of equipment performance. On the other hand, traditional internal audits and defect identification usually use explicit events such as "alarm records," "quality control exceeding limits," and "significant deviations in EQA" as entry points, making it difficult to capture this subcritical risk state of "deviation without alarms" in a timely manner. This leads to internal audit findings being mostly retrospective rather than pre-emptive.

[0004] Therefore, the latent faults of medium and large-sized instruments and the random fluctuations of reagent / quality control systems constitute a composite source of distortion at the physical level that is mechanistically invisible and statistically dilutionable. This directly weakens the ability of traditional quality management and document review mechanisms to perceive real risks. This intertwining of latent degradation and reagent differences is a key technical challenge that AI-driven medical laboratory internal audit defect prediction and closed-loop management systems must address head-on: only through long-term modeling of multi-source operational data, quality control trajectories, and batch information can systemic quality hazards hidden beneath the "qualified appearance" be identified in advance, before equipment alarms are triggered or results exceed limits, thus achieving a shift from passive response to proactive early warning and closed-loop control. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an AI-driven medical laboratory internal audit defect prediction and closed-loop management system to address the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: AI-driven medical laboratory internal audit defect prediction and closed-loop management system, including data collection module, critical risk assessment module, and closed-loop rectification module; The data collection module is used to acquire hidden degradation information of medium and large-scale automated testing equipment through the hidden degradation unit and to acquire reagent batch difference information through the batch difference unit. The critical pressure risk assessment module is used to construct a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and output critical pressure risk status labels that have not triggered alarms but have hidden systematic deviations. The closed-loop rectification module is used to implement a closed-loop rectification strategy based on the critical risk status label. The closed-loop rectification strategy includes adjusting the calibration frequency through the calibration unit and adjusting the available reagent batches through the reagent batch adjustment unit.

[0007] In a preferred embodiment, the occult degradation information of the medium and large-scale automated detection equipment includes the optical particle size signal drift coefficient of the laser particle size analyzer, the single-channel metrological deviation, and the electrode response time anomaly coefficient. The reagent batch variation information includes the reagent batch storage temperature fluctuation coefficient and the reagent batch historical exposure cumulative coefficient.

[0008] In a preferred embodiment, the logic for obtaining the optical graininess signal drift coefficient is as follows: Obtain the particle size distribution of the laser particle size analyzer output relative to a stable reference sample at time t: ,in Let j be the center particle size of the j-th particle size range. The number of particle size ranges. The number fraction within the particle size range measured at time t; Calculate the average particle size at time t: ,in The average particle size; Set reference time The baseline average particle size was calculated based on the baseline time. ,in Based on the average particle size, As the baseline distribution; At a later time point Calculate the average particle size at each time step. Based on the average particle size at each time point Calculate the average particle size drift rate over time: ,in The average particle size drift rate, ,in Here, N is the index number of the time point, and N is the total number of observed time points; time point Particle size distribution Compared with the baseline distribution Comparison and calculation of distribution shape drift rate: ,in The shape shift rate is the distribution. Assign importance weights to different particle size ranges; Calculate the optical particle size signal drift coefficient based on the average particle size drift rate and the distribution shape drift rate: ,in The optical particle size signal drift coefficient. , These are preset proportionality coefficients for the average particle size drift rate and the distribution shape drift rate, respectively. , All are greater than 0.

[0009] In a preferred embodiment, the logic for obtaining the single-channel metering deviation is as follows: Obtain multichannel pipette arm in time Measured volume of the same target: ,in Channel number , Total number of channels; In the time window Inside, access channel of Valid measurement results Calculate the time window inner passage Average measured volume: ,in To represent the average measured volume, Let i be the end time of the i-th time window. For channel Obtain the sequence number of the measured volume. The length of the time window; The relative measurement deviation of the channel is calculated by comparing the average measured volume with the nominal target volume. ,in The nominal target volume; In the time window First, calculate the average deviation of all channels: ,in The average deviation across all channels; Calculate the dispersion of the channel group: ,in The standard deviation of the channel group; The single-channel measurement deviation is obtained by dividing the difference between the relative measurement deviation of a single channel and the population average deviation by the standard deviation of the population deviation. ,in For single-channel measurement deviation, for This is a very small constant used to prevent division by zero.

[0010] In a preferred embodiment, the logic for obtaining the electrode response time anomaly coefficient is as follows: Acquire the time-varying electrode signal output by the electrode during actual measurement. ; The initial value of the electrode signal is marked as Stable values ​​are marked as The time required for the electrode signal to reach 95% of its stable value is marked as the electrode response time. ; In the time window Electrode response time for multiple measurements Calculate the single response time deviation: ,in The single response time deviation is calculated for the g-th measurement; Calculate the electrode response time anomaly coefficient: ,in This is the electrode response time anomaly coefficient.

[0011] In a preferred embodiment, the logic for obtaining the reagent batch storage temperature fluctuation coefficient is as follows: Temperature sensors are placed within the reagent storage environment, and sampling intervals are set. Record temperature sequences: ,in For the r-th time point The temperature sensor collects the reagent storage temperature, where R is the number of sampling time points; The collected reagents were stored at the same temperature as the recommended storage temperature. Compare and calculate the instantaneous deviation of reagent storage temperature: ,in This is an instantaneous deviation; Calculate the maximum and minimum fluctuations in reagent storage temperature: ,in This represents the maximum to minimum fluctuation in reagent storage temperature. Calculate the cumulative deviation: ,in This is the cumulative deviation. By combining the maximum-minimum temperature fluctuation and cumulative deviation of reagent storage, the batch temperature fluctuation coefficient of reagent is obtained: ,in This refers to the temperature fluctuation coefficient during batch storage of the reagent. The maximum and minimum temperature fluctuations during the storage of reagent batch h are given. This represents the cumulative deviation of batch h of reagents. This represents the allowable fluctuation range for batch h of reagents. The total monitoring time for batch h of reagents. , This represents the number of sampling times for the h-th batch of reagents.

[0012] In a preferred embodiment, the logic for obtaining the historical exposure cumulative coefficient of reagent batches is as follows: Obtain and number the exposure events that occur during the lifecycle of the target batch of reagents: Where e is the exposure event number, The total number of exposure events; each exposure event includes the following identifier: exposure start time. Exposure end time Exposure duration Average ambient temperature during exposure Recommended storage temperature ; Based on the average ambient temperature during the exposure period Recommended storage temperature Calculate effective overtemperature : ; Based on effective overtemperature Calculate the temperature-time combined exposure for the e-th exposure event: ,in This represents the combined temperature-time exposure. The permissible over-temperature buffer range for the reagent. For reference time scale; Combined with the number of bottles opened during each exposure event Number of cross-environment / cross-device transfers Number of times exposed to strong light and ultraviolet rays Calculate the total exposure amount for the e-th exposure event: ,in Let e ​​be the total exposure amount of the e-th exposure event. These represent preset proportionality coefficients for temperature-time combined exposure, number of bottle openings, number of cross-environment / cross-equipment transfers, and number of exposures to strong light and ultraviolet radiation, respectively. All are greater than 0; For the same reagent batch, the exposure amounts of all exposure events within its lifecycle are summed to obtain the historical cumulative exposure amount for the entire lifecycle: ,in This represents the cumulative historical exposure amount for batch h of reagents. Calculate the historical exposure cumulative factor for reagent batches based on historical usage cumulative exposure: ,in This represents the cumulative exposure factor based on historical usage of the reagent batch. The threshold for safe cumulative exposure.

[0013] In a preferred embodiment, the step of constructing a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and outputting critical pressure risk status labels that have not triggered alarms but contain implicit systematic deviations, includes the following steps: The hidden degradation information of medium and large-scale automated testing equipment and the batch-to-batch difference information of reagents are used as the input feature vector of the critical pressure risk assessment model: ; Construct a quantization function for the criticality risk assessment model based on the input feature vector: ,in This is the critical risk assessment value. Let d be the index in the input feature vector. Let be the preset proportional coefficient for the d-th indicator, and All are greater than 0; The critical risk assessment value will be compressed. With pressure critical risk threshold Compare and output the critical risk state label. : Among them, the critical risk state label This indicates that although the instrument did not trigger an alarm and the quality control results were still within the acceptable range, the system was approaching a critical state of potential deviation, which could trigger closed-loop intervention.

[0014] In a preferred embodiment, when the critical risk state label is pressed... At that time, a closed-loop rectification strategy was implemented; The adjustment of the calibration frequency via the calibration unit includes: calculating the calibration frequency adjustment coefficient based on the critical pressure risk assessment value. ,in To calibrate the frequency adjustment factor, This is the current calibration frequency for medium and large-scale automated testing equipment. To adjust the sensitivity coefficient; when and When the calibration frequency is increased, the system automatically generates a new calibration schedule and sends it to the device management terminal; The adjustment of available reagent batches through the reagent batch adjustment unit includes: the reagent batch adjustment unit retrieves all in-stock reagent batch difference information corresponding to the target detection item from the data collection module, including the reagent batch storage temperature fluctuation coefficient and the reagent batch historical use exposure cumulative coefficient for each batch; Calculate the batch sort value for each available and pending batch: ,in This is the sorting value for batch numbered b among the available and pending batches. Preset proportional coefficients for the reagent batch storage temperature fluctuation coefficient and the reagent batch historical usage exposure cumulative coefficient, and All are greater than 0; Sort all batches by batch sorting value from lowest to highest, and divide them into: Normal risk zone: Marked as preferred batch; Critical risk zone: Marked as a restricted batch; High-risk areas: Batch marked as disabled; The reagent batch adjustment unit automatically adjusts the list of available reagent batches accordingly.

[0015] The technical effects and advantages of this invention are as follows: 1. This invention achieves a fundamental breakthrough over the limitations of traditional quality management and internal audit systems that "only look at alarms and out-of-bounds violations" by jointly modeling hidden degradation information of equipment and batch-specific differences of reagents within the same framework. Through a data collection module, it continuously acquires fine-grained operational characteristics of medium-to-large-scale automated testing equipment, such as optical particle size signal drift coefficients, single-channel metrological deviations, and electrode response time anomalies. Simultaneously, it collects batch-level difference characteristics, such as reagent batch storage temperature fluctuation coefficients and historical usage exposure cumulative coefficients. A critical pressure risk assessment module constructs a critical pressure risk assessment model that integrates hidden degradation characteristics from both the equipment and material sides, outputting a critical pressure risk status label indicating "no alarm triggered but a systematic deviation trend already exists." Furthermore, this invention integrates risk identification and specific intervention actions through a closed-loop rectification module. With tight coupling, when a critical risk status tag is triggered, the system can automatically drive the calibration unit to dynamically increase the calibration frequency of relevant instruments and channels. At the same time, the reagent batch adjustment unit performs risk classification and availability rearrangement of in-use batches, simultaneously reducing potential sources of deviation from both the equipment and reagent sides. Thus, in terms of technical effect, this invention achieves a shift from "discovering explicit faults after the fact and passively correcting deviations" to "identifying implicit degradations in advance and actively managing them in a closed loop." This not only improves the long-term stability and traceability of laboratory test results, but also significantly enhances the internal audit system's ability to perceive subcritical risk statuses, avoiding systemic quality hazards that are masked by quality control appearances and alarm logic. It provides strong support for medical laboratories to achieve intelligent, refined, and forward-looking quality management without interrupting daily operations. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of the system according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example: The present invention provides, as follows Figure 1 The AI-driven medical laboratory internal audit defect prediction and closed-loop management system shown includes a data collection module, a critical risk assessment module, and a closed-loop rectification module. The data collection module is used to acquire hidden degradation information of medium and large-scale automated testing equipment through the hidden degradation unit and to acquire reagent batch difference information through the batch difference unit. The critical pressure risk assessment module is used to construct a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and output critical pressure risk status labels that have not triggered alarms but have hidden systematic deviations. The closed-loop rectification module is used to implement a closed-loop rectification strategy based on the critical risk status label. The closed-loop rectification strategy includes adjusting the calibration frequency through the calibration unit and adjusting the available reagent batches through the reagent batch adjustment unit. The data collection module is used to acquire hidden degradation information of medium and large-scale automated testing equipment through the hidden degradation unit and to acquire reagent batch difference information through the batch difference unit. The hidden degradation information of the medium and large-scale automated testing equipment includes the optical particle size signal drift coefficient of the laser particle size analyzer, the single-channel metrological deviation, and the electrode response time anomaly coefficient. The reagent batch variation information includes the reagent batch storage temperature fluctuation coefficient and the reagent batch historical usage exposure cumulative coefficient. In this embodiment of the invention, to characterize the slow degradation of the optical detection link of a laser particle size analyzer under conditions where the manufacturer's alarm threshold has not been triggered, an optical particle size signal drift coefficient is introduced as a key characterization parameter for the implicit degradation information of medium and large-scale automated detection equipment. The reason for this is that during long-term operation, the laser particle size analyzer is affected by physical factors such as light source brightness decay, optical path contamination, and changes in detector sensitivity. Statistically, the output particle size distribution signal will show a trend of gradual shift in the distribution center position and slight distortion in the distribution shape. This slow, continuous, small-amplitude drift usually does not cause quality control violations or equipment self-test errors in the short term, but it can be superimposed with reagent batch differences to form a critical risk state of "results appearing to be under control, but the system actually deviating." The optical particle size signal drift coefficient is a comprehensive index used to quantify the degree of slow shift of the particle size detection signal of a laser particle size analyzer relative to a reference state during long-term operation. It is a numerical measure characterizing the long-term stability of the optical detection link of the laser particle size analyzer, reflecting the systematic drift intensity of particle size distribution characteristics (such as mean particle size, median particle size, or characteristic quantiles) relative to the initial reference state. The logic for obtaining the optical grain size signal drift coefficient is as follows: Obtain the particle size distribution of the laser particle size analyzer output relative to a stable reference sample at time t: ,in Let j be the center particle size of the j-th particle size range. The number of particle size ranges. The number fraction within the particle size range measured at time t; Calculate the average particle size at time t: ,in The average particle size; Set reference time (For example, the initial calibration time after equipment installation, or the reference period recommended by the manufacturer), and the reference average particle size calculated based on the reference time: ,in Based on the average particle size, As the baseline distribution; At a later time point ( ,in (where N is the index number of the time point and N is the total number of observed time points), and calculate the average particle size at each time point. Based on the average particle size at each time point Calculate the average particle size drift rate over time: ,in The average particle size drift rate, ,in Here, N is the index number of the time point, and N is the total number of observed time points; Looking at the average particle size alone is not enough, because optical degradation (such as light source attenuation and changes in detector response) can also change the distribution shape; time point Particle size distribution Compared with the baseline distribution Comparison and calculation of distribution shape drift rate: ,in The shape shift rate is the distribution. Assign importance weights to different particle size ranges (for example, the initial value can be set to 1, and can be adjusted later according to the project's sensitivity to specific particle size ranges). Optical degradation manifests both in the shift of the average particle size (energy center) and in the deformation of the distribution shape (response curve). Therefore, the optical particle size signal drift coefficient is calculated based on the average particle size drift rate and the distribution shape drift rate. ,in The optical particle size signal drift coefficient. , These are preset proportionality coefficients for the average particle size drift rate and the distribution shape drift rate, respectively. , All are greater than 0; It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization, Z-Score standardization, etc., which will not be elaborated here. , The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. , The initial value can be 0.5, 0.5; In this embodiment of the invention, to characterize the subtle, gradual metering inconsistencies that arise during long-term operation of multi-channel pipetting or metering units in medium-to-large-scale automated testing equipment, single-channel metering deviation is introduced as another key parameter characterizing the equipment's latent degradation information. The reason for this is that during long-term, high-frequency operation, multi-channel pipetting arms or multi-well injection modules are affected by physical factors such as mechanical wear, aging of seals, localized blockages, and changes in lubrication conditions, resulting in small but cumulative deviations in the actual volume of liquid aspirated or displaced in each channel. These deviations often occur only in one or a few channels, and their absolute value is insufficient to immediately trigger quality control breaches or equipment alarms. However, when superimposed with the sensitivity patterns of specific reagent batches or specific projects, they manifest as "systematically higher or lower results in a certain testing channel," forming a typical latent criticality risk source. Without quantitative monitoring, this "few channels falling behind" situation is difficult to identify in a timely manner through traditional internal audits and routine quality control. Single-channel measurement deviation refers to the degree of deviation of the measured volume of a specific channel from the reference measurement level and the overall level of other channels in a multi-channel pipetting or metering unit during actual long-term operation.

[0019] The logic for obtaining the single-channel measurement deviation is as follows: Obtain multichannel pipette arm in time Measured volume of the same target: ,in Channel number , Total number of channels; It should be noted that the number of channels in a multichannel pipette is determined by the actual pipette being evaluated, such as 8-channel or 16-channel pipettes. Data sources can include: online calibration, measurement records during routine quality control, volume estimates calculated from built-in flow / pressure sensors, and measured volumes obtained using weighing / conductivity methods. In the time window Inside, access channel of Valid measurement results Calculate the time window inner passage Average measured volume: ,in To represent the average measured volume, Let i be the end time of the i-th time window. For channel Obtain the sequence number of the measured volume. The length of the time window; The relative measurement deviation of the channel is calculated by comparing the average measured volume with the nominal target volume. ,in The nominal target volume; In the time window First, calculate the average deviation of all channels: ,in The average deviation across all channels; Calculate the dispersion of the channel group: ,in The standard deviation of the channel group; The single-channel measurement deviation is obtained by dividing the difference between the relative measurement deviation of a single channel and the population average deviation by the standard deviation of the population deviation. ,in For single-channel measurement deviation, for This is to prevent division by zero by a very small constant (generally taken as...). ); It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization, Z-Score standardization, etc., which will not be elaborated here. In this embodiment of the invention, the detection performance of ISE (ion-selective electrode) commonly used in medium and large-scale automated detection equipment is affected by factors such as membrane aging, deposit adhesion, changes in electrolyte concentration, temperature fluctuations, and changes in electrode interface contact impedance during long-term use. These physical and chemical mechanisms can cause the electrode to exhibit delayed or nonlinear responses when responding to the sample. Even if the final measurement result is still within the instrument alarm threshold, there may be "hidden degradation," affecting system stability and result reliability. This type of delay usually manifests as an increased time for the electrode to reach a stable signal or an increase in transient fluctuations, which is a typical hidden risk source. The electrode response time anomaly coefficient refers to the degree of deviation of the actual signal response time of an ion-selective electrode (ISE) or similar electrode from the healthy baseline response time and its historical response trend during sample measurement in automated detection equipment. The logic for obtaining the electrode response time anomaly coefficient is as follows: Acquire the time-varying electrode signal output by the electrode during actual measurement. ; The initial value of the electrode signal is marked as Stable values ​​are marked as The time required for the electrode signal to reach 95% of its stable value is marked as the electrode response time. ; In the time window Electrode response time for multiple measurements Calculate the single response time deviation: ,in The single response time deviation is calculated for the g-th measurement; The maximum deviation is used to measure the most severe latent degradation in the calculation of the electrode response time anomaly coefficient. ,in This is the electrode response time anomaly coefficient; It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization, Z-Score standardization, etc., which will not be elaborated here. In this embodiment of the invention, the quality and stability of reagent batches are affected not only by the manufacturing process but also by fluctuations in storage conditions. Temperature fluctuations, in particular, can accelerate the degradation of certain chemical or biological reagents, leading to decreased activity, reduced reaction efficiency, and even slightly affecting measurement results. Even if the reagent has not reached its expiration date or alarm threshold, the cumulative effect of temperature fluctuations may still cause hidden degradation risks, manifesting as systematic deviations in specific projects or sensitive channels. The reagent batch storage temperature fluctuation coefficient refers to the degree of deviation of the temperature fluctuation amplitude of a reagent batch during storage from its allowable fluctuation range in a medical laboratory. The logic for obtaining the reagent batch storage temperature fluctuation coefficient is as follows: Temperature sensors are placed within the reagent storage environment, and sampling intervals are set. Record temperature sequences: ,in For the r-th time point The temperature sensor collects the reagent storage temperature, where R is the number of sampling time points; The collected reagents were stored at the same temperature as the recommended storage temperature. Compare and calculate the instantaneous deviation of reagent storage temperature: ,in This is an instantaneous deviation; To capture extreme fluctuations during batch storage and reflect the risk of momentary loss of control of storage equipment, the maximum-minimum fluctuation of reagent storage temperature was calculated: ,in This represents the maximum to minimum fluctuation in reagent storage temperature. Calculating cumulative deviation measures the implicit risk of degradation from long-term micro deviations: ,in This is the cumulative deviation. By combining the maximum-minimum temperature fluctuation and cumulative deviation of reagent storage, the batch temperature fluctuation coefficient of reagent is obtained: ,in This refers to the temperature fluctuation coefficient during batch storage of the reagent. The maximum and minimum temperature fluctuations during the storage of reagent batch h are given. This represents the cumulative deviation of batch h of reagents. This represents the allowable fluctuation range for batch h of reagents. The total monitoring time for batch h of reagents. , This represents the number of sampling times for the h-th batch of reagents; It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization, Z-Score standardization, etc., which will not be elaborated here. In this embodiment of the invention, a single reagent batch undergoes multiple temperature exposures, bottle openings and closings, workstation stays, and cross-equipment transfers throughout its entire lifecycle, from initial opening to final consumption. Even if the batch as a whole is still within its nominal shelf life, frequent environmental exposures and usage behaviors will cause the reagent activity to gradually decline through mechanisms such as evaporation, oxidation, light exposure, and contamination. The impact on test results is often slow, cumulative, and non-sudden, manifesting as a slight but persistent systematic shift in results within a specific time period or under a specific equipment-channel-project combination. Traditional quality control and internal audits struggle to identify this hidden source of degradation in a timely manner. To transform this "cumulative exposure risk during use" into a quantifiable engineering indicator, this invention introduces a historical usage exposure cumulative coefficient for reagent batches as the core characterization parameter for reagent batch variation information. The basic idea is to map key exposure behaviors of a reagent batch throughout its entire lifecycle, such as the number of bottle openings, cumulative opening time, duration of over-temperature exposure, time spent outside the cold chain, and number of cross-environment transfers, into a unified quantitative indicator to characterize the potential performance degradation degree of the batch of reagents relative to its original state. The logic for obtaining the cumulative exposure coefficient of reagent batches based on historical usage is as follows: Obtain and number the exposure events that occur during the lifecycle of the target batch of reagents: Where e is the exposure event number, The total number of exposure events; each exposure event includes the following identifier: exposure start time. Exposure end time Exposure duration Average ambient temperature during exposure Recommended storage temperature ; It should be noted that the entire lifecycle of a reagent batch h, from opening to use / disposal, includes potentially harmful behaviors such as: repeatedly removing the reagent from a refrigerated environment → placing it on a room temperature workbench → then returning it to the refrigerator; prolonged exposure on the workbench (temperature rise, light, air contact); refrigerator / freezer malfunction leading to overall temperature exceeding the limit for a period of time; frequent opening of bottles and repeated insertion and removal of pipette tips, resulting in localized contamination and volatilization; and multiple transfers between different experimental areas / equipment. All of these behaviors constitute exposure events. Based on the average ambient temperature during the exposure period Recommended storage temperature Calculate effective overtemperature : ; Based on effective overtemperature Calculate the temperature-time combined exposure for the e-th exposure event: ,in This represents the combined temperature-time exposure. This refers to the permissible over-temperature buffer range for the reagent, for example, it can be temporarily stored at 25°C for 2 hours. For reference time scale; It should be noted that the reference time scale This refers to a reference timescale used for dimensionless processing of single exposure duration. It is typically selected based on reagent stability validation results or the manufacturer's instructions, preferably set to 1 hour. However, those skilled in the art can adjust the timescale according to the stability requirements of different reagents. Adjust to other time scales such as 30 minutes and 2 hours; Combined with the number of bottles opened during each exposure event Number of cross-environment / cross-device transfers Number of times exposed to strong light and ultraviolet rays Calculate the total exposure amount for the e-th exposure event: ,in Let e ​​be the total exposure amount of the e-th exposure event. These represent preset proportionality coefficients for temperature-time combined exposure, number of bottle openings, number of cross-environment / cross-equipment transfers, and number of exposures to strong light and ultraviolet radiation, respectively. All are greater than 0; It should be noted that, The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. The initial value can be 0.25, 0.25, 0.25, 0.25; For the same reagent batch, the exposure amounts of all exposure events within its lifecycle are summed to obtain the historical cumulative exposure amount for the entire lifecycle: ,in This represents the cumulative historical exposure amount for batch h of reagents. Calculate the historical exposure cumulative factor for reagent batches based on historical usage cumulative exposure: ,in This represents the cumulative exposure factor based on historical usage of the reagent batch. The threshold for safe cumulative exposure; It should be noted that the safe cumulative exposure threshold This can be determined through validation experiments or historical experience. For example, in stability validation, if it is found that performance begins to deviate significantly near a certain cumulative exposure level, that level can be set as [the appropriate level]. The above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here. The critical pressure risk assessment module is used to construct a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and output critical pressure risk status labels that have not triggered alarms but have hidden systematic deviations. The step of constructing a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and outputting critical pressure risk status labels that have not triggered alarms but contain hidden systematic deviations, includes the following steps: The hidden degradation information of medium and large-scale automated testing equipment and the batch-to-batch difference information of reagents are used as the input feature vector of the critical pressure risk assessment model: ; Construct a quantization function for the criticality risk assessment model based on the input feature vector: ,in This is the critical risk assessment value. Let d be the index in the input feature vector. Let be the preset proportional coefficient for the d-th indicator, and All are greater than 0; It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization, Z-Score standardization, etc., which will not be elaborated here. The settings can be made according to the actual situation. For example, the expert empowerment method can be adopted, which means inviting experts in relevant fields to determine the preset ratio coefficients of each indicator through professional opinion surveys and comprehensive evaluations. For example, the initial value of each indicator can be 0.2. To identify states where latent offsets have already occurred but have not yet triggered alarms, a critical risk threshold is introduced. The critical pressure risk assessment value is compared with the critical pressure risk threshold, and the critical pressure risk status label is output. : Among them, the critical risk state label This indicates that although the instrument did not trigger an alarm and the quality control results were still within the allowable range, the system was close to or had reached a critical state of potential deviation, which could trigger closed-loop intervention. It should be noted that the critical risk status label is... It can be obtained from historical quality control data and statistics on equipment failure / reagent offset events; The closed-loop rectification module is used to implement a closed-loop rectification strategy based on the critical risk status label. The closed-loop rectification strategy includes adjusting the calibration frequency through the calibration unit and adjusting the available reagent batches through the reagent batch adjustment unit. When the critical risk state label is applied At that time, a closed-loop rectification strategy was implemented; The adjustment of the calibration frequency via the calibration unit includes: calculating the calibration frequency adjustment coefficient based on the critical pressure risk assessment value. ,in To calibrate the frequency adjustment factor, This is the current calibration frequency for medium and large-scale automated testing equipment. The sensitivity coefficient can be adjusted according to the device type and historical offset trend; when and When the calibration frequency is increased, the system automatically generates a new calibration schedule and sends it to the equipment management terminal. At the same time, all calibration adjustment operations, frequency changes, risk labels and rectification reasons are automatically generated into an electronic audit trail for internal audit or quality management review. The adjustment of available reagent batches through the reagent batch adjustment unit includes: the reagent batch adjustment unit retrieves all in-stock reagent batch difference information corresponding to the target detection item from the data collection module, including the reagent batch storage temperature fluctuation coefficient and the reagent batch historical use exposure cumulative coefficient for each batch; Calculate the batch sort value for each available and pending batch: ,in This is the sorting value for batch numbered b among the available and pending batches. Preset proportional coefficients for the reagent batch storage temperature fluctuation coefficient and the reagent batch historical usage exposure cumulative coefficient, and All values ​​are greater than 0, used to balance the impact of storage temperature fluctuations and historical exposure accumulation on reagent performance; It should be noted that, The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. The initial value can be 0.5, 0.5; Sort all batches by batch sorting value from lowest to highest, and divide them into: Normal risk zone: Marked as preferred batch; Critical risk zone: Marked as a restricted batch; High-risk areas: Batches marked as disabled or pending evaluation; The reagent batch adjustment unit automatically adjusts the list of available reagent batches accordingly: The preferred batch is placed in the high-priority supply queue and used as the default batch. For batches in the critical risk zone, their use is restricted to non-critical projects or only for short-term use if additional quality control is passed. For batches from high-risk areas, their selectable status in LIS / instrument is automatically frozen, triggering a review or scrapping process by the quality manager; Meanwhile, the reagent batch adjustment unit automatically generates electronic records for each batch sorting value, status change (normal → critical pressure → frozen / scrapped), adjustment reason, and the scope of impact on the testing process (equipment, items, and sample quantity involved), and writes them into the audit trail log.

[0020] This invention achieves a fundamental breakthrough over the limitations of traditional quality management and internal audit systems that "only look at alarms and out-of-bounds violations" by jointly modeling hidden degradation information of equipment and batch-specific differences of reagents within the same framework. Through a data collection module, it continuously acquires fine-grained operational characteristics of medium-to-large-scale automated testing equipment, such as optical particle drift coefficients, single-channel metrological deviations, and electrode response time anomalies. Simultaneously, it collects batch-level difference characteristics, such as reagent batch storage temperature fluctuation coefficients and historical exposure accumulation coefficients. A critical pressure risk assessment module constructs a critical pressure risk assessment model that integrates both machine-material hidden degradation characteristics, outputting a critical pressure risk status label indicating "no alarm triggered but a systematic deviation trend already exists." Furthermore, this invention closely integrates risk identification with specific intervention actions through a closed-loop rectification module. The system is coupled so that when a critical risk status tag is triggered, the system can automatically drive the calibration unit to dynamically increase the calibration frequency of relevant instruments and channels. At the same time, the reagent batch adjustment unit performs risk classification and availability rearrangement of in-use batches, reducing potential sources of deviation from both the equipment and reagent sides. Thus, in terms of technical effect, this invention achieves a shift from "discovering explicit faults after the fact and passively correcting deviations" to "identifying implicit degradations in advance and actively managing them in a closed loop." This not only improves the long-term stability and traceability of laboratory test results, but also significantly enhances the internal audit system's ability to perceive subcritical risk statuses, avoiding systemic quality hazards that are masked by quality control appearances and alarm logic. It provides strong support for medical laboratories to achieve intelligent, refined, and forward-looking quality management without interrupting daily operations.

[0021] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0022] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0023] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0024] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0025] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AI-driven medical laboratory internal audit defect prediction and closed-loop management system, characterized by: It includes a data collection module, a critical pressure risk assessment module, and a closed-loop rectification module; The data collection module is used to acquire hidden degradation information of medium and large-scale automated testing equipment through the hidden degradation unit and to acquire reagent batch difference information through the batch difference unit. The critical pressure risk assessment module is used to construct a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and output critical pressure risk status labels that have not triggered alarms but have hidden systematic deviations. The closed-loop rectification module is used to implement a closed-loop rectification strategy based on the critical risk status label. The closed-loop rectification strategy includes adjusting the calibration frequency through the calibration unit and adjusting the available reagent batches through the reagent batch adjustment unit.

2. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 1, characterized in that: The hidden degradation information of the medium and large-scale automated testing equipment includes the optical particle size signal drift coefficient of the laser particle size analyzer, the single-channel metrological deviation, and the electrode response time anomaly coefficient. The reagent batch variation information includes the reagent batch storage temperature fluctuation coefficient and the reagent batch historical exposure cumulative coefficient.

3. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 2, characterized in that: The logic for obtaining the optical grain size signal drift coefficient is as follows: Obtain the particle size distribution of the laser particle size analyzer output relative to a stable reference sample at time t: ,in Let j be the center particle size of the j-th particle size range. The number of particle size ranges. The number fraction within the particle size range measured at time t; Calculate the average particle size at time t: ,in The average particle size; Set reference time The baseline average particle size was calculated based on the baseline time. ,in Based on the average particle size, As the baseline distribution; At a later time point Calculate the average particle size at each time step. Based on the average particle size at each time point Calculate the average particle size drift rate over time: ,in The average particle size drift rate, ,in Here, N is the index number of the time point, and N is the total number of observed time points; time point Particle size distribution Compared with the baseline distribution Comparison and calculation of distribution shape drift rate: ,in The shape shift rate is the distribution. Assign importance weights to different particle size ranges; Calculate the optical particle size signal drift coefficient based on the average particle size drift rate and the distribution shape drift rate: ,in The optical particle size signal drift coefficient. , These are preset proportionality coefficients for the average particle size drift rate and the distribution shape drift rate, respectively. , All are greater than 0.

4. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 2, characterized in that: The logic for obtaining the single-channel measurement deviation is as follows: Obtain the time of multichannel pipette arm Measured volume of the same target: ,in Channel number , Total number of channels; In the time window Inside, access channel of Valid measurement results Calculate the time window inner passage Average measured volume: ,in To represent the average measured volume, Let i be the end time of the i-th time window. For channel Obtain the sequence number of the measured volume. The length of the time window; The relative measurement deviation of the channel is calculated by comparing the average measured volume with the nominal target volume. ,in The nominal target volume; In the time window First, calculate the average deviation of all channels: ,in The average deviation across all channels; Calculate the dispersion of the channel group: ,in The standard deviation of the channel group; The single-channel measurement deviation is obtained by dividing the difference between the relative measurement deviation of a single channel and the population average deviation by the standard deviation of the population deviation. ,in For single-channel measurement deviation, for This is a very small constant used to prevent division by zero.

5. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 2, characterized in that: The logic for obtaining the electrode response time anomaly coefficient is as follows: Acquire the time-varying electrode signal output by the electrode during actual measurement. ; The initial value of the electrode signal is marked as Stable values ​​are marked as The time required for the electrode signal to reach 95% of its stable value is marked as the electrode response time. ; In the time window Electrode response time for multiple measurements Calculate the single response time deviation: ,in The single response time deviation is calculated for the g-th measurement; Calculate the electrode response time anomaly coefficient: ,in This is the electrode response time anomaly coefficient.

6. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 2, characterized in that: The logic for obtaining the reagent batch storage temperature fluctuation coefficient is as follows: Temperature sensors are placed within the reagent storage environment, and sampling intervals are set. Record the temperature sequence: ,in For the r-th time point The temperature sensor collects the reagent storage temperature, where R is the number of sampling time points; The collected reagents were stored at the same temperature as the recommended storage temperature. Compare and calculate the instantaneous deviation of reagent storage temperature: ,in This is an instantaneous deviation; Calculate the maximum and minimum fluctuations in reagent storage temperature: ,in This represents the maximum to minimum fluctuation in reagent storage temperature. Calculate the cumulative deviation: ,in This is the cumulative deviation. By combining the maximum-minimum temperature fluctuation and cumulative deviation of reagent storage, the batch temperature fluctuation coefficient of reagent is obtained: ,in This refers to the batch storage temperature fluctuation coefficient of the reagent. The maximum and minimum temperature fluctuations during the storage of reagent batch h are given. This represents the cumulative deviation for batch h of reagents. This represents the allowable fluctuation range for batch h of reagents. The total monitoring time for batch h of reagents. , This represents the number of sampling times for the h-th batch of reagents.

7. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 2, characterized in that: The logic for obtaining the cumulative exposure coefficient of reagent batches based on historical usage is as follows: Obtain and number the exposure events that occur during the lifecycle of the target batch of reagents: Where e is the exposure event number, The total number of exposure events; each exposure event includes the following identifier: exposure start time. Exposure end time Exposure duration Average ambient temperature during exposure Recommended storage temperature ; Based on the average ambient temperature during the exposure period Recommended storage temperature Calculate effective overtemperature : ; Based on effective overtemperature Calculate the temperature-time combined exposure for the e-th exposure event: ,in This represents the combined temperature-time exposure. The permissible over-temperature buffer range for the reagent. For reference time scale; Combined with the number of bottles opened during each exposure event Number of cross-environment / cross-device transfers Number of times exposed to strong light and ultraviolet rays Calculate the total exposure amount for the e-th exposure event: ,in Let e ​​be the total exposure amount of the e-th exposure event. These represent preset proportionality coefficients for temperature-time combined exposure, number of bottle openings, number of cross-environment / cross-equipment transfers, and number of exposures to strong light and ultraviolet radiation, respectively. All are greater than 0; For the same reagent batch, the exposure amounts of all exposure events within its lifecycle are summed to obtain the historical cumulative exposure amount for the entire lifecycle: ,in This represents the cumulative historical exposure amount for batch h of reagents. Calculate the historical exposure cumulative factor for reagent batches based on historical usage cumulative exposure: ,in This represents the cumulative exposure factor based on historical usage of the reagent batch. The threshold for safe cumulative exposure.

8. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 1, characterized in that: The step of constructing a critical pressure risk assessment model based on the hidden degradation information of medium and large-scale automated testing equipment and reagent batch difference information, and outputting critical pressure risk status labels that have not triggered alarms but contain hidden systematic deviations, includes the following steps: The hidden degradation information of medium and large-scale automated testing equipment and the batch-to-batch difference information of reagents are used as the input feature vector of the critical pressure risk assessment model: ; Construct a quantization function for the criticality risk assessment model based on the input feature vector: ,in This is the critical risk assessment value. Let d be the index in the input feature vector. Let be the preset proportional coefficient for the d-th indicator, and All are greater than 0; The critical risk assessment value will be compressed. With pressure critical risk threshold Compare and output the critical risk state label. : Among them, the critical risk state label This indicates that although the instrument did not trigger an alarm and the quality control results were still within the acceptable range, the system was approaching a critical state of potential deviation, which could trigger closed-loop intervention.

9. The AI-driven medical laboratory internal audit defect prediction and closed-loop management system according to claim 8, characterized in that: When the critical risk state label is applied At that time, a closed-loop rectification strategy was implemented; The adjustment of the calibration frequency via the calibration unit includes: calculating the calibration frequency adjustment coefficient based on the critical pressure risk assessment value. ,in To calibrate the frequency adjustment factor, This is the current calibration frequency for medium and large-scale automated testing equipment. To adjust the sensitivity coefficient; when and When the calibration frequency is increased, the system automatically generates a new calibration schedule and sends it to the device management terminal; The adjustment of available reagent batches through the reagent batch adjustment unit includes: the reagent batch adjustment unit retrieves all in-stock reagent batch difference information corresponding to the target detection item from the data collection module, including the reagent batch storage temperature fluctuation coefficient and the reagent batch historical use exposure cumulative coefficient for each batch; Calculate the batch sort value for each available and pending batch: ,in This is the sorting value for batch numbered b among the available and pending batches. Preset proportional coefficients for the reagent batch storage temperature fluctuation coefficient and the reagent batch historical usage exposure cumulative coefficient, and All are greater than 0; Sort all batches by batch sorting value from lowest to highest, and divide them into: Normal risk zone: Marked as preferred batch; Critical risk zone: Marked as a restricted batch; High-risk areas: Batch marked as disabled; The reagent batch adjustment unit automatically adjusts the list of available reagent batches accordingly.

Images