An intelligent triage scheduling system based on patient data flow

Abstract

This invention relates to the field of machine learning scheduling technology, specifically to an intelligent triage and scheduling system based on patient data streams. The system includes a vital sign acquisition module, a deterioration distribution module, an equipment triggering module, an access control addressing module, and a triage and scheduling module. In this invention, multidimensional patient monitoring data is extracted and compared with a safety baseline within a continuous observation window. Deviations are continuously accumulated and cross-aggregated to construct an objective quantitative baseline characterizing the degree of deterioration in the patient's overall condition. When the deterioration score exceeds the takeover threshold, the system skips the software queuing process to locate an available physical emergency space. The underlying network switches electronic access control and emergency equipment relays to the activation level, forcibly opening the physical emergency channel. This eliminates the time loss caused by critically ill patients waiting for room cleaning and equipment activation, achieving an instantaneous transition from critical vital sign data to hardware-level scheduling and control commands, ensuring that critically ill patients receive timely and effective treatment.

Description

Technical Field

[0001] This invention relates to the field of machine learning scheduling technology, and in particular to an intelligent triage and scheduling system based on patient data streams. Background Technology

[0002] The field of machine learning scheduling technology encompasses the core content of determining resource allocation and task execution order by mining features from past samples and real-time inputs. It relies on probabilistic statistical inference, extracting attribute features from historical allocation cases and establishing parameter weights in iterations to form a decision-making process for assigning new task sequences to corresponding execution nodes. Among these, intelligent triage and scheduling systems refer to technologies for prioritizing patients in hospital outpatient and emergency departments and allocating resources to doctors' offices. Typically, triage staff manually measure and record patients' basic physiological indicators (temperature, pulse, blood pressure, heart rate) and their reported pain levels. Based on a pre-established medical emergency and criticality level scoring standard table, each physiological indicator value is scored individually and then summed. Subsequently, based on the total score, patients are ranked in descending order and placed into the corresponding specialist clinic's appointment list, awaiting their number to be called by the doctor.

[0003] In the current triage and scheduling process, medical staff mainly rely on manually measuring and recording single static physiological values ​​and performing comparison and scoring. Its operation is essentially a software-level queue sorting and updating mechanism. A single static assessment cannot capture the continuous evolution trend of the rapid deterioration of the patient's vital signs. Simply arranging patients into the regular clinic list according to the total score and waiting for their numbers to be called in sequence breaks the underlying linkage mechanism between medical data flow and physical rescue resources. As a result, even after critically ill patients obtain the highest priority, they are still limited by physical obstacles such as access control in the emergency room and the power being turned on to the rescue equipment. This can easily delay the opportunity for emergency intervention and cause serious delays in triage and scheduling. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent triage and scheduling system based on patient data streams.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent triage and scheduling system based on patient data streams, comprising:

[0006] The vital signs acquisition module collects the patient's heart rate monitoring data, compares the patient's heart rate monitoring data with the preset heart rate safety benchmark, and calculates the deviation between the two to obtain the vital signs deviation difference.

[0007] The deterioration distribution module determines a preset time observation window limit, accumulates the deviation difference of the vital signs for each occurrence within a measurement cycle that continuously reaches the preset time observation window limit, extracts the heart rate deviation accumulation, synchronously collects the patient's blood oxygen monitoring data and compares it with a preset blood oxygen safety benchmark, extracts the blood oxygen deviation accumulation accumulated over the same measurement cycle, and performs cross-convergence of the heart rate deviation accumulation and blood oxygen deviation accumulation to construct a concurrent deterioration integral.

[0008] The equipment triggering module compares the concurrent deterioration score with the preset emergency equipment takeover triggering threshold. If the concurrent deterioration score is greater than the preset emergency equipment takeover triggering threshold, the patient is transferred to the emergency department and a takeover triggering identifier is established.

[0009] The access control addressing module, based on the takeover trigger identifier, obtains all emergency room electronic access control network addresses and emergency instrument power relay control addresses, filters emergency room electronic access control network addresses that are in an idle state, and outputs linkage addressing records;

[0010] The triage and scheduling module, based on the linked addressing record, switches the electronic access control network address and the relay control address to the start level, synchronously extracts the patient's identity information and displays it on the terminal scheduling screen, thus constructing an intelligent triage and scheduling sequence.

[0011] As a further aspect of the present invention, when comparing the concurrent deterioration score with the preset emergency equipment takeover trigger threshold, if the concurrent deterioration score is not greater than the preset emergency equipment takeover trigger threshold, the patient is transferred to a regular outpatient clinic.

[0012] As a further aspect of the present invention, the deviation of vital signs includes the amplitude of rhythm abnormality, pulsatility overload reading, and ischemia variability index; the calculation parameters of the complication deterioration score include the quantitative value of the severity of the condition and the quantitative value of the hypoxia compensation rating; the takeover trigger identifier includes the highest priority emergency rescue identifier, green passage instruction tag, and specialist triage certificate; the linkage addressing record includes the access control medium access control bit, the defibrillator power supply gateway, and the ventilator node road sign; the intelligent triage scheduling sequence includes spatial allocation priority, physician intervention shift, and instrument start-stop time axis.

[0013] As a further aspect of the present invention, the vital sign acquisition module includes:

[0014] The monitoring and receiving submodule collects patient heart rate monitoring data output by the triage station acquisition device, synchronously acquires a preset heart rate safety benchmark, extracts the internal time node identifiers of the patient heart rate monitoring data and the preset heart rate safety benchmark, sorts the time node identifiers according to the time sequence, maps the original acquisition content according to the sorting position, and generates a time-series distribution quantity.

[0015] The comparison execution submodule separates the time-series distribution quantity carrying the monitoring value and the predetermined benchmark, performs a layer-by-layer logical comparison operation between the monitoring value and the predetermined benchmark, filters the difference status identifiers, collects the node scales corresponding to the difference status identifiers, and obtains the fluctuation range value.

[0016] The deviation statistics submodule extracts the fluctuation range value along with the classification attribute label, performs a recombination and aggregation action based on the classification attribute label, extracts the same type of integrated state produced by the recombination and aggregation action, integrates the discrete performance of the same type of integrated state, and obtains the deviation difference of vital signs.

[0017] As a further aspect of the present invention, the deterioration distribution module includes:

[0018] The heart rate normalization submodule extracts a preset time observation window limit, performs time-series node arrangement on the vital sign deviation difference within the corresponding period when the preset time observation window limit is reached, performs numerical accumulation operation, and generates heart rate deviation accumulation.

[0019] The blood oxygen accumulation submodule collects the patient's blood oxygen monitoring data and a preset blood oxygen safety benchmark, subtracts the patient's blood oxygen monitoring data from the preset blood oxygen safety benchmark, extracts the deviation sequence, and performs a step-by-step summation operation on the deviation sequence according to the same measurement period to obtain the blood oxygen deviation accumulation.

[0020] The cross-convergence submodule, based on the heart rate deviation accumulation and the blood oxygen deviation accumulation, obtains the critical reference quantity of heart rate, the critical reference quantity of blood oxygen, and the physiological lag reference period, performs compensated cross-product calculation, and outputs the concurrent deterioration integral.

[0021] As a further aspect of the present invention, the device triggering module includes:

[0022] The threshold comparison submodule obtains the preset emergency equipment takeover trigger threshold, divides the concurrent deterioration integral by the preset emergency equipment takeover trigger threshold to solve the proportional quotient value, determines the digital region where the proportional quotient value is relative to the constant, extracts the digital region polarity mapping parameter, and generates the comparison difference quantity.

[0023] The diversion and scheduling submodule filters the comparison difference to a negative decision domain corresponding to a constant, parses the outpatient waiting scheduling vector according to the negative decision domain, issues the outpatient waiting scheduling vector to execute the ordinary outpatient transfer action, and obtains the regular waiting record.

[0024] The takeover identification submodule filters the comparison difference to a positive decision domain corresponding to a value greater than a constant, parses the emergency dispatch vector based on the positive decision domain, executes the emergency transfer action, writes the intervention code into the logic register, and establishes a takeover trigger identification.

[0025] As a further aspect of the present invention, the access control addressing module includes:

[0026] The address extraction submodule, based on the takeover trigger identifier, parses and obtains the network access control address and the instrument power network address, extracts the local area network identifier to which the network access control address and the instrument power network address belong, performs physical network segmentation, and generates a node distribution list;

[0027] The idle filtering submodule extracts the communication occupancy digital parameter within the node distribution list, compares the communication occupancy digital parameter with the idle state benchmark value, removes the device array units corresponding to the idle state benchmark value, and obtains the available node matrix.

[0028] The routing linkage submodule extracts the communication underlying routing path contained in the available node matrix, arranges the sending end communication protocol quantity and the receiving end communication protocol quantity along the communication underlying routing path, and performs a concatenation operation on the sending end communication protocol quantity and the receiving end communication protocol quantity to obtain the linkage addressing record.

[0029] As a further aspect of the present invention, the triage and scheduling module includes:

[0030] The level switching submodule parses the linkage addressing record to obtain the electronic access control network address and the relay control address, extracts preset high and low level parameters, switches the electronic access control network address and the relay control address to the start level according to the preset high and low level parameters, and generates node level status quantities.

[0031] The encoding and parsing submodule, based on the synchronization control timing parsed from the node level state quantity, regulates the identity acquisition device to receive local data in the corresponding timing window, performs regular expression matching extraction on the local data, separates redundant interference characters, and obtains patient identity information.

[0032] The sequence construction submodule imports the patient's identity information into the internal display memory area of ​​the terminal scheduling screen, executes the interface mapping display action, arranges the interface mapping items in chronological order, and establishes an intelligent triage scheduling sequence.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] In this invention, multidimensional monitoring data of patients is extracted and compared with safety benchmarks through a continuous observation window. The deviation difference is continuously accumulated and cross-aggregated to construct an objective quantitative benchmark characterizing the degree of deterioration of the patient's overall condition. When the deterioration score exceeds the takeover threshold, the software queuing process is skipped to locate an available physical emergency space. The underlying network switches the electronic access control and the relays of the emergency instruments to the start level to forcibly open the physical emergency channel, eliminating the time loss caused by critically ill patients waiting for room cleaning and equipment startup. This enables the instantaneous conversion of critical vital signs data into hardware-level scheduling and control commands, ensuring that critically ill patients can receive effective treatment in a timely manner. Attached Figure Description

[0035] Figure 1 This is a system flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the vital sign acquisition module of the present invention;

[0037] Figure 3 This is a flowchart of the deterioration distribution module of the present invention;

[0038] Figure 4 This is a flowchart of the device triggering module of the present invention;

[0039] Figure 5 This is a flowchart of the access control addressing module of the present invention;

[0040] Figure 6 This is a flowchart of the triage and scheduling module of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention 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 merely illustrative and not intended to limit the invention.

[0042] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0043] Please see Figure 1 An intelligent triage and scheduling system based on patient data streams includes:

[0044] The vital signs acquisition module collects the patient's heart rate monitoring data from the triage station's acquisition equipment, and simultaneously acquires a preset heart rate safety benchmark. The patient's heart rate monitoring data is compared with the preset heart rate safety benchmark, and the deviation between the two is calculated to obtain the vital signs deviation difference.

[0045] The deterioration distribution module extracts the preset time observation window limit, accumulates the deviation difference of vital signs for each measurement cycle that continuously reaches the preset time observation window limit, extracts the heart rate deviation accumulation, collects the patient's blood oxygen monitoring data and compares it with the preset blood oxygen safety benchmark, extracts the blood oxygen deviation accumulation accumulated over the same measurement cycle, and performs cross-convergence of the heart rate deviation accumulation and blood oxygen deviation accumulation to construct the concurrent deterioration integral;

[0046] The equipment triggering module obtains the preset emergency equipment takeover triggering threshold, compares the concurrent deterioration score with the preset emergency equipment takeover triggering threshold, and judges whether the concurrent deterioration score is greater than the preset emergency equipment takeover triggering threshold. If the concurrent deterioration score is greater than the preset emergency equipment takeover triggering threshold, the patient is transferred to the emergency department and a takeover triggering mark is established. If the concurrent deterioration score is not greater than the preset emergency equipment takeover triggering threshold, the patient is transferred to the general outpatient waiting area.

[0047] The access control addressing module obtains all emergency room electronic access control network addresses and emergency instrument power relay control addresses based on the takeover trigger identifier, filters emergency room electronic access control network addresses that are in an idle state, and outputs linkage addressing records;

[0048] The triage and scheduling module, based on the linkage addressing record, switches the electronic access control network address and the relay control address to the start level, synchronously extracts the patient's identity information and displays it on the terminal scheduling screen, thus constructing an intelligent triage and scheduling sequence.

[0049] Deviations in vital signs include abnormal rhythm amplitude, pulsatility overload readings, and ischemia variability index; complications and deterioration scores include records of the severity of the condition and hypoxia compensation ratings; takeover trigger indicators include the highest priority emergency resuscitation indicator, green access instruction sign, and specialist triage certificate; linkage addressing records include access control media access control bits, defibrillator power supply gateways, and ventilator node road signs; intelligent triage and scheduling sequences include spatial allocation priority, physician intervention shifts, and instrument start-stop timelines.

[0050] Please see Figure 2 The vital signs acquisition module includes:

[0051] The monitoring and receiving submodule collects the patient's heart rate monitoring data output by the triage station's data acquisition equipment, simultaneously acquires the preset heart rate safety benchmark, extracts the internal time node identifiers of the patient's heart rate monitoring data and the preset heart rate safety benchmark, sorts the time node identifiers according to the order of time, maps the original collection content according to the sorting position, and generates time-series distribution data.

[0052] The monitoring and receiving submodule uses a microcontroller unit integrated within a medical-grade fingertip pulse oximeter to drive alternating light emission from a 660nm red LED and a 905nm infrared LED. The light penetrates the capillary bed at the patient's fingertip and is received by a photodetector. The photodetector converts the received light intensity changes into a continuous microampere-level current signal. A transimpedance amplifier circuit converts this current signal into a millivolt-level analog voltage signal. Subsequently, a built-in 16-bit high-precision analog-to-digital converter discretizes the analog voltage signal at a sampling frequency of 125 times per second, converting it into a raw digital heart rate waveform sequence. To eliminate baseline drift caused by patient movement and environmental power frequency electromagnetic interference, the monitoring and receiving submodule incorporates a one-dimensional convolutional neural network hardware inference acceleration engine. The input layer of this acceleration engine contains a buffer register array that receives the original waveform tensor with a length of 512 sampling points. The first hidden layer is configured as a one-dimensional convolutional layer with 64 convolutional kernels of size 7 and a stride of 1. The convolution operation is performed by a hardware multiply-accumulator array. The output channel is connected to a linear rectified function activation circuit to truncate all negative noise features. The subsequent max pooling layer has a window size of 2 and a stride of 2 to achieve data downsampling while preserving the peak features of the waveform. The second hidden layer is a fully connected layer with 128 neurons. Finally, the output layer outputs smoothed and denoised patient heart rate monitoring data through regression mapping, specifically 85 beats per minute. This one-dimensional convolutional neural network, a classic architecture in machine learning for temporal signal processing and deep feature extraction, underwent offline training before leaving the factory. The training set consisted of 10,000 segments of annotated, absolutely clean heart rate waveforms from an open-source medical database. The loss function used the mean squared error formula to calculate the mean squared difference between the network's output waveform and the standard waveform. Backpropagation employed an adaptive moment estimation optimization algorithm to update the weight distribution of each layer. The learning rate hyperparameter was selected as 0.0005 through grid search cross-validation, the batch size was set to 32, and the number of iterations was set to 150. This parameter combination exhibited the fastest and non-oscillating convergence trajectory during gradient descent.

[0053] The monitoring receiving submodule synchronously reads the preset heart rate safety baseline stored in the internal read-only memory. This baseline is defined according to the World Health Organization's adult heart rate health standards, with a lower limit of 60 beats per minute and an upper limit of 100 beats per minute. The submodule extracts the internal time node identifiers based on the Coordinated Universal Time (CoUT) standard that accompany the patient's heart rate monitoring data. The hardware sorter uses a bubble sort logic array to sort up to 1024 timestamps in the buffer pool in ascending order. Based on the recombined timestamp index address, the sorter remaps the heart rate values ​​in the random access memory through the direct memory access channel, combining them to form a time-series distribution containing both time and numerical dimensions, which is stored in a dual-port static random access memory.

[0054] The comparison execution submodule separates the timing distribution quantity carrying the monitoring value and the predetermined benchmark, performs a layer-by-layer logical comparison operation between the monitoring value and the predetermined benchmark, filters the difference status identifiers, collects the node scales corresponding to the difference status identifiers, and obtains the fluctuation range value.

[0055] The comparison execution submodule reads the static random access memory via the data bus. The hardware parser decomposes the timing distribution into 8-bit monitoring value bus signals and predetermined upper and lower limit signals. The comparison execution submodule has a built-in bipolar digital comparator logic array that performs a layer-by-layer logical comparison operation between the monitoring value and the predetermined benchmark. The first-level comparator compares the monitoring value 85 with the upper limit value 100, outputting a negative sign bit, indicating that the upper limit is not exceeded. The second-level comparator compares the monitoring value 85 with the lower limit value 60, outputting a positive sign bit, indicating that the lower limit is exceeded. The logic gate circuit outputs a normal identification code based on the results of the two comparisons. If the input monitoring value is 115, the sign bit of the first-level comparator flips, outputting an over-limit interrupt signal, which the hardware logic encodes as 1, indicating a high abnormality. If the monitoring value is 55, it is encoded as -1, indicating a low abnormality, thus filtering out the difference status indicators. The counter circuit collects the corresponding node scales where the difference status flag is not 0 within 100 consecutive sampling periods. When 8 consecutive high abnormal flags are detected, the arithmetic logic unit calculates the area of ​​the difference between the peak extreme value and the safety upper limit to obtain the fluctuation range value. For example, the over-limit integral is calculated to be 45.

[0056] The deviation statistics submodule extracts the fluctuation range value with attached classification attribute labels, performs recombination and aggregation actions based on the classification attribute labels, extracts the same type of integrated state from the recombination and aggregation actions, integrates the discrete performance of the same type of integrated state, and obtains the deviation difference of vital signs.

[0057] The deviation statistics submodule receives a data packet containing fluctuation interval values ​​with an out-of-limit integral of 45, and reads the additional classification attribute label field. The classification attribute labels are obtained through a lookup table; the data dictionary defines the label mapping code for "tachycardia" as 0x01 and the label mapping code for "arrhythmia" as 0x02. Based on the classification attribute label 0x01, the microprocessor activates the hardware hash mapping table, importing fluctuation interval values ​​with the same label into the corresponding physical memory block for recombination and aggregation. For all data within the 0x01 block, the mean calculator extracts the aggregated state produced by the recombination and aggregation action, calculating the average deviation amplitude of this interval as 22. The adder tree array integrates the single-point discrete performance of the aggregated state, calculates the variance value as 3.5, and packages and compresses the average deviation amplitude and variance value to obtain the vital sign deviation difference data packet, which is output to the lower-level module communication interface.

[0058] Table 1 Comparison of Feature Extraction Schemes for Heart Rate Monitoring Data

[0059]

[0060] Please see Figure 3 The deterioration distribution module includes:

[0061] The heart rate normalization submodule extracts the preset time observation window limit, performs time sequence node arrangement on the deviation difference of vital signs within the corresponding period of reaching the preset time observation window limit, performs numerical accumulation operation, and generates heart rate deviation accumulation.

[0062] The heart rate normalization submodule extracts a preset time observation window limit from the control register. This limit is set based on the physiological cycle of vital sign changes in critically ill patients, specifically a value of 300 seconds. A hardware timer starts counting down. Within the corresponding 300-second period, the timing arranger within the heart rate normalization submodule extracts all vital sign deviation difference data packets stored in the static random access memory (SRAM) for that period. Based on the timestamp field in the packet header, the timing arranger calls the fast sorting hardware logic circuit to perform a strict ascending order sorting of the vital sign deviation differences. The accumulator hardware logic unit then follows the sorted data link, sequentially feeding the average deviation value extracted from each data packet into a 64-bit adder register to perform numerical accumulation. For example, if five valid deviations are captured within the current 300-second period, with values ​​of 12, 15, 18, 10, and 20 respectively, the accumulator will add 12 and 15 to get 27, then add 18 to get 45, and so on. Finally, the register locks the sum value, generates the specific value of heart rate deviation accumulation as 75, and writes it to a dedicated shared memory area, marking the valid bit for subsequent module calls.

[0063] The blood oxygen accumulation submodule collects the patient's blood oxygen monitoring data and the preset blood oxygen safety benchmark, subtracts the patient's blood oxygen monitoring data from the preset blood oxygen safety benchmark, extracts the deviation sequence quantity, and performs a term-by-term summation operation on the deviation sequence quantity according to the same measurement period to obtain the blood oxygen deviation accumulation.

[0064] The pulse oximetry accumulation submodule synchronously activates the pulse oximetry data acquisition and communication interface. The microcontroller unit reads the patient's pulse oximetry monitoring data from the pulse oximeter front-end module via the integrated circuit built-in memory bus protocol. The raw data read is the pulse oximetry saturation percentage calculated based on light absorption rate, with the specific monitoring value for the current cycle being 91. The pulse oximetry accumulation submodule retrieves the preset pulse oximetry safety benchmark from non-volatile memory. This benchmark is based on the criteria for severe respiratory illnesses and is rigidly set to 95. The subtraction logic unit receives the patient's pulse oximetry monitoring data 91 and the preset pulse oximetry safety benchmark 95, performs a two's complement subtraction operation on the two, discards the sign bit, and extracts the absolute value to obtain a deviation sequence quantity of 4. Within the same measurement cycle of 300 seconds, which is equivalent to the preset time observation window limit, the shift register continuously captures the deviation sequence quantity obtained from the new round of sampling, for example, subsequently acquiring 5, 6, 4, and 7 in sequence. The hardware adder array performs a step-by-step summation operation on this set of deviation sequences according to the same measurement period. The operation process is 4 plus 5 to get 9, then 6 plus 6 to get 15, and so on, until the final output shows the specific value of the accumulated blood oxygen deviation as 26.

[0065] The cross-convergence submodule obtains the critical baseline values ​​of heart rate and blood oxygen and the physiological hysteresis reference period based on the heart rate deviation accumulation and blood oxygen deviation accumulation, performs compensated cross-product calculation, and outputs the concurrent deterioration integral.

[0066] The cross-convergence submodule synchronously receives the specific values ​​of heart rate deviation accumulation (75) and blood oxygen deviation accumulation (26) generated by the preceding link via its internal high-speed data bus. Subsequently, this submodule retrieves the pre-set critical heart rate and blood oxygen threshold values ​​from the configuration flash memory via direct memory access. These two threshold values ​​are medically calibrated to 125 and 40, respectively, based on clinical experience in intensive care. Simultaneously, it acquires the physiological hysteresis reference period, reflecting the conduction delay between the patient's cardiovascular and respiratory systems. This parameter is quantized into a hysteresis compensation coefficient at the underlying level, with a fixed value of 1.584. Under the scheduling of the hardware-level floating-point unit, the logic circuit first performs normalization processing, dividing the heart rate deviation accumulation (75) by the critical heart rate threshold (125) to obtain a relative heart rate deviation rate of 0.600; similarly, dividing the blood oxygen deviation accumulation (26) by the critical blood oxygen threshold (40) to obtain a relative blood oxygen deviation rate of 0.650. Next, the cross-multiplier array multiplies these two deviation rates, resulting in an initial cross product of 0.390. Considering the inherent physiological time lag between the decrease in blood oxygen saturation and the compensatory increase in heart rate, the module must perform time deviation correction. The computational core performs a compensated cross-product operation with the initial cross-product of 0.390 and the compensation coefficient of 1.584 corresponding to the physiological hysteresis reference period, completing the fine-tuning of the temporal dimension weights. Finally, the precise output value of the concurrent deterioration integral is 0.618, which is directly stored in the high-speed buffer register, providing a key input for the takeover judgment of the next-level equipment.

[0067] Please see Figure 4 The equipment trigger module includes:

[0068] The threshold comparison submodule obtains the preset emergency equipment takeover trigger threshold, divides the concurrent deterioration integral by the preset emergency equipment takeover trigger threshold to solve the proportional quotient value, determines the digital region where the proportional quotient value is relative to the constant, extracts the digital region polarity mapping parameter, and generates the comparison difference quantity.

[0069] The threshold comparison submodule retrieves the preset emergency equipment takeover trigger threshold from the rule base of the hospital information management core server via a secure socket layer encrypted channel. This threshold parameter has undergone retrospective receiver operating characteristic (ROC) curve analysis testing using large-scale clinical historical samples to select the inflection point corresponding to the optimal Youden index, with the specific fixed value precisely set to 0.550. This inflection point selection is not static; the backend synchronously integrates and introduces the K-means clustering algorithm logic from unsupervised machine learning. Through periodic unsupervised iterative clustering analysis of the vital sign data of hundreds of newly added ICU patients each day, the preset emergency equipment takeover trigger threshold can dynamically learn and self-adjust as the overall patient population characteristics drift across different epidemic cycles. The digital signal processor of the threshold comparison submodule receives the concurrent degradation integral of 0.618 from the previous link, calls the high-precision floating-point divider logic, uses the concurrent degradation integral of 0.618 as the dividend, and the preset emergency equipment takeover trigger threshold of 0.550 as the divisor, performs a 32-bit floating-point division operation, and obtains a ratio quotient of 1.123. The comparator logic unit then intervenes to determine the digit region of the ratio quotient 1.123 relative to the constant 1. The comparator compares it with the hard-coded constant 1.000, and detects that the most significant bit does not generate a borrow, determining that it falls into the positive judgment region greater than the constant 1. The hardware mapping table extracts the digit region polarity mapping parameters according to the memory address index of the region. Since it is in the positive region, the polarity mapping parameter outputs the hexadecimal status word 0xFF, representing the alarm activation polarity. The logic controller merges and packages this polarity parameter with the original calculated deviation value to generate a long-byte comparison difference data packet.

[0070] The triage and scheduling submodule filters and compares the differences and assigns them to a negative decision domain that is no greater than a constant. Based on the negative decision domain, it parses the outpatient waiting scheduling vector, issues the outpatient waiting scheduling vector to execute the ordinary outpatient transfer action, and obtains the regular waiting record.

[0071] The triage scheduling submodule is equipped with a bidirectional multiplexer that continuously monitors the comparison difference data stream on the bus. The multiplexer's control end performs conditional branch judgments. For another patient condition, if the detected ratio quotient is 0.850, the hardware comparator determines that it belongs to the corresponding negative decision domain not greater than a constant 1, and the polarity mapping parameter outputs a hexadecimal status word 0x00. The triage scheduling submodule filters the patient's condition associated with this 0x00 flag, activates the read-only memory sector associated with the negative decision domain, and parses out the preset outpatient waiting scheduling vector. This vector encapsulates specific triage target address parameter groups, such as floor identifier code set to 2, department identifier code set to Internal Medicine Outpatient Room 13, and queuing priority code set to 3. The network interface controller encapsulates this outpatient waiting scheduling vector into a Hypertext Transfer Protocol request body and sends the outpatient waiting scheduling vector through the Ethernet physical layer chip, instructing the outpatient electronic queuing terminal to perform a regular outpatient transfer action. The database write engine synchronously captures the confirmation receipt of this action and adds a new row of data to the triage database table using Structured Query Language, obtaining the regular waiting record.

[0072] The takeover identifier submodule filters and compares the differences and assigns them to the positive decision domain corresponding to the constant. Based on the positive decision domain, it parses the emergency dispatch vector, executes the emergency transfer action, writes the intervention code into the logic register, and establishes the takeover trigger identifier.

[0073] The takeover identifier submodule is synchronously connected in parallel to another output channel of the multiplexer. When it filters and intercepts the hexadecimal status word 0xFF with polarity mapping parameters for the comparison difference in this embodiment, it confirms that it belongs to the positive decision domain corresponding to a value greater than the constant 1. At this time, the priority interrupt controller immediately pulls the highest-level unmasked interrupt pin high, suspending all low-priority background tasks. The takeover identifier submodule parses the emergency dispatch vector containing extremely high privileges based on the high-dimensional memory address space mapped by the positive decision domain. The parameters covered by this vector include the emergency resuscitation room equipment matrix network mask, the defibrillator preheating activation command word, and the infusion pump channel opening command word. The central processing unit core takes over this vector, executes the emergency transfer action, and controls the automated guided vehicle in the emergency resuscitation room to prepare for transfer. At the same time, the state machine in the field-programmable gate array is triggered, directly writing the emergency intervention code (specifically configured as 0xAA55) to a specific non-volatile ferroelectric random access memory logic register (physical address 0x40021000). After successful writing, the state machine locks the address area, establishes an immutable takeover trigger flag, and releases the highest level of hardware authorization for subsequent network addressing and direct control of underlying physical medical devices.

[0074] Table 2. Scheduling Vector Control Parameter Mapping Configuration Table

[0075]

[0076] Referring to Table 2, this hardware-level threshold comparison and vector scheduling mechanism can seize the critical golden window of opportunity for rescue.

[0077] Please see Figure 5 The access control addressing module includes:

[0078] The address extraction submodule parses and obtains the network access control address and the instrument power network address based on the takeover trigger identifier, extracts the local area network identifier of the network access control address and the instrument power network address, performs physical network segmentation, and generates a node distribution list.

[0079] The address extraction submodule periodically polls the non-volatile ferroelectric random access memory (FRAM) logic register located at physical address 0x40021000 via its internal high-speed peripheral component interconnect bus. When the polling reads the value 0xAA55, the address extraction submodule immediately activates the hard-wired network topology resolution engine based on this takeover trigger flag. The resolution engine reads the pre-configured device asset dynamic host configuration protocol binding table, performs a hash matching lookup in the table, and resolves to obtain the specific network access control address responsible for the intensive care area (specifically extracted as IPv4 address 10.10.2.150) and the instrument power network address of the supporting resuscitation and emergency equipment (specifically extracted as IPv4 address 10.10.3.200). Subsequently, the address extraction submodule performs a bitwise AND operation with the obtained IP address using the subnet mask 255.255.255.0 to extract the LAN identifier of the network access control address as 10.10.2.0 and the LAN identifier of the instrument power supply network address as 10.10.3.0. The network controller sends a configuration message to the core switch, modifies the VLAN tag control information, and performs logical network segmentation to completely isolate the above two network segments from the regular office network at the Layer 2 data link layer. It also writes the media access control address list of all online devices within the isolated network segments into memory, generating a node distribution list containing 35 devices.

[0080] The idle filtering submodule extracts the communication occupancy digital parameters within the node distribution list, compares the communication occupancy digital parameters with the idle state benchmark value, removes the device array units corresponding to the idle state benchmark value, and obtains the available node matrix.

[0081] The idle filtering submodule sends status query packets to each physical media access control address registered in the node distribution list via a simple network management protocol (SMART) polling interface. The query messages penetrate the underlying network to reach the network interface cards (NICs) of each device terminal. The NICs return data frames containing the current CPU utilization, remaining memory space, and real-time port traffic status. The idle filtering submodule extracts communication utilization parameters from the node distribution list within the returned data frames. In this data processing process, the idle filtering submodule not only relies on the static status information fed back by these NIC ports in real time, but its internal core computing unit also calls a machine learning hidden Markov model based on probabilistic statistical algorithms in parallel. Based on the network congestion state transition probability matrix of the local area network's physical location at various historical moments, it proactively predicts and extrapolates the network bandwidth consumption trend of each underlying communication node in the next time window, thereby minimizing transient jitter errors. Based on this composite mechanism, the submodule specifically extracts that the overall communication bandwidth utilization rate of a smart access control node on the emergency transport channel is 65%, and the network utilization rate of another defibrillator is 15%. The idle filtering submodule retrieves the idle state baseline value set in the cache area within the control chip. This baseline value is set to 30% based on the requirement of ensuring emergency response commands. A digital comparator compares the extracted communication occupancy parameter (65%) with the idle state baseline value (30%). If 65% is detected as significantly exceeding the baseline value, the logic state machine triggers a removal instruction, erasing the logical mapping of the corresponding device from the available pool and removing the array unit corresponding to the idle state baseline value. For nodes with an occupancy rate of 15%, their logical mapping is retained. After a full list traversal and filtering, the idle filtering submodule obtains a matrix of available nodes with guaranteed latency and smooth operation. This matrix is ​​stored in static random access memory as a multi-dimensional array.

[0082] The routing linkage submodule extracts the available node matrix containing the underlying communication routing path, arranges the sending end communication protocol quantity and the receiving end communication protocol quantity along the underlying communication routing path, and performs a concatenation operation on the sending end communication protocol quantity and the receiving end communication protocol quantity to obtain the linkage addressing record;

[0083] The routing linkage submodule obtains the memory pointer of the available node matrix and calls the internal open shortest path first (BSF) routing calculation hardware core module. The routing calculation core extracts the underlying communication routing paths from the available node matrix and uses a fixed Dijkstra's shortest path algorithm logic gate array to calculate the optimal physical link from the main triage station to the resuscitation room equipment gateway (e.g., triage station network card -> core routing switch port 01 -> floor aggregation switch port 05 -> terminal equipment gateway). Along this generated underlying communication routing path, the protocol encapsulator begins arranging the network control commands to be issued. The encapsulator pushes the sending-end communication protocol quantity into the front end of the data packet, including the transmission control protocol header information, the source port number 8080, and the triage station IP address; and pushes the receiving-end communication protocol quantity into the rear end of the data packet, including the target device port number 502 (a typical industrial control protocol port), the target device gateway IP address, and a cyclic redundancy check (CRC) code. The data bus splicer uses a shift register queue to perform a rigorous bit-by-bit concatenation operation on the transmitting and receiving communication protocol quantities in a streaming manner, filling in specific emergency takeover load instructions in between, resulting in a 128-byte, structurally complete linkage addressing record. This record is directly pushed into the transmission queue of the network card's advanced programmable interrupt controller, awaiting transmission, ensuring the directional and unstoppable nature of network link transmission.

[0084] Please see Figure 6 The triage and scheduling module includes:

[0085] The level switching submodule parses the linkage addressing record to obtain the electronic access control network address and relay control address, extracts the preset high and low level parameters, switches the electronic access control network address and relay control address to the start level according to the preset high and low level parameters, and generates node level status quantities;

[0086] The network receiver adapter of the level switching submodule captures the linkage addressing record on the bus and calls the internal hardware-level deep packet inspection and decapsulation circuit to perform layer-by-layer stripping of the 128-byte data packet. By reading the offset address, it parses the linkage addressing record to obtain the electronic access control network address (e.g., specific decimal address code 301 corresponds to emergency room door number one) and relay control address (e.g., specific decimal code 505 corresponds to emergency room life support equipment power strip) located in the load segment. The level switching submodule consults the underlying pin mapping register table to extract the preset high and low level parameters. These parameters are set according to the industrial-grade relay operation specifications, specifically setting the drive engagement level to a high level of 5 volts and the disconnect release level to a low level of 0 volts. The microprocessor controls the digital-to-analog converter module of the general-purpose input / output port to instantaneously increase the control voltage of the physical pin mapped to the electronic access control network address 301 and the physical pin mapped to the relay control address 505 according to the preset high and low level parameters of 5 volts, switching to the start level of the drive engagement state. When the access control electromagnetic lock relay and power switch contactor are activated by a high level, the state latch inside the level switching submodule captures the current change fed back by the pin and generates a node level state quantity that indicates the open state, marking that the physical channel and device have been successfully prepared.

[0087] The encoding and parsing submodule uses the synchronization control timing parsed from the node level status quantity to regulate the identity acquisition device to receive local data in the corresponding timing window, performs regular expression matching extraction on the local data, separates redundant interference characters, and obtains patient identity information.

[0088] After the encoding and parsing submodule detects that the node level output of the state latch has changed to a high logic level, it sends a clock activation signal to the high-definition camera and near-field communication RFID reader located on the triage desk, activating the identity acquisition device. The RFID reader generates a 13.56 MHz high-frequency electromagnetic field through a coil, exciting the RFID chip inside the patient's hospital card, and receives the returned local data sequence containing the identity verification code and magnetic stripe encryption verification information; the high-definition camera simultaneously acquires facial contour video streams as auxiliary local data. To maximize the fault tolerance and parsing speed of identity verification in critical care situations, when processing the continuous facial contour video stream input from the high-definition camera, the underlying image processing unit activates a pre-loaded deep residual network visual feature extraction model based on machine learning mechanisms. Within a time window of hundreds of milliseconds, it performs cross-comparison and identification of key facial biometric anchor points of the patient, thereby assisting in locking in matching items in the patient database. For the read radio frequency local data string containing jumbled tabs and prefix codes (such as "#ID_NO:32010219800101XXXX_END@"), the hardware-level regular expression matching coprocessor built into the encoding and parsing submodule performs regular expression matching extraction. The coprocessor loads the fixed matching pattern "\d{18}" and performs character-by-character scanning and comparison using a finite state automaton to accurately separate redundant interference characters such as "#ID_NO:" and "_END@", obtaining a pure and accurate 18-digit national identity card number string as the patient's identity information in this emergency situation.

[0089] The sequence construction submodule imports patient identity information into the internal display memory area of ​​the terminal scheduling screen, executes interface mapping display actions, arranges interface mapping items according to time sequence, and establishes an intelligent triage scheduling sequence.

[0090] The sequence construction submodule, through a high-speed direct memory access controller, directly imports the acquired high-risk patient identity information data block into a dedicated double-buffered video memory area within the terminal scheduling screen. The graphics processing unit takes over the data in this memory block, calls the built-in 2D rendering engine, extracts the red warning color scheme template and extra-large display font parameters for emergency priority, executes interface mapping and display actions, binds and merges the identity information with the previously calculated concurrent deterioration integral, generates a striking red alert layer, and pushes it to the top of the main display. Simultaneously, the sequence construction submodule queries the coordinate information of other waiting patients in the video memory, calls the hardware sorting tree logic, forcibly interrupts the original registration order logic, and strictly follows the time sequence of the high-risk alerts to forcibly shift and reconstruct all interface mapping entries. The mapping entry for the high-risk patient identity information is absolutely placed at the top of the 0th row, 0th column pixel coordinate starting position, establishing a completely new intelligent triage and scheduling sequence. The video output interface projects the sequence onto a large screen at a refresh rate of 60 Hz via a high-definition multimedia interface protocol. In the event of a sudden critical illness, the scheduling delay time for updating the triage interface is reduced, ensuring the real-time access to information by medical staff.

[0091] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An intelligent triage and scheduling system based on patient data stream, characterized in that, include: The vital signs acquisition module collects the patient's heart rate monitoring data, compares the patient's heart rate monitoring data with the preset heart rate safety benchmark, and calculates the deviation between the two to obtain the vital signs deviation difference. The deterioration distribution module determines a preset time observation window limit, accumulates the deviation difference of the vital signs for each occurrence within a measurement cycle that continuously reaches the preset time observation window limit, extracts the heart rate deviation accumulation, synchronously collects the patient's blood oxygen monitoring data and compares it with a preset blood oxygen safety benchmark, extracts the blood oxygen deviation accumulation accumulated over the same measurement cycle, and performs cross-convergence of the heart rate deviation accumulation and blood oxygen deviation accumulation to construct a concurrent deterioration integral. The deterioration distribution module includes: The heart rate normalization submodule extracts a preset time observation window limit, performs time-series node arrangement on the vital sign deviation difference within the corresponding period when the preset time observation window limit is reached, performs numerical accumulation operation, and generates heart rate deviation accumulation. The blood oxygen accumulation submodule collects the patient's blood oxygen monitoring data and a preset blood oxygen safety benchmark, subtracts the patient's blood oxygen monitoring data from the preset blood oxygen safety benchmark, extracts the deviation sequence, and performs a step-by-step summation operation on the deviation sequence according to the same measurement period to obtain the blood oxygen deviation accumulation. The cross-convergence submodule, based on the heart rate deviation accumulation and the blood oxygen deviation accumulation, obtains the critical heart rate reference quantity, the critical blood oxygen reference quantity, and the physiological hysteresis reference period, performs compensated cross-product calculation, and outputs the concurrent deterioration integral. The equipment triggering module compares the concurrent deterioration score with the preset emergency equipment takeover triggering threshold. If the concurrent deterioration score is greater than the preset emergency equipment takeover triggering threshold, the patient is transferred to the emergency department and a takeover triggering identifier is established. The access control addressing module, based on the takeover trigger identifier, obtains all emergency room electronic access control network addresses and emergency instrument power relay control addresses, filters emergency room electronic access control network addresses that are in an idle state, and outputs linkage addressing records; The triage and scheduling module, based on the linked addressing record, switches the electronic access control network address and the relay control address to the start level, synchronously extracts the patient's identity information and displays it on the terminal scheduling screen, thus constructing an intelligent triage and scheduling sequence.

2. The intelligent triage and scheduling system based on patient data stream according to claim 1, characterized in that: When comparing the concurrent deterioration score with the preset emergency equipment takeover trigger threshold, if the concurrent deterioration score is not greater than the preset emergency equipment takeover trigger threshold, the patient is transferred to a regular outpatient clinic.

3. The intelligent triage and scheduling system based on patient data stream according to claim 1, characterized in that: The deviation of vital signs includes the amplitude of abnormal rhythm, pulsatility overload reading, and ischemia variability index; the calculation parameters of the complication deterioration score include the quantitative value of the severity of the condition and the quantitative value of the hypoxia compensation rating; the takeover trigger identifiers include the highest priority emergency rescue identifier, green passage instruction token, and specialist triage certificate; the linkage addressing record includes the access control medium access control bit, defibrillator power supply gateway, and ventilator node road sign; the intelligent triage scheduling sequence includes spatial allocation priority, physician intervention shift, and instrument start-stop time axis.

4. The intelligent triage and scheduling system based on patient data stream according to claim 1, characterized in that, The vital sign acquisition module includes: The monitoring and receiving submodule collects patient heart rate monitoring data output by the triage station acquisition device, synchronously acquires a preset heart rate safety benchmark, extracts the internal time node identifiers of the patient heart rate monitoring data and the preset heart rate safety benchmark, sorts the time node identifiers according to the time sequence, maps the original acquisition content according to the sorting position, and generates a time-series distribution quantity. The comparison execution submodule separates the time-series distribution quantity carrying the monitoring value and the predetermined benchmark, performs a layer-by-layer logical comparison operation between the monitoring value and the predetermined benchmark, filters the difference status identifiers, collects the node scales corresponding to the difference status identifiers, and obtains the fluctuation range value. The deviation statistics submodule extracts the fluctuation range value along with the classification attribute label, performs a recombination and aggregation action based on the classification attribute label, extracts the same type of integrated state produced by the recombination and aggregation action, integrates the discrete performance of the same type of integrated state, and obtains the deviation difference of vital signs.

5. The intelligent triage and scheduling system based on patient data stream according to claim 1, characterized in that, The device triggering module includes: The threshold comparison submodule obtains the preset emergency equipment takeover trigger threshold, divides the concurrent deterioration integral by the preset emergency equipment takeover trigger threshold to solve the proportional quotient value, determines the digital region where the proportional quotient value is relative to the constant, extracts the digital region polarity mapping parameter, and generates the comparison difference quantity. The diversion and scheduling submodule filters the comparison difference to a negative decision domain corresponding to a constant, parses the outpatient waiting scheduling vector according to the negative decision domain, issues the outpatient waiting scheduling vector to execute the ordinary outpatient transfer action, and obtains the regular waiting record. The takeover identification submodule filters the comparison difference to a positive decision domain corresponding to a value greater than a constant, parses the emergency dispatch vector based on the positive decision domain, executes the emergency transfer action, writes the intervention code into the logic register, and establishes a takeover trigger identification.

6. The intelligent triage and scheduling system based on patient data stream according to claim 1, characterized in that, The access control addressing module includes: The address extraction submodule, based on the takeover trigger identifier, parses and obtains the network access control address and the instrument power network address, extracts the local area network identifier to which the network access control address and the instrument power network address belong, performs physical network segmentation, and generates a node distribution list; The idle filtering submodule extracts the communication occupancy digital parameters within the node distribution list, compares the communication occupancy digital parameters with the idle status baseline value, and obtains the available node matrix. The routing linkage submodule extracts the communication underlying routing path contained in the available node matrix, arranges the sending end communication protocol quantity and the receiving end communication protocol quantity along the communication underlying routing path, and performs a concatenation operation on the sending end communication protocol quantity and the receiving end communication protocol quantity to obtain the linkage addressing record.

7. The intelligent triage and scheduling system based on patient data stream according to claim 6, characterized in that: The available node matrix specifically refers to the remaining device array units after removing device array units that are greater than the idle state baseline value.

8. The intelligent triage and scheduling system based on patient data stream according to claim 1, characterized in that, The triage and scheduling module includes: The level switching submodule parses the linkage addressing record to obtain the electronic access control network address and the relay control address, extracts preset high and low level parameters, switches the electronic access control network address and the relay control address to the start level according to the preset high and low level parameters, and generates node level status quantities. The encoding and parsing submodule, based on the synchronization control timing parsed from the node level state quantity, regulates the identity acquisition device to receive local data in the corresponding timing window, performs regular expression matching extraction on the local data, separates redundant interference characters, and obtains patient identity information. The sequence construction submodule imports the patient's identity information into the internal display memory area of ​​the terminal scheduling screen, executes the interface mapping display action, arranges the interface mapping items in chronological order, and establishes an intelligent triage scheduling sequence.

Images