Pain evaluation model generation device, pain evaluation model generation method, and recording medium storing program for performing same

Abstract

The present invention relates to a pain evaluation model generation method. One embodiment of the present invention provides a pain evaluation model generation method comprising the steps of: acquiring a photoplethysmogram signal of a subject; selecting at least one input index from the acquired photoplethysmogram signal; and generating a pain evaluation model capable of evaluating the pain of the subject by performing machine learning using the selected input index as input data, wherein the acquired photoplethysmogram includes a first photoplethysmogram acquired at a first time point at which the subject is anesthetized and a second photoplethysmogram acquired at a second time point at which the subject recovers from the anesthesia after the first time point.

Description

Pain assessment model creation device, pain assessment model creation method, and recording medium storing a program for performing the same

[0001] The present invention relates to a pain assessment model generation device, a pain assessment model generation method, and a recording medium storing a program for performing the same.

[0002] In general, accurately assessing a patient's pain level is crucial in clinical settings and actual surgical settings. Therefore, medical staff can accurately and promptly assess a patient's pain to provide appropriate treatment and management.

[0003] To assess the presence and severity of pain during or after surgery, a patient's photoplethysmography (PPG) can be measured to determine the presence or severity of pain. Specifically, PPG is measured by changes in blood flow using sensors attached to the skin, and physiological signals such as heart rate changes can be used to assess pain levels.

[0004] In this case, the patient is anesthetized during surgery, so the impact of their psychological and emotional state on photoplethysmography is minimal. Therefore, the reliability of pain assessment using photoplethysmography is relatively high. However, after waking from anesthesia, the patient's psychological and emotional factors, as well as their cognitive abilities, can influence photoplethysmography measurements, potentially reducing the reliability of pain assessment.

[0005] For this reason, ongoing efforts are being made to develop a pain assessment model that can ensure reliability not only during anesthetized surgery but also after surgery.

[0006] The technical problem to be solved by the present invention is to create a pain assessment model that can reliably evaluate both the pain level of a patient under anesthesia and the pain level of a patient who has recovered from anesthesia.

[0007] One embodiment of the present invention provides a method for generating a pain assessment model, including the steps of: obtaining a photoplethysmography signal of a subject; selecting at least one input index from the obtained photoplethysmography signal; and performing machine learning using the selected input index as input data to generate a pain assessment model capable of evaluating pain of a subject, wherein the obtained photoplethysmography includes a first photoplethysmography acquired at a first time point when the subject is anesthetized and a second photoplethysmography acquired at a second time point when the subject wakes up from anesthesia after the first time point.

[0008] A method for generating a pain assessment model according to one embodiment of the present invention is generated by integrating indices obtained from a photoplethysmogram of a patient obtained during anesthesia and indices obtained from a photoplethysmogram obtained after the patient has recovered from anesthesia, thereby having the effect of reliably evaluating the degree or presence of pain in a patient regardless of whether the patient is anesthetized.

[0009] FIG. 1 is a flowchart illustrating a method for generating a pain assessment model according to one embodiment of the present invention.

[0010] Fig. 2 is a flowchart for explaining a method of selecting an input index using the photoplethysmogram obtained in Fig. 1.

[0011] Figure 3 is a diagram for explaining the process of creating a pain assessment model using the photoplethysmography acquired in Figure 1.

[0012] Figure 4 is a diagram for explaining individual waveforms forming the first photovolume pulse wave and the second photovolume pulse wave.

[0013] Figures 5 to 7 are diagrams for explaining indices extracted from individual waveforms forming a photoplethysmographic pulse.

[0014] Figure 8 is a block diagram illustrating a pain assessment model generation device according to one embodiment of the present invention.

[0015] The present invention is capable of various modifications and embodiments. Specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, as well as the methods for achieving them, will become clearer with reference to the embodiments described in detail below, along with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms.

[0016] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. When describing with reference to the drawings, identical or corresponding components are given the same reference numerals and redundant descriptions thereof will be omitted.

[0017] In the examples below, singular expressions include plural expressions unless the context clearly indicates otherwise.

[0018] In the examples below, terms such as “include” or “have” mean that a feature or component described in the specification is present, and do not preclude the possibility that one or more other features or components may be added.

[0019] In some embodiments, where implementations are otherwise feasible, specific process sequences may be performed in a different order than described. For example, two processes described in succession may be performed substantially simultaneously, or in a reverse order from the described order.

[0020] For convenience of explanation, the sizes of components in the drawings may be exaggerated or reduced. For example, the sizes and thicknesses of each component shown in the drawings are arbitrarily indicated for convenience of explanation, and thus the following embodiments are not necessarily limited to those shown.

[0021] In the description of each component, when it is described as being formed on or under, on and under include both those formed directly or through the intervention of other components, and the standards for on and under are explained based on the drawings.

[0022] In the following examples, when a part is said to be “connected” to another part, this includes not only cases where it is “directly connected” but also cases where it is “electrically connected” with another element in between.

[0023] FIG. 1 is a flowchart for explaining a method for generating a pain assessment model according to one embodiment of the present invention, FIG. 3 is a diagram for explaining a process for generating a pain assessment model using the photoplethysmography obtained in FIG. 1, and FIG. 4 is a diagram for explaining individual waveforms constituting the first photoplethysmography and the second photoplethysmography.

[0024] Referring to FIG. 1, in the step (110) of obtaining a photoplethysmogram (PPG) signal (PPG1, PPG2) of a subject, a processor (820) can obtain a photoplethysmogram signal (PPG1, PPG2) from an optical sensor that senses pulse information of the subject.

[0025] In this specification, pulse information obtained from a subject or object is described as being limited to photoplethysmographic signals (PPG1, PPG2) obtained using light, but is not limited thereto, and the pulse information obtained in the present invention may be obtained using a pressure sensor, an acoustic sensor, a thermal sensor, or a camera sensor.

[0026] The processor (820) can acquire photoplethysmography signals (PPG1, PPG2) at a first time point when the subject is anesthetized and at a second time point when the subject wakes up from anesthesia, respectively. Hereinafter, the 'first photoplethysmography signal (PPG1)' is defined as the photoplethysmography signal (PPG1, PPG2) acquired at the first time point when the subject is anesthetized, and the 'second photoplethysmography signal (PPG2)' is defined as the photoplethysmography signal (PPG1, PPG2) acquired at the second time point when the subject wakes up from anesthesia.

[0027] In addition, 'Time point 1' can be interpreted as the time point during anesthesia surgery, and 'Time point 2' can be interpreted as the time point when the subject wakes up after anesthesia surgery. That is, at Time point 1, since the subject is anesthetized, factors such as the subject's subjective experience, emotional state, or cognitive ability may not have a significant effect on the first photoplethysmography signal (PPG1), and at Time point 2, since the subject is awake from anesthesia, factors such as the subject's subjective experience, emotional state, or cognitive ability may affect the second photoplethysmography signal (PPG2).

[0028] Referring to FIG. 4, the processor (820) can obtain photoplethysmography signals (PPG1, PPG2) from a subject and separate and extract individual waveforms (IP1, IP2) constituting the photoplethysmography signals (PPG1, PPG2). Specifically, the processor (820) can separate the photoplethysmography signals (PPG1, PPG2) into individual waveforms (IP1, IP2) based on the pulse onset of the individual pulse of the photoplethysmography signals (PPG1, PPG2).

[0029] Hereinafter, individual waveforms extracted separately from the first photoplethysmography signal (PPG1) are referred to as first individual waveforms (IP1), and individual waveforms extracted separately from the second photoplethysmography signal (PPG2) are referred to as second individual waveforms (IP2).

[0030] The processor (820) can extract multiple indices from individual waveforms (IP1, IP2), and a description thereof will be provided later in step 221 or step 222 of FIG. 2.

[0031] In the step (120) of selecting at least one input index (IF1, IF2) from the acquired photoplethysmography signal (PPG1, PPG2), the processor (820) can select at least one input index (IF1, IF2) from a plurality of indices extracted from the photoplethysmography signal (PPG1, PPG2). Meanwhile, the process of selecting the input index (IF1, IF2) from a plurality of indices extracted from the photoplethysmography signal (PPG1, PPG2) by the processor (820) will be described later in steps 221 to 241 or steps 222 to 242 of FIG. 2.

[0032] In the step (130) of generating a pain assessment model using selected input indices (IF1, IF2), the processor (820) integrates a first input indices (IF1) selected from a plurality of indices extracted from a first photoplethysmography signal (PPG1) and a second input indices (IF2) selected from a plurality of indices extracted from a second photoplethysmography signal (PPG2), and can generate a pain assessment model using the integrated indices as input data.

[0033] Referring to FIGS. 1 and 3, the processor (820) can generate a pain assessment model capable of evaluating the pain of a subject by performing machine learning using both the first input indicator (IF1) and the second input indicator (IF2) as input data.

[0034] In this specification, the 'subject' is the subject of the photoplethysmography signal (PPG1, PPG2) acquired to extract input data or input indices (IF1, IF2) for generating a pain assessment model. Specifically, the photoplethysmography signal (PPG1, PPG2) of the 'subject' can provide data for machine learning of the pain assessment model or provide input indices (IF1, IF2) that become input data of the pain assessment model.

[0035] Therefore, the 'subject' can be interpreted as a group of people / animals providing the first photoplethysmographic signal (PPG1) and the second photoplethysmographic signal (PPG2).

[0036] Meanwhile, the 'subject' can be interpreted as a human / animal whose pain level is evaluated by the pain assessment model generated above. Specifically, in the present invention, a pain assessment model can be generated using the photoplethysmography signals (PPG1, PPG2) of the 'subject', and the pain level of the 'subject' can be evaluated using the photoplethysmography signals (PPG1, PPG2) of the 'subject' and the generated pain assessment model.

[0037] The processor (820) can perform model training and validation processes using the first input index (IF1) and the second input index (IF2) as input data. In one embodiment, the processor (820) can perform K-fold cross validation, in which case, when verifying the performance of the pain assessment model, the Area Under the ROC Curve (AUC), Balanced Accuracy (BAC), Sensitivity (SE), Specificity (SP), Positive Predictive Value (PPV), or Receiver Operating Characteristic (ROC) Curve can be calculated.

[0038] The processor (820) can perform a K-fold cross-validation process using the bio-signals of the subjects, such as photoplethysmography signals (PPG1, PPG2) obtained from the subjects, as data, and the processor (820) can perform a training and verification process using the first input index (IF1) and the second input index (IF2) as input data of the K-fold cross-validation.

[0039]

[0040] FIG. 2 is a flowchart for explaining a method of selecting an input index from a photoplethysmography signal obtained in FIG. 1, and FIGS. 5 to 7 are drawings for explaining indices extracted from individual waveforms forming a photoplethysmography signal.

[0041] Referring to FIG. 2, in the step (211) of acquiring a first photoplethysmography signal (PPG1) at a time point when the subject is anesthetized, the processor (820) can receive the first photoplethysmography signal (PPG1) acquired at a first time point from the subject, and can separate and extract a plurality of first individual waveforms (IP1) from the received first photoplethysmography signal (PPG1).

[0042] In the step (221) of extracting a plurality of indices from the first photoplethysmography signal (PPG1), the processor (820) can extract a plurality of indices from the first individual waveform (IP1) constituting the first photoplethysmography signal (PPG1).

[0043] The above multiple indicators may include a basic indicator reflecting the characteristics of the first individual waveform (IP1), a normalized indicator normalized by the basic indicators, and an additional indicator developed additionally.

[0044] Additionally, the plurality of indicators may include the maximum, median, and minimum values ​​of the basic indicators of each of the plurality of first individual waveforms (IP1) acquired within the analysis interval, the maximum, median, and minimum values ​​of the normalization indicators, and the maximum, median, and minimum values ​​of the additional indicators.

[0045] The basic indices obtained by the processor (820) from the first individual waveform (IP1) are as shown in [Table 1] below. However, this is only one embodiment, and the composition of the basic indices is not limited to [Table 1] below.

[0046] NNo. Index Definition 1 (max, mid, min) Atotal Pulse area (maximum, median, minimum) 2 (max, mid, min) Asys Systolic area (maximum, median, minimum) 3 (max, mid, min) Adia Diastolic area (maximum, median, minimum) 4 (max, mid, min) TriAtotal Triangular area of ​​the pulse (maximum, median, minimum) 5 (max, mid, min) TriA sys Triangular area (max, mid, min) defined by individual waveforms during the systolic phase6(max, mid, min)TriA dia Triangular area defined by individual waveforms in the diastolic phase (maximum, median, minimum)7(max, mid, min) LsysSystolic length (maximum, median, minimum)8(max, mid, min) LdiaDiastolic length (maximum, median, minimum)9(max, mid, min)PPI sys Time interval (maximum, median, minimum) between systolic extremes of two consecutive individual waveforms10(max, mid, min) PPIonsetTime interval (maximum, median, minimum) between beat onsets of two consecutive individual waveforms11(max, mid, min) PskewSkewness of a waveform (maximum, median, minimum)12(max, mid, min) PkurKurtosis of a waveform (maximum, median, minimum)13(max, mid, min)PW 30 Pulse width (maximum, mid, minimum) at an amplitude equal to 30% of the maximum amplitude of the waveform 14(max, mid, min)PW 50 Pulse width (maximum, mid, minimum) at an amplitude equal to 50% of the maximum amplitude of the waveform 15 (max, mid, min) PW 70Pulse width (maximum, mid, minimum) at an amplitude equal to 70% of the maximum amplitude of the waveform 16 (max, mid, min) PW 90 Pulse width (maximum, mid, minimum) at an amplitude equal to 90% of the maximum amplitude of the waveform17(max, mid, min)ACA bl AC amplitude (max, median, min) at baseline18(max, mid, min)ACAonsetAC amplitude (max, median, min) in the previous diastolic period19(max, mid, min)ACV bl AC change systolic (max, median, min) 20 (max, mid, min) ACVonset AC change diastolic (max, median, min) 21 (max, mid, min) RSrise slope (max, median, min) 22 (max, mid, min)FSdescent slope (max, median, min) 23 (max, mid, min)LRSrise slope length (max, median, min) 24 (max, mid, min)LFSrise slope length (max, median, min) 25 (max, mid, min) RSmax Maximum of rise slope (max, median, min)

[0047] In addition, the normalization indices obtained by the processor (820) from the above-mentioned basic indices are as shown in [Table 2] below. However, this is only one embodiment, and the composition of the normalization indices is not limited to [Table 2] below.

[0048] NNo. Indicator Definition 1 (max, mid, min) Asys / A totalA sys Wow A total The ratio of (max, mid, min)2(max, mid, min) Adia / A totalA dia Wow A total The ratio of (max, mid, min)3(max, mid, min) Asys / A diaA sys Wow A diaThe ratio of (max, mid, min)4(max, mid, min) Atotal / ACA blA 와tal Wow ACA bl The ratio of (max, mid, min)5(max, mid, min)A sys / ACA bl A sys Wow ACA bl The ratio of (max, mid, min)6(max, mid, min)A dia / ACA bl A dia Wow ACA bl The ratio of (max, mid, min)7(max, mid, min)TriA sys / TriATriA sys and the ratio of TriA (max, mid, min)8(max, mid, min)TriA dia / TriATriA dia and the ratio of TriA (max, mid, min)9(max, mid, min)TriA sys / TriA dia TriA sys Wow TriA dia The ratio of (max, mid, min) 10 (max, mid, min) Lsys / PPI onsetL sys Wow PPI onset The ratio of (max, mid, min) Ldia / PPI onsetL dia Wow PPI onset The ratio of (max, mid, min) 12 (max, mid, min) Lsys / L diaL sys Wow L dia The ratio of (max, mid, min)ACV13(max, mid, min) bl / ACA onset ACV bl Wow ACA onset The ratio of (max, mid, min)14(max, mid, min)ACV bl / ACA bl ACV blWow ACA bl Ratio of (max, mid, min) 15(max, mid, min) ACVonset / ACA onsetACV onset Wow ACA onset The ratio of (max, mid, min)16(max, mid, min)ACV onset / ACA bl ACV onset Wow ACA bl Ratio of (max, mid, min) 17(max, mid, min) RS max / RSRS max and the ratio of RS (max, mid, min)18(max, mid, min)RS / FS and the ratio of RS to FS (max, mid, min)19(max, mid, min)RS / ACA bl RS and ACA bl The ratio of (max, mid, min)20(max, mid, min) (Atotal / ACA bl ) / L sys(A total / ACA bl ) and L sys The ratio of (max, mid, min)21(max, mid, min) (Atotal / ACA bl ) / L dia(A total / ACA bl ) and L dia The ratio of (max, mid, min)22(max, mid, min) (Atotal / ACA bl ) / PPI onset(A total / ACA bl ) and PPI onset The ratio of (max, mid, min)23(max, mid, min)(A sys / ACA bl ) / L sys (A sys / ACA bl ) and L sys The ratio of (max, mid, min)24(max, mid, min)(A sys / ACA bl ) / L dia (Asys / ACA bl ) and L dia The ratio of (max, mid, min)25(max, mid, min)(A sys / ACA bl ) / PPI onset (A sys / ACA bl ) and PPI onset The ratio of (max, mid, min)26(max, mid, min)(A dia / ACA bl ) / L sys (A dia / ACA bl ) and L sys The ratio of (max, mid, min)27(max, mid, min)(A dia / ACA bl ) / L dia (A dia / ACA bl ) and L dia The ratio of (max, mid, min)28(max, mid, min)(A dia / ACA bl ) / PPI onset (A dia / ACA bl ) and PPI onset ratio of (maximum, median, minimum)

[0049] Additionally, additional indicators are as shown in [Table 3] below. However, this is only one example, and the composition of additional indicators is not limited to [Table 3] below.

[0050] NNo. Indicator Definition 1 (max, mid, min)AL Area under an individual waveform to the left of the midpoint of the maximum pulse width (maximum, median, minimum) 2 (max, mid, min)AR Area under an individual waveform to the right of the midpoint of the maximum pulse width (maximum, median, minimum) 3 (max, mid, min)ALR diff Difference between AR and AL (maximum, median, minimum)4(max, mid, min) ALRdiffratioALR diff Wow A leftAbsolute difference (max, mid, min) 5(max, mid, min) LdiffL dia Wow L sys Difference (max, median, min)6(max, mid, min) PVsys Amplitude difference (max, median, min)7(max, mid, min) PVonset Amplitude difference (max, median, min)8(max, mid, min) between beat onsets of adjacent individual waveforms / A totalALR diff Wow A total The ratio of (max, mid, min)9(max, mid, min)AR / ACA bl AR and ACA bl The ratio of (max, mid, min)10(max, mid, min)AL / ACA bl AL and ACA bl The ratio of (maximum, median, minimum)11(max, mid, min)PW 30 / PPI onset PW 30 Wow PPI onset The ratio of (maximum, median, minimum)12(max, mid, min)PW 50 / PPI onset PW 50 Wow PPI onset The ratio of (maximum, median, minimum)13(max, mid, min)PW 70 / PPI onset PW 70 Wow PPI onset The ratio of (maximum, median, minimum)14(max, mid, min)PW 90 / PPI onset PW 90 Wow PPI onset The ratio of (max, mid, min)15(max, mid, min)(RS / ACA bl ) / PPI onset (RS / ACA bl ) and PPI onset The ratio of (max, mid, min)16(max, mid, min)(AR / ACA bl) / PP Ionset (AR / ACA bl ) and PPI onset The ratio of (max, mid, min)17(max, mid, min)(AL / ACA bl ) / PPI onset (AL / ACA bl ) and PPI onset ratio of (maximum, median, minimum)

[0051] That is, in order to select the first input index (IF1), the indexes extracted from the first individual waveform (IP1) are the maximum value (max), median value (mid), and minimum value (min) of 25 basic indexes, 28 normalized indexes, and 17 additional indexes, and thus a total of 210 indexes were extracted. Referring to FIG. 3, the processor (820) selected an index with high performance among the plurality of indexes as the first input index (IF1) through model learning.

[0052] Specifically, the processor (820) repeatedly calculates Shapley values ​​of multiple indices through bootstrapping, and selects an indices having a high average of absolute values ​​of the Shapley values ​​obtained through repeated calculations as the first input indices (IF1).

[0053] Hereinafter, the ‘representative Shapley value’ of a specific indicator is defined as ‘the average of the absolute values ​​of the Shapley values ​​of a specific indicator obtained by repeatedly calculating through bootstrapping.’

[0054] In the step (231) of repeatedly calculating Shapley values ​​of a plurality of indicators, the processor (820) performs bootstrapping through model learning using a decision tree-based ensemble model, thereby repeatedly calculating Shapley values ​​of a plurality of indicators, and specifically, the processor (820) can repeatedly calculate Shapley values ​​of a plurality of indicators based on a model based on XGBoost (eXtreme Gradient Boosting) as a decision tree-based ensemble model.

[0055] Additionally, the processor (800) can calculate a representative Shapley value by calculating the average of the absolute values ​​of Shapley values ​​repeatedly calculated for each indicator.

[0056] Referring to FIG. 3, as a result of the processor (820) calculating the representative Shapley values ​​of the plurality of indices extracted from the first individual waveform (IP1), the skewness of the individual waveform (IP1) having the smallest absolute value among the skewnesses of the plurality of individual waveforms (IP1) forming the first photoplethysmographic signal (PPG1) is min P The representative Shapley value of skew was calculated to be large.

[0057] Below, the skewness of the individual waveform (IP1) P Explains the skew index. The skewness of a waveform can be defined as the degree to which the waveform is skewed from the center, and a waveform with a large absolute value of skewness can form a shape that is relatively more biased to one side than a waveform with a relatively small absolute value of skewness.

[0058] Referring to Fig. 5, the 1-1 individual waveform (IP1-1) and the 1-2 individual waveform (IP1-2) are different individual waveforms (IP1) extracted separately from the 1st photoplethysmographic signal (PPG1). The 1-1 individual waveform (IP1-1) has a shape that is more left-biased with respect to the center of the waveform than the 1-2 individual waveform (IP1-2), and accordingly, the absolute value of the skewness of the 1-1 individual waveform (IP1-1) is measured to be relatively larger than the absolute value of the skewness of the 1-2 individual waveform (IP1-2).

[0059] The processor (820) calculates the representative Shapley values ​​of the multiple indices extracted from the first individual waveform (IP1), and the pulse width (PW) at an amplitude corresponding to 50% of the maximum amplitude of the waveform 50 ) and the time interval (PPI) between the onsets of two consecutive individual waveforms (IP1). onset ) maximum value of the ratio (max PW) 50 / PPI onset ) was calculated to be a large representative Shapley value.

[0060] Below, the pulse width (PW) at an amplitude corresponding to 50% of the maximum amplitude of the waveform 50 ) and the time between the onset of the beats of two consecutive individual waveforms (IP1, IP2) (PPI onset ) interval ratio (max PW 50 / PPI onset ) explains the indicators.

[0061] Referring to Fig. 6, the pulse width (PW) at an amplitude corresponding to 50% of the maximum amplitude of the individual waveforms (IP1, IP2) 50 ) is the maximum amplitude (Amp) of individual waveforms (IP1, IP2) sys ) can be interpreted as the time interval between two points having an amplitude value corresponding to 50% of the pulse rate. In addition, the time interval (PPI) between the onset points of the pulses of two individual waveforms (IP1, IP2) onset) can be interpreted as the time interval between the beat onset of one individual waveform (IP1, IP2) and the beat onset of the adjacent individual waveform (IP1, IP2).

[0062] Through this, in the step (241) of selecting an index with a high representative Shapley value as the first input index (IF1), the processor (820) min P skew and max PW 50 / PPI onset can be selected as the first input index (IF1), and the process is selected from the indices of the first photoplethysmographic signal (PPG1). min P skew and max PW 50 / PPI onset It can be used as input data for creating a pain assessment model using the first input indicator (IF1).

[0063] In this specification, 'amplitude' can be interpreted as a displacement value at a specific time.

[0064] As a result, as a result of the processor (820) calculating the representative Shapley values ​​of the plurality of indices extracted from the first photoplethysmogram signal (PPG1), the first input indices (IF1) can be extracted as the skewness of the individual waveforms (IP1) constituting the first photoplethysmogram signal (PPG1). Specifically, the first input indices (IF1) can be extracted as the skewness of the individual waveforms (IP1) having the smallest absolute value among the skewnesses of the plurality of individual waveforms (IP1) constituting the first photoplethysmogram signal (PPG1).

[0065] In addition, as a result of the processor (820) calculating the representative Shapley value of a plurality of indices extracted from the first photoplethysmography signal (PPG1), the first input indices (IF1) can be extracted based on the pulse width at a preset amplitude of the individual waveform (IP1) constituting the first photoplethysmography signal (PPG1), and specifically, the first input indices (IF1) can be calculated as a ratio of the pulse width at the preset amplitude of the individual waveform (IP1) and the time interval between the start point of the pulsation of the individual waveform (IP1), and the preset amplitude can be half of the maximum amplitude of the individual waveform (IP1).

[0066] In this specification, 'pulse width' can be interpreted as the length in the X-axis direction of the coordinate system shown in FIGS. 5 to 8, and can be interpreted as the period of the waveform or the time interval between two points on the waveform.

[0067] Referring to FIG. 2, in the step (212) of acquiring a second photoplethysmography signal (PPG2) at the time when the subject wakes up from anesthesia, the processor (820) can receive the second photoplethysmography signal (PPG2) acquired at the second time from the subject, and can separate and extract a plurality of second individual waveforms (IP2) from the received second photoplethysmography signal (PPG2).

[0068] In the step (222) of extracting a plurality of indices from the second photoplethysmography signal (PPG2), the processor (820) can extract a plurality of indices from the second individual waveform (IP2) constituting the second photoplethysmography signal (PPG2).

[0069] Since the above-described plurality of indicators are the same as the plurality of indicators extracted from the first individual waveform (IP1) described above, a description of the types of the plurality of indicators extracted from the second individual waveform (IP2) is omitted.

[0070] That is, the indicators extracted from the second individual waveform (IP2) to select the second input indicator (IF2) are the maximum, median, and minimum values ​​of 25 basic indicators, 28 normalized indicators, and 17 additional indicators, and thus a total of 210 indicators were extracted.

[0071] Referring to FIG. 3, the processor (820) selects a second input indicator (IF2) among the multiple indicators with high performance through model learning.

[0072] Specifically, the processor (820) repeatedly calculates Shapley values ​​of multiple indices extracted from the second individual waveform (IP2) through bootstrapping, and selects an index with a high absolute value average of the multiple Shapley values ​​obtained through repeated calculations as the second input index (IF2).

[0073] The step (232) of repeatedly calculating Shapley values ​​of a plurality of indices extracted from the second individual waveform (IP2) is the same process as the step (231) of repeatedly calculating Shapley values ​​of a plurality of indices extracted from the first individual waveform (IP1), so the description in the overlapping range is omitted.

[0074] Referring to FIG. 3, as a result of the processor (820) calculating the representative Shapley values ​​of the plurality of indices extracted from the second individual waveform (IP2), the minimum value of the triangular area defined by the individual waveform (IP2) in the systolic phase of the individual waveform (IP2) forming the second photoplethysmographic signal (PPG2) is min TriA sys ) was calculated to be a representative Shapley value for the corresponding indicator.

[0075] Below, the triangle area (TriA) defined by the individual waveforms (IP1, IP2) in the systolic phase of the individual waveforms (IP1, IP2) sys ) explains the indicators.

[0076] Referring to Figure 7, the triangle defined by the individual waveforms (IP1, IP2) in the systolic phase of the individual waveforms (IP1, IP2) is the time interval (RT) between the systolic poles from the onset of the beat of the individual waveforms (IP1, IP2). sys ) as the base, and the height is the amplitude difference between the systolic peak and the pulsation onset of the individual waveforms (IP1, IP2). Alternatively, the triangle defined by the individual waveforms (IP1, IP2) in the systolic phase of the individual waveforms (IP1, IP2) can be defined as a right triangle whose hypotenuse is the line segment connecting the systolic peak and the pulsation onset of the individual waveforms (IP1, IP2).

[0077] The processor (820) calculates the representative Shapley values ​​of the multiple indices extracted from the second individual waveform (IP2), and the maximum value (max PV) of the amplitude difference between the beat onset points of adjacent individual waveforms (IP2) onset ) was calculated to be a large representative Shapley value.

[0078] Through this, in the step (242) of selecting an index with a high representative Shapley value as the second input index (IF2), the processor (820) min TriA sys Wow max PV onset can be selected as the second input index (IF2), and the process is selected from the indices of the second photoplethysmographic signal (PPG2). min TriA sys Wow max PV onset It can be used as input data for creating a pain assessment model using the second input indicator (IF2).

[0079] As a result, as a result of the processor (820) calculating the representative Shapley values ​​of the plurality of indices extracted from the second photoplethysmography signal (PPG2), the second input indices (IF2) can be calculated as the product of the interval from the start point of the pulse of the individual waveform (IP2) constituting the second photoplethysmography signal (PPG2) to the maximum amplitude of the individual waveform (IP2) and the maximum amplitude of the individual waveform (IP2).

[0080] In addition, as a result of the processor (820) calculating the representative Shapley values ​​of the plurality of indices extracted from the second photoplethysmography signal (PPG2), the second input indices (IF2) can be calculated as the difference in amplitude between the pulsation starting point of one individual waveform (IP2) among the plurality of individual waveforms (IP2) constituting the second photoplethysmography signal (PPG2) and the pulsation starting point of the individual waveform (IP2) adjacent to the one individual waveform (IP2).

[0081] In the step (130) of creating a pain assessment model using the aforementioned input indices (IF1, IF2) as input data, the processor (820) can learn, verify, and create a pain assessment model using both the selected first input indices (IF1) and the second indices as input data.

[0082] That is, the processor (820) can generate a pain assessment model using both the first input index (IF1) acquired at the time the patient was anesthetized and the second input index (IF2) acquired at the time after the patient woke up from anesthesia, and thus, the pain assessment model generated by the present invention is generated by considering both the data at the time the patient was anesthetized and the data at the time the patient woke up from anesthesia, thereby having the effect of being able to evaluate the pain of the subject with high reliability regardless of the anesthesia state of the subject.

[0083]

[0084] Figure 8 is a block diagram illustrating a pain assessment model generation device according to one embodiment of the present invention.

[0085] A pain assessment model generation device (800) according to one embodiment of the present invention may include at least a memory (810) and a processor (820). Only components related to the embodiment are illustrated in the pain assessment model generation device (800) of FIG. 8. Therefore, those skilled in the art will appreciate that the pain assessment model generation device (800) may further include other general-purpose components in addition to the components illustrated in FIG. 8.

[0086] The processor (820) controls the overall operation of the pain assessment model generation device (800). For example, the processor (820) may control the overall operation of the memory (810), a communication module (not shown), an input unit (not shown), and / or an output unit (not shown) by executing programs stored in the memory (810). The processor (820) may control the operation of the pain assessment model generation device (800) by executing programs stored in the memory (810).

[0087] The processor (820) can control at least a part of the operations of the pain assessment model generation method described above with reference to FIGS. 1 to 7. For example, the processor (820) can obtain the photoplethysmography signals (PPG1, PPG2) of the subject to select input indices (IF1, IF2), and can generate a pain assessment model using the selected input indices (IF1, IF2) as input data. Furthermore, the processor (820) can evaluate the degree / presence of pain of the subject by using the photoplethysmography signals (PPG1, PPG2) obtained from the subject and the pain assessment model.

[0088] Meanwhile, a specific example of how the processor (820) operates is the same as that described above with reference to FIGS. 1 to 7. Therefore, a specific description of the operation of the processor (820) is omitted below.

[0089] The processor (820) may be implemented using at least one of ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), DSPDs (Digital Signal Processing Devices), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), controllers, micro-controllers, microprocessors, and other electrical units for performing functions.

[0090] The memory (810) is hardware that stores various data processed within the pain assessment model generation device (800), and can store programs for various operations, processing, and control of the processor (820).

[0091] The memory (810) may include a random access memory (RAM) such as a dynamic random access memory (DRAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a CD-ROM, a Blu-ray or other optical disk storage, a hard disk drive (HDD), a solid state drive (SSD), or a flash memory (810).

[0092] The communication module may include at least one component that enables the pain assessment model generation device (800) to perform wired / wireless communication with a device that acquires a photoplethysmography signal and / or another external device. For example, the communication module may include a wired communication unit for implementing Ethernet, serial communication, or optical communication, and / or a wireless communication unit for implementing Wi-Fi, Bluetooth, or cellular network-based communication.

[0093] Meanwhile, embodiments according to the present invention may be implemented in the form of a computer program that can be executed through various components on a computer, and such a computer program may be recorded on a computer-readable medium. At this time, the medium may include, but is not limited to, magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program instructions, such as ROMs, RAMs, and flash memory (810).

[0094] Meanwhile, the computer program may be specifically designed and constructed for the present invention, or may be one known and available to those skilled in the computer software field. Examples of computer programs may include not only machine language code, such as that generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

[0095] Methods according to various embodiments of the present invention may be provided as included in a computer program product. The computer program product may be traded as a product between a seller and a buyer. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or may be distributed online (e.g., downloaded or uploaded) via an application store (e.g., Play Store™) or directly between two user devices. In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily generated in a machine-readable storage medium, such as a memory (810) of a manufacturer's server, an application store's server, or a relay server.

[0096] Unless the steps constituting the method according to the present invention are explicitly described in a specific order or are otherwise described in a different order, the steps may be performed in any appropriate order. The present invention is not necessarily limited to the order in which the steps are described. The use of all examples or exemplary terms (e.g., “for example,” “etc.”) in the present invention is merely intended to illustrate the present invention in more detail, and the scope of the present invention is not limited by the examples or exemplary terms unless otherwise defined by the claims. Furthermore, those skilled in the art will appreciate that various modifications, combinations, and variations can be configured according to design conditions and factors within the scope of the appended claims or their equivalents.

[0097] The spirit of the present invention should not be limited to the embodiments described above, and all scopes equivalent to or equivalent to the scope of the following claims, as well as the scope of the present invention, are considered to fall within the scope of the present invention.

[0098] According to one embodiment of the present invention, a pain assessment model generation device, a pain assessment model generation method, and a recording medium storing a program for performing the same are provided. Furthermore, embodiments of the present invention can be applied to methods or devices for evaluating physical data used in industry.

Claims

1. A step of acquiring a photoplethysmographic signal of a subject; A step of selecting at least one input index from the acquired photoplethysmography signal; and A step of generating a pain assessment model capable of evaluating the pain of a subject by performing machine learning on the selected input indicators as input data; A method for creating a pain assessment model, wherein the photoplethysmogram obtained above includes a first photoplethysmogram obtained at a first time point when the subject is anesthetized and a second photoplethysmogram obtained at a second time point after the first time point when the subject wakes up from anesthesia.

2. In paragraph 1, The step of selecting the above input indicator is: A step of selecting at least one first input index from the first photoplethysmogram; and A method for creating a pain assessment model, comprising: a step of selecting at least one second input index different from the first input index from the second photoplethysmogram; 3. In paragraph 2, A method for creating a pain assessment model, wherein the first input index is extracted as the skewness of an individual waveform forming the first photoplethysmographic wave.

4. In paragraph 3, A method for creating a pain assessment model, wherein the first input index is extracted as the skewness of an individual waveform having the smallest absolute value among the skewnesses of a plurality of individual waveforms forming the first photoplethysmographic wave.

5. In paragraph 2, A method for creating a pain assessment model, wherein the first input indicator is extracted based on a pulse width at a preset amplitude of an individual waveform forming the first photoplethysmographic wave.

6. In paragraph 5, A method for creating a pain assessment model, wherein the preset amplitude is half the maximum amplitude of the individual waveform.

7. In paragraph 5, A method for creating a pain assessment model, wherein the first input index is calculated as a ratio of the pulse width at the preset amplitude of the individual waveform and the time interval (PPIonset) between the start point of the pulse of the individual waveform.

8. In paragraph 2, A method for creating a pain assessment model, wherein the second input index is calculated as the product of the interval from the start point of the pulse of each individual waveform forming the second photoplethysmographic wave to the maximum amplitude of each individual waveform and the maximum amplitude of each individual waveform.

9. In paragraph 2, A method for creating a pain assessment model, wherein the second input index is calculated as the difference in amplitude between the pulsation starting point of one individual waveform among a plurality of individual waveforms forming the second photoplethysmographic wave and the pulsation starting point of an individual waveform adjacent to the one individual waveform.

10. In paragraph 2, The steps for creating the above pain assessment model are: A method for creating a pain assessment model, wherein at least one of the first input indicators and at least one of the second input indicators are both input data.

11. In paragraph 1, The step of selecting the above input indicator is: A step of repeatedly calculating Shapley values ​​of multiple indices extracted from the photoplethysmographic pulse; and A method for creating a pain assessment model, comprising: a step of selecting an index having a high absolute value average of the Shapley values ​​obtained by repeatedly calculating the above as the input index; 12. In paragraph 11, A method for creating a pain assessment model, wherein the Shapley values ​​of the above plurality of indicators are calculated based on an XGBoost (eXtreme Gradient Boosting)-based model as an ensemble model based on a decision tree.

13. A computer-readable recording medium storing a program including commands for performing each step according to the method described in any one of claims 1 to 12.

14. Memory in which at least one program is stored; and A processor that operates by executing at least one program; The above processor, A pain assessment model can be created by acquiring a photoplethysmography signal of a subject, selecting at least one input index from the acquired photoplethysmography signal, and performing machine learning on the selected input index as input data. A pain assessment model generation device, wherein the acquired photoplethysmogram includes a first photoplethysmogram acquired at a first time point when the subject is anesthetized and a second photoplethysmogram acquired at a second time point after the first time point when the subject wakes up from anesthesia.

Images