Arousal level measurement method and arousal level measurement apparatus

Abstract

To provide an arousal level measurement method that identifies the occurrence order of specific ocular and pupillary movements that appear with a decrease in arousal, and determines the arousal level.SOLUTION: An arousal level measurement method measures changes in eye movement and pupil diameter by performing ocular and pupillary measurements, detects, as an ocular alertness index (Ocular Alertness Index) for evaluating arousal, a shortening of intersaccadic intervals (Frequent saccade), a decrease in peak saccadic velocity (Slow saccade), slow fluctuations in eye position (Slow eye movement), and the co-occurrence (Div-mio) of divergence contrary to the near reflex (Divergence) and miosis (Miosis), and evaluates the decrease level of arousal on the basis of the occurrence order of the ocular alertness indices.SELECTED DRAWING: Figure 14

Description

Technical Field

[0001] The present invention relates to a method for measuring an arousal level and an apparatus for measuring an arousal level, which determine an arousal level based on eye pupil measurement.

Background Art

[0002] In recent years, for the purpose of preventing human errors that lead to fatal accidents, and providing an environment suitable for quick sleep and effective work / research performance, the need to monitor arousal levels has been increasing. Therefore, an arousal monitoring system for detecting various levels of arousal decline is required so that appropriate interventions can be made at each level of arousal.

[0003] Conventionally, as biological indicators indicating arousal levels, electroencephalograms, heart rate variability, eye reactions such as pupil dilation and constriction, and vestibulo-ocular reflexes have been used.

[0004] It is known that various brain states are reflected in eye pupil movement in various forms. For example, it has been demonstrated that a decrease in pupil constriction, divergence, and vestibulo-ocular reflex (VOR) appears before we recognize our own sleepiness (Patent Document 1).

[0005] Also, after becoming aware of one's own sleepiness, the number of blinks increases, the inter-saccadic interval (ISI) decreases, the peak velocity of saccades (SC) decreases, slow eye movement (SEM) of slow eye position appears, and the gains of optokinetic eye movement and smooth pursuit eye movement decrease.

[0006] Thus, specific eye pupil movements (Ocular Alertness Index: OAI) that rarely appear in a high arousal state can be a reliable indicator for objective vigilance monitoring.

Prior Art Documents

Patent Document

[0007]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0008] However, in previous studies, a decrease in arousal was detected only from specific OAIs, and other potential OAIs were not evaluated simultaneously. Furthermore, the relationship between each OAI and the level of arousal has not been clarified either.

[0009] Therefore, an object of the present invention is to provide a method for measuring an arousal level and an apparatus for measuring an arousal level, which specify the order of appearance of specific ocular pupil movements (OAIs) that appear as the arousal level decreases, and determine the arousal level.

Means for Solving the Problems

[0010] In order to achieve the above object, in the invention described in claim 1, a method for measuring an arousal level, measures changes in eye movement and pupil diameter by eye pupil measurement, as an ocular alertness index for evaluating arousal, based on the eye movement measurement results, detects a shortening of the interval between saccades (Frequent saccade), a decrease in the peak velocity of saccades (Slow saccade), and a slow oscillation of the eye position (Slow eye movement), based on the measurement results of eye movement and pupil diameter, detects the co-occurrence of divergence (Div-Mio) and miosis (Div-Mio) that are contrary to the near reflex, is a method for measuring an arousal level that evaluates the level of decrease in arousal based on the order of appearance of the ocular alertness index, Frequent saccade is Detect those below the saccade interval threshold set based on the saccade intervals that cannot occur in a highly awakened state as Frequent saccade, Slow saccade and Slow eye movement are In the main sequence diagram of saccades consisting of the amplitude and peak velocity of eye movements, set a region where the peak velocity is low with respect to the amplitude that cannot occur in a highly awakened state, and among the data belonging to that region, detect those below the amplitude threshold set based on the amplitude that cannot occur in a highly awakened state as Slow saccade, and detect those exceeding the amplitude threshold as Slow eye movement, Set the maximum velocities of divergence and miosis that occur during a fully awakened state as thresholds, and classify the co-occurrence of divergence and miosis that simultaneously exceed these thresholds as Div-Mio, using the following technical means.

[0011] In the invention according to claim 2, in the method for measuring the arousal level according to claim 1, The eye arousal index appears in the order of Frequent saccade, Slow saccade, Div-Mio, Slow eye movement, and it is evaluated that the arousal level decreases in this order of appearance. using the following technical means.

[0012] In the invention according to claim 3, in the method for measuring the arousal level according to claim 2, As the eye arousal index, further adopt the elongation of eyelid closure duration (Elongated eyelid closure duration), and evaluate that the arousal level when this appears is equal to or lower than the arousal level when Slow saccade appears. using the following technical means.

[0013] In the invention according to claim 4, an arousal level measuring device, An eye pupil measuring device that measures eye movements and changes in pupil diameter, As an Ocular Alertness Index for evaluating the level of alertness, based on the eye movement measurement results obtained by an eye pupil measurement device, Frequent saccade, Slow saccade, and Slow eye movement are detected, and an eye alertness index detection device that detects the co-occurrence of Divergence and Miosis contrary to the near vision reflex based on the measurement results of eye movement and pupil diameter is provided. An alertness level measurement device that implements the alertness level measurement method according to claim 1 or claim 2 and determines the alertness level. The technical means as described above is used.

Effect of the Invention

[0014] As an OAI for evaluating the alertness level, Frequent saccade, Slow saccade, Div-Mio, and Slow eye movement are adopted to develop an algorithm that can be accurately detected, and the relationship between the appearance order of Frequent saccade, Slow saccade, Div-Mio, and Slow eye movement and the alertness level is clarified. Thereby, a plurality of adopted OAIs can be accurately measured, and the alertness level can be determined based on their appearance order. As an OAI, by adding EECD, the alertness level can also be determined.

Brief Description of the Drawings

[0015]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Mode for Carrying Out the Invention

[0016] The method for measuring the arousal level of the present invention is a method for evaluating the decrease level of arousal based on the order of appearance of an ocular arousal index (OAI) for evaluating arousal by measuring changes in eye movement and pupil diameter by known eye pupil measurement. The method for measuring the arousal level of the present invention will be described with reference to the drawings.

[0017] Regarding the OAI, which is an eye pupil movement that does not appear when the arousal level is high and appears when the arousal level decreases, as shown in the examples, when the order of its appearance and the arousal level were evaluated, the following contents were found. The abbreviations used in this specification are shown in Table 1.

[0018]

Table 1

[0019] As OAI, there are 1) shortening of the inter-saccade interval (ISI): Frequent SC, 2) decrease in the peak velocity of saccades (SC): Slow SC, 3) decrease in the convergence angle, 4) miosis, 5) delay in saccadic eye movement: SEM, 6) prolongation of eyelid closure, 7) blunting of blinking, 8) vestibulo-ocular reflex: VOR, 9) blunting of optokinetic response, 10) blunting of smooth pursuit eye movement, etc.

[0020] Regarding the above 3) and 4), divergence accompanied by mydriasis occurs due to changes in the viewing distance regardless of the level of arousal, and miosis may be induced due to changes in the light environment regardless of the level of arousal. Therefore, they were not adopted.

[0021] Instead, in the present invention, the co-occurrence of divergence (Div) and miosis (Mio) (Div-Mio) was evaluated.

[0022] 6) and 7) are the prolongation of the eyelid closure duration, and were evaluated together as Elongated eyelid closure duration (EECD). That is, as OAI, Frequent SC, Slow SC, Div-Mio, SEM, and EECD were adopted.

[0023] The method for measuring the arousal level of the present invention will be described with reference to the drawings.

[0024] The eye pupil movement is measured by known means. Then, based on the temporal changes in the pupil diameter and the positions of both eyes, the characteristics of the eye pupil movement that appear when the arousal level decreases, such as when sleepy, that is, OAI, can be extracted.

[0025] Frequent SC, Slow SC, and SEM are detected based on the eye movement measurement results, and Div-Mio is detected based on the measurement results of eye movement and pupil diameter.

[0026] (Frequent SC) Frequent SC is an SC that appears frequently with a shorter inter-saccade interval (ISI) than normal SC. Frequent SC is detected as an SC that is below the inter-saccade interval threshold set based on the inter-saccade interval that cannot occur in a highly awakened state among SCs. The inter-saccade interval is, for example, about 0.1 seconds. Also, the amplitude of Frequent SC is such that 56.4% is less than once, and most of them are equivalent to microsaccades. An example of Frequent SC extracted by the algorithm described later is shown in FIG. 9G.

[0027] (Slow SC and SEM) When the arousal level decreases, the angle and amplitude of saccades take values that cannot occur in a highly awakened state, and slow saccades (Slow SC) with a slow peak velocity and saccadic eye movement delays (SEM) as shown in FIGS. 9H and 10L appear. Here, "slow" means "slower" than the velocity of line-of-sight movement (normal SC) in a highly awakened state.

[0028] In normal SC in a highly awakened state, when the amplitude is small, the peak velocity is small, and when the amplitude is large, the peak velocity is large. Slow SC and SEM set a region where the peak velocity with respect to the amplitude is low and cannot occur in a highly awakened state in the main sequence diagram of saccades consisting of the amplitude and peak velocity of eye movement. Among the data belonging to that region, those below the amplitude threshold set based on the amplitude that cannot occur in a highly awakened state are detected as Slow SC, and those exceeding the amplitude threshold are detected as SEM.

[0029] Specifically, as shown in FIG. 1, in the main sequence diagram (main sequence) of saccades with the horizontal axis being the amplitude and the vertical axis being the peak velocity of the vergence angle change, based on the SC measured in a highly awakened state, the lower limit value of the peak velocity of the angle with respect to the amplitude is set as the angle threshold θ as the threshold. Also, based on the amplitude that cannot occur in a highly awakened state, the amplitude threshold TS is set.

[0030] In FIG. 1, the region above the straight line of the angle threshold θ is the region where normal SCs appear in a highly awakened state, and Slow SCs and SEMs appear in the region below the straight line of the angle threshold θ.

[0031] Among the eye movements that appear in the region below the straight line TL of the angle threshold, those with an amplitude equal to or less than the amplitude threshold TS are detected as Slow SCs, and those exceeding the amplitude threshold TS are detected as SEMs.

[0032] (Div-Mio) In a highly awakened state, when fixating the line of sight from far to near or vice versa under a certain light environment, an increase in the convergence angle (convergence) and a decrease in the pupil diameter (miosis), or divergence and mydriasis are simultaneously induced. This reflex is called the near response. Among Div-Mio, the Div part is divergence where the convergence angle becomes smaller, and the Mio part is miosis where the pupil diameter shrinks. The present inventors have discovered that in a low-awakened state, for example, in a sleepy subject, when the convergence angle decreases, the pupil contracts, which is different from the normal near response. In the present invention, this phenomenon contrary to the normal near response is adopted as Div-Mio in OAI.

[0033] (EECD) PERCLOS (Percent of Eyelid Closure) is usually defined as the ratio of the eyelid closure time including blinks in one minute, and is widely used as an OAI that characterizes eyelid closure. When this ratio exceeds 80%, it is determined that the awakening level has decreased. PERCLOS takes into account both the frequency and duration of eyelid closure including blinks.

[0034] Here, since the blink frequency varies not only depending on the awakening level but also on mental factors such as tension and anxiety, only the duration of eyelid closure (eyelid closure duration) was considered. It is known that the eyelid closure duration becomes longer as the awakening level decreases, and the duration (EECD) was evaluated for long-term eyelid closure.

[0035] Here, the maximum value of the eyelid closure duration that appeared during the fully awake state was set as the threshold. This threshold was regarded as the upper limit of the normal eyelid closure duration of each subject observed in the fully awake state. When the eyelid closure duration in the resting state exceeded the threshold, it was determined that EECD occurred due to a decrease in arousal level.

[0036] Figures 4Q and 4R show detection examples of EECD. The black portions of the pupil diameter data in Figure 4Q are the eyelid closure periods during which the pupil was not visible (obstructed by the eyelids). To obtain the eyelid closure time data, the time of each black portion was measured and compared with the set threshold. In Figure 4R, it was measured with a value exceeding the set threshold and could be detected as EECD.

[0037] (Detection algorithm for OAI) The detection algorithm for OAI will be described with reference to the figures. In the present invention, a conventional algorithm was adopted and partially improved for more appropriately detecting the currently evaluated OAI.

[0038] In Figure 2, an algorithm for detecting eye movements is shown, and in Figures 3 and 4, flowcharts of algorithms for classifying them into each OAI are shown.

[0039] When the arousal level decreases, Slow SC with a slow peak velocity and SEM with a slow drift of the line of sight appear. Here, "slow" means "slower" than the velocity of normal eye movement (normal SC). Conventional SC detection algorithms employ two thresholds, one for SC velocity and the other for SC duration.

[0040] Here, in the conventional SC detection algorithm, only normal SC with an amplitude of 0.5 degrees or more was evaluated. Therefore, as exemplified in Figure 10I, high-frequency noise in the eye movement data may be misdetected as SC. Hereinafter, "normal" indicates "a state with a high arousal level".

[0041] In the algorithm shown in FIG. 3, the aim is to separate these slow saccades (Slow SC and SEM) from normal SC without misdetecting noise as normal SC or slow eye movements. By setting an appropriate threshold for the SC speed, these slow eye movements can be separated from normal eye movements. Here, the peak speed of normal SC is relatively small for microsaccades with small amplitudes, making it difficult to distinguish from slow eye movements that appear with a decrease in arousal. Also, using only the speed threshold may cause noise to be misdetected in the process of detecting microsaccades.

[0042] In the present invention, in order to handle smaller SCs and other eye movements (amplitude 0.07 degrees or more), in the improved algorithm, first, normal SC and slow eye movements are detected together, and the slow eye movements are separated from normal eye movements by two-stage speed threshold processing and using a duration threshold and an amplitude threshold (FIG. 2). In this way, the detection is changed so as to detect normal SC, Slow SC, and SEM without misdetecting noise.

[0043] Frequent SC is mainly a horizontal eye movement, but may also be accompanied by a vertical eye movement. Therefore, in the algorithms of FIGS. 2 and 3, a threshold is set for the diagonal eye speed or amplitude (c) calculated by the following formula.

[0044] (Equation 1) c =(a 2 + b 2 ) 1 / 2 Here, a is the horizontal eye speed or amplitude, and b is the vertical eye speed or amplitude.

[0045] In the first step of the two-stage speed threshold, instead of setting the speed threshold to 5 times that of the conventional algorithm, it is set to 3 times the standard deviation (SD) of the eye speed data to detect normal SC, Slow SC, and SEM with relatively high peak speeds (FIG. 2a).

[0046] In the second step, another velocity threshold was set for 1 SD of the eye velocity data after excluding what was detected in the first step (Fig. 2b). Further, these thresholds were reset every 10 seconds and adaptively set according to changes in the eye velocity characteristics corresponding to the subject's arousal level. In this way, small normal SCs, slow SCs, and SEMs with all equal peak velocities can be reliably detected.

[0047] The start times of these eye movements are assigned just before exceeding 2 degrees / second twice before the eye velocity exceeds the threshold (3 or 1 SD of the eye velocity of each subject) (Fig. 2c). Their end times are assigned at the timing when the eye velocity first drops below 2 degrees / second after exceeding the threshold if there is no overshoot, or after performing a line-of-sight correction following the overshoot (Fig. 2c). The end time of one of these eye movements may overlap with the start time of another eye movement. If these two are in the same direction, they are regarded as a continuous single eye movement (Fig. 2d). To strictly prevent false detection, the improved algorithm of the present invention adopts a long duration threshold of 0.04 seconds or more instead of 0.006 seconds used in the conventional algorithm (Fig. 2e) and adopts an amplitude threshold of 0.07 degrees or more (Fig. 2f).

[0048] Next, the algorithm shown in Fig. 3 separates and detects normal SC, Slow SC, SEM, and Frequent SC in this order.

[0049] Conventionally, to detect SEM, an algorithm based on the thresholds of the average velocity and amplitude of SCs called the main sequence diagram has been used. In this main sequence diagram, the average velocity of individual SCs is adopted on the vertical axis, but this is not for the purpose of detecting slow SCs.

[0050] The peak velocity is known to be proportional to the SC amplitude up to 15 - 20 degrees. In the improved algorithm (Figure 3), a main sequence diagram with the peak velocity on the vertical axis instead of the average velocity is used to separate normal SC, Slow SC, and SEM.

[0051] Evaluate the angle formed by the horizontal axis and the line connecting each data point representing each eye movement detected so far in the main sequence diagram and the origin. Data points with a high peak eye velocity (vertical axis) and a small amplitude (horizontal axis) have a large angle.

[0052] Incorporate the parameters of the fully awake state data recorded in the control session (described later in the examples) for each subject. Set the minimum angle of the fully awake data as the angle threshold θ for each subject.

[0053] Apply this angle threshold to the data of the test session (described later in the examples), and as shown in Figure 1, those with an angle greater than or equal to the angle threshold θ are detected as normal SC (Figure 3a).

[0054] Among the rest, those with an amplitude threshold TS (for example, 5 deg equivalent to the conventional algorithm) greater than that are classified as SEM, and the rest are detected as Slow SC (Figure 3b).

[0055] Frequent SC is an eye movement that occurs more frequently with a shorter saccade interval (ISI) than normal SC.

[0056] For those including both those detected as normal SC in Figure 3a and Slow SC, calculate all ISIs, and those with an ISI shorter than the shortest 0.15 seconds (interval threshold) in normal adults are detected as Frequent SC (Figure 3c).

[0057] Here, as the interval threshold, the minimum ISI in a fully awake state was not used. This is because, in the examples, even in a fully awake state, the ISI was shorter than 0.15 seconds for 10 out of 14 subjects.

[0058] In the detection algorithm of EECD (Figure 4), based on the measurement result of the pupil diameter, the duration of eyelid closure is detected. The input values are the pupil diameter and the second derivative of the change in pupil diameter. A known algorithm for detecting EECD was adopted, and in the present invention, a unique threshold for accurately detecting EECD was set.

[0059] First, using the pupil data in which specific values were recorded during eye closure, the duration of each individual eyelid closure in the fully awake state in the control session (example) of each subject is detected (Figure 4a). In this algorithm, the maximum value of the eyelid closure duration that appeared during the fully awake state (for example, the test session in the example) was set as the threshold (Figure 4b). This threshold is regarded as the upper limit of the normal eyelid closure duration of each subject observed in the fully awake state, and when the eyelid closure duration in the resting state exceeds the threshold, it is determined that EECD has occurred due to a decrease in arousal level.

[0060] An algorithm for detecting Div-Mio of the present invention will be described. First, the right horizontal eye position data is subtracted from the left horizontal eye position data to obtain the convergence angle data such that a positive convergence angle represents an increase in the convergence angle. Here, the horizontal eye position data in the fixation state during eye position correction is corrected to 0 degrees.

[0061] In the algorithm for detecting Div-Mio, the maximum speeds of divergence and miosis that appeared during the fully awake state were set as the thresholds (Div and Mio thresholds) for each subject, and specific divergences and mioses that exceeded these thresholds simultaneously were classified and detected as Div-Mio.

[0062] It should be noted that the thresholds set by each algorithm may vary depending on the subject, measurement conditions, and detection purpose, and can be appropriately set according to the conditions and purposes.

[0063] (Appearance order and arousal level of OAI) As revealed in the examples, as the arousal level decreases, the above-mentioned OAI according to the arousal level Frequent SC→Slow SC→Div-Mio→SEM appears in this order. The arousal level is represented by levels 1-3, and the arousal level decreases in this order.

[0064] Level 1: Appearance of Frequent SC Level 2: Appearance of Slow SC Level 3: Appearance of Div-Mio and SEM

[0065] Thus, by detecting the appearance timing of the proposed OAI, the process of decreasing arousal level can be monitored. Also, the potential decrease in arousal level can be evaluated.

[0066] Here, EECD may appear at any time after Frequent SC. That is, when EECD appears, it can be determined that the level is 2 or higher.

[0067] (Arousal level measuring device) The arousal level measuring device includes an eyeball pupil measuring device and an eyeball arousal index detecting device. The eyeball pupil measuring device is a known device that measures eye movement and changes in pupil diameter. The eyeball arousal index detecting device is, for example, a computer connected to the eyeball pupil measuring device. Based on the eye movement measurement results, as an eyeball arousal index (OAI) for evaluating arousal level, Frequent saccade, Slow saccade, and Slow eye movement are detected, and based on the measurement results of eye movement and pupil diameter, a reflection (Div-Mio) that is contrary to the co-occurrence near reflex of divergence and miosis is detected. The arousal level is judged based on the detected OAI and output.

[0068] The arousal level measuring device may include a warning device that issues warnings such as sounds and lights according to the arousal level.

[0069] (Effects of the Embodiment) As OAIs for evaluating arousal level, Frequent SC, Slow SC, Div-Mio, and SEM were adopted to develop an algorithm that can be accurately detected. The relationship between the appearance order of Frequent SC, Slow SC, Div-Mio, and SEM and the arousal level was clarified. Thereby, the above-mentioned proposed multiple OAIs can be accurately measured, and the arousal level can be determined based on their appearance order. As an OAI, EECD can also be added to determine the arousal level.

Example

[0070] The relationship between the arousal level and the appearance order of OAIs was investigated.

[0071] The subjects of the experiment were 14 male college students aged 22 to 26. In order to prevent maintaining a high arousal level or suddenly falling asleep during the experiment, they were not allowed to eat caffeinated drinks or food one hour before the experiment.

[0072] The subjects participated in two experimental sessions: a "control session" and a "test session". The data from the control session was regarded as a completely awake state and used to detect specific eye pupil movement behaviors from the data of the test session.

[0073] The experiment was conducted in a quiet and dark room as shown in FIG. 5. The subjects lay supine in a reclining chair, and the horizontal and vertical eye positions and pupil diameters of both eyes were measured at a sampling rate of 256 Hz for about 2 hours by a head-mounted eye tracker EyeSeeCam (EyeSeeTech, Germany).

[0074] The recorded data of the pupils and eye positions of both eyes were imported into MATLAB (registered trademark) (Mathworks, USA) for offline analysis, and OAIs were detected using the algorithm of the present invention.

[0075] During the control session, the subject remained fully awake in a reclining chair in a dark room, maintaining the state by performing a mental arithmetic task. The task was to subtract 7 from 999, 1000, or 1001 randomly assigned to each subject.

[0076] The subject was instructed to remain still, blink and breathe naturally, and fixate on the surroundings of the "penlight" in Fig. 5. The session lasted for about 200 seconds, and the arousal level before the eyelids closed in the initial stage of falling asleep was evaluated. The 30-second eye pupil movement data correctly answered by each subject during the control session was regarded as that in a fully awake state.

[0077] During the test session, the subject remained in the same state and performed the same procedure as the control session except for the mental arithmetic task. The subject was instructed not to resist sleepiness. From the data measured in the test session, OAI was detected and its order of appearance was examined.

[0078] (Subjective evaluation of sleepiness) Before and after each test and control session, the subject reported subjective sleepiness using a visual analog scale (VAS). In the VAS, the words "very sleepy" and "not sleepy at all" were written at the left and right ends of a 100-mm long horizontal line, and the subject was asked to draw a vertical line on the horizontal line somewhere between the left and right ends to represent the current subjective sleepiness. The distance from "not sleepy at all" to the vertical line was quantified as the subjective sleepiness level.

[0079] Here, subjects who reported 0 - 33 mm after the test were classified into the "awake group", and subjects who reported 0 - 33 mm before the test and 34 - 100 mm after the test, that is, subjects who reported wakefulness before the test and an increase in sleepiness after the test, were classified into the "sleepiness group". Others (subjects who did not increase subjective sleepiness after the test, or subjects who already reported subjective sleepiness before the test) were classified into the "other group". For subjects classified into the sleepiness group while in an awake state, appropriate data for setting the threshold for detecting OAI might not have been obtained, so the order of appearance of OAI was not evaluated.

[0080] Table 2 shows the results of the Visual Analogue Scale (VAS) for subjective sleepiness of all subjects in the awake and resting states. Before is the report before the session (the larger the number, the stronger the sleepiness), After is the report after the session, and Change shows the change before and after the session. A positive change indicates an increase in subjective sleepiness. Among the 14 subjects, 2 reported feeling sleepy in the VAS tests conducted before and after the control session. Since the data of the control session might not fully reflect a completely awake state, they were excluded from the subsequent data analysis.

[0081] [Table 2]

[0082] For the remaining 12 subjects, appropriate thresholds were determined from the data of the control session and applied to detect OAI in the test session data. Table 3 shows the time when these 12 subjects were completely awake and answered arithmetic questions correctly.

[0083] [Table 3]

[0084] (Detection of OAI by algorithm) An example of applying an algorithm to OAI detection is shown below. Here, the threshold was set based on 30 seconds of control session data when the subject was fully awake.

[0085] Figure 6 shows the horizontal eye position (A), vertical eye position (B), convergence angle, and pupil diameter trace (C) for the entire control session of subject g. The data for 30 seconds in the dashed square portions a to f from A to C are the data when the subject was fully awake, that is, the data when the subject answered correctly by mental calculation during the control session, and are shown enlarged in D - F.

[0086] Figures 7 and 8 show the horizontal eye position (A), vertical eye position (B), convergence angle, and pupil diameter trace (C) for the entire test session of a representative subject (subject g). The dashed square portions in Figures 7A - C are enlarged in Figure 8D - F, and Figure 8D - F is further enlarged in Figures 9, 10G - L, Figure 12O, and Figure 13Q.

[0087] Table 4 shows all the thresholds set for each of the 12 subjects for detecting OAI.

[0088]

Table 4

[0089] Angle threshold: The threshold for separating Slow SC and SEM from normal SC (Figures 3a and 11N). Amplitude threshold: The threshold for separating Slow SC and SEM (Figures 3b and 11N). Saccade interval threshold: The threshold for detecting Frequent SC from Normal SC and Slow SC (Figure 3c). Div threshold: The threshold for detecting the Div part of Div - Mio (Figure 12P). Mio threshold: Threshold for detecting the Mio part of Div-Mio (Fig. 12P). EECD threshold: Threshold for detecting EECD (Figs. 4b and 13R).

[0090] The algorithm for detecting OAI (Figs. 2 - 4) whose threshold was set based on data when fully awake in the control session was applied to the data of the test session. The results of OAI detection by the algorithm are shown below.

[0091] Fig. 9G shows a detection example of Frequent SC with an ISI shorter than 0.15 seconds. Fig. 9H shows a detection example of Slow SC.

[0092] Fig. 10J shows detection examples of Slow SC and SEM, and normal SC detected by the modified algorithm. For comparison, Fig. 10I shows the detection results by the conventional algorithm. In the detection results by the modified algorithm shown in Fig. 10J, the noise components misdetected by the conventional algorithm shown in Fig. 10I were not misdetected. In all 12 subjects, the modified algorithm reduced misdetection by 96.6% compared with the conventional algorithm. Also, while only individual SCs could be detected by the conventional algorithm, (I), the modified algorithm could detect eye movements formed continuously by multiple eye movements ("one continuous eye movement" in J). Such eye movements were often specifically slow eye movements (Slow SC or SEM).

[0093] Figs. 10L and 11N show detection examples of Slow SC and SEM by the modified algorithm. For comparison, Figs. 10K and 11M show detection examples of SEM by the conventional algorithm. The conventional algorithms shown in Figs. 10K and 11M were not aimed at detecting Slow SC. The modified algorithm could detect not only SEM but also Slow SC as shown in Figs. 10L and 10N.

[0094] Figure 12O shows an example of Div-Mio detection by an algorithm. The negative convergence angle and pupil diameter indicate divergence and miosis, respectively. Figure 12P shows the first derivative waveforms of the changes in the convergence angle and pupil diameter. In Figure 12P, Div-Mio was detected at the point where the derivative waveform of the convergence angle crossed the Div threshold set to determine that the waveform had shifted closer to divergence from convergence. That is, in Figure 12, the hatched portion could be detected as Div-Mio.

[0095] Figure 13 shows an example of EECD detection by an algorithm. The black line portion of the pupil diameter data in Figure 13Q automatically detected by the algorithm (Figure 4) is the eyelid closure period during which the pupil was blocked by the eyelid and not visible. To obtain the eyelid closure time data, the time of each black line portion was measured. By comparing the measured time with a threshold, EECD could be detected by the algorithm of the present invention as shown in Figure 13R.

[0096] (Determination of the order of appearance of OAI) Using the data from each subject's test session, the relative appearance timing of each OAI was evaluated. The number of subjects in which a certain OAI appeared before another OAI was counted, and the possibility of the reverse order was also evaluated. Furthermore, a one-sided paired t-test was used to evaluate whether the difference in the first appearance timing of pairs of OAIs was significant. When there was concern about a type I error, the p-value was corrected using false discovery rate correction (Benjamini & Hochberg, 1995).

[0097] Table 5 shows the first appearance timing of each OAI for seven subjects (a, b, e, f, h, i, l) who were awake in the VAS test before the test session and sleepy in the VAS test after the test session, and two subjects (c, g) who were awake both before and after the test session. "-" indicates that each OAI was not detected. Frequent SC appeared in all nine subjects. Slow SC and SEM appeared in eight and seven subjects, respectively, and Div-Mio and EECD appeared in five subjects, respectively.

[0098] [Table 5]

[0099] Table 6 lists the timing of the appearance of each OAI (except EECD) listed in Table 5, ordered by arousal level. Of the eight subjects who were presented with both Frequent and Slow SC, Frequent SC appeared before Slow SC in six subjects, while the remaining two subjects had the reverse order. SEM and Div-Mio always appeared after Frequent SC or Slow SC. Of the five subjects who were presented with both Div-Mio and SEM, Div-Mio appeared before SEM in four subjects.

[0100] [Table 6]

[0101] The order in which these OAIs appear is typically: Frequent SC→Slow SC→Div-Mio→SEM In some subjects (2 out of 8), the order of appearance of frequent and slow SCs was reversed, but in only 1 out of 5 subjects.

[0102] Figure 14 shows an example of a raster plot showing black dots representing all occurrences of each OAI for subject 1. The circled plots indicate the first occurrence of each OAI.

[0103] OAI is the same as in Table 6 Frequent SC→Slow SC→Div-Mio→SEM EECD appeared sporadically and later than Frequent SC.

[0104] The difference in the first appearance timing between Frequent SC and Slow SC was, on average, 40.28 seconds (maximum 187.0 seconds, minimum -15.33 seconds). (t(7) = 1.63, p → 0.05, Cohen's d = 0.92). The difference in the first appearance timing between Slow SC and Div-Mio was, on average, 70.54 seconds (maximum 131.6 seconds, minimum 25.96 seconds). (t(4) = 3.86, false discovery rate corrected (p < 0.05, Cohen's d = 0.66). The difference in the first appearance timing between Div-Mio and SEM was, on average, 55.33 seconds (maximum 105.5 seconds, minimum -71.04 seconds). (t(4) = 1.72, false discovery rate corrected p → 0.05, Cohen's d = 1.63). Among the OAIs, only EECD appeared at completely different timings for each subject, but the first appearance timing of EECD was later than that of Frequent SC (the interval was, on average, 95.25 seconds, maximum 150.3 seconds, minimum 35.73 seconds, t(4) = 4.07, false discovery rate corrected p < 0.05, Cohen's d = 3.19). There was a difference of 40.28 to 95.25 seconds in the appearance timings of these OAIs.

[0105] The difference in the appearance timings of the OAIs was as large as 40.28 to 95.25 seconds, indicating that the OAIs are broad indicators of arousal.

[0106] From the above, OAI corresponds to a decrease in the arousal level Frequent SC→Slow SC→Div-Mio→SEM was confirmed to appear in this order.

Claims

1. By measuring the eye pupil to measure changes in eye movement and pupil diameter, As an ocular alertness index for evaluating arousal level, Based on the eye movement measurement results, detect shortening of the saccade interval (Frequent saccade), decrease in the peak velocity of saccades (Slow saccade), and slow oscillation of eye position (Slow eye movement), Based on the measurement results of eye movement and pupil diameter, detect the co-occurrence of divergence and miosis (Div-Mio) that is contrary to the near reflex, A method for measuring the arousal level that evaluates the level of decrease in arousal based on the order of appearance of the ocular alertness index, Frequent saccade is Detect as Frequent saccade those that are below the saccade interval threshold set based on the saccade interval that cannot occur in a highly aroused state, Slow saccade and Slow eye movement are In the main sequence diagram of saccades consisting of the amplitude and peak velocity of eye movement, set a region where the peak velocity is low with respect to the amplitude that cannot occur in a highly aroused state, and among the data belonging to that region, detect as Slow saccade those that are below the amplitude threshold set based on the amplitude that cannot occur in a highly aroused state, and detect as Slow eye movement those that exceed the amplitude threshold, Set the maximum velocities of divergence and miosis that occurred during a fully awake state as thresholds, and classify the co-occurrence of divergence and miosis that simultaneously exceed these thresholds as Div-Mio. A method for measuring the arousal level characterized by this.

2. The ocular alertness index appears in the order of Frequent saccade, Slow saccade, Div-Mio, Slow eye movement, and it is evaluated that the arousal level decreases in this order of appearance. The method for measuring the arousal level according to Claim 1.

3. As an ocular alertness index, further adopt an elongated eyelid closure duration, and evaluate that the arousal level when this appears is equal to or lower than the arousal level when Slow saccade appears. The method for measuring the arousal level according to Claim 2.

4. An eye pupil measurement device that measures changes in eye movement and pupil diameter, As an eye arousal index (OAI) for evaluating arousal level, based on the eye movement measurement results by the eye pupil measurement device, Frequent saccade, Slow saccade and Slow eye movement are detected, and based on the measurement results of eye movement and pupil diameter, an eye arousal index detection device that detects the co-occurrence of divergence and miosis that are contrary to the near reflex, and An arousal level measurement device characterized by implementing the arousal level measurement method according to claim 1 or claim 2 and determining the arousal level.

Images