QUANTIFICATION OF MOVEMENT OR LOAD ASYMMETRIES

Abstract

A computer-implemented method for quantifying a movement or loading asymmetry of a pair of extremities for computer-assisted diagnosis and / or for assessing the effectiveness of a surgery or treatment is presented. The method comprises: providing (102) a first pattern representing a movement or loading pattern of a first extremity; providing (104) a second pattern representing a movement or loading pattern of a second extremity; performing (106) a linear transformation of the data values ​​of the first and second data points to obtain scaled first and second patterns, wherein the data values ​​of the scaled first and second data points have been transformed into a common data value range; identifying (108) pairs of first and second scaled data points that represent the same point in time;Calculating (110) a normalized asymmetry score pattern comprising a plurality of normalized asymmetry scores, each asymmetry score being calculated as a function of the difference between the data value of a pair of first and second scaled data points; and outputting (112) an indication of the presence and / or magnitude of the subject's asymmetry as a function of the normalized asymmetry score patterns to enable use of the indication for computer-assisted diagnosis and / or for evaluating the effectiveness of a surgical procedure or treatment to compensate for or reduce movement asymmetries.

Description

FIELD OF THE INVENTION

[0001] The invention relates to the quantification of movement and load asymmetries for computer-assisted diagnosis and for assessing the effectiveness of an operation or treatment. BACKGROUND

[0002] Nowadays, various methods are used to quantify gait asymmetry for diagnostic and therapeutic purposes, e.g., to monitor postoperative functional recovery in clinical rehabilitation.

[0003] For example, unilateral lower limb injuries and degenerative musculoskeletal disorders often result in asymmetries in loading during movement and / or asymmetries in limb movement on the left and right sides of the body. Therefore, movement asymmetries, also known as gait asymmetries, and / or loading asymmetries are often used as metrics to assess disease status and monitor a patient's functional recovery or disease progression. Movement or loading asymmetries can also be used for computer-assisted diagnosis. Movement asymmetries, for example, can be an indicator of a neurological disorder.

[0004] There are now a number of methods for quantifying asymmetries, such as the symmetry index (SI) (Robinson, RO, Herzog, W., and Nigg, BM (1987): “Use of force platform variables to quantify the effects of chiropractic manipulation on gait symmetry”, J. Manipulative Physiol. Ther. 10, 172-176); the ratio index (RI) (Ganguli, S., Mukherji, P., and Bose, KS (1974): “Evaluation of the gait pattern of unilateral transtibial amputees fitted with patellar tendon prostheses”, J. Indian Med. Assoc. 63, 256-259); the symmetry angle (SA Zifchock, RA, Davis, I., Higginson, J., and Royer, T., 2008: “the symmetry angle: a novel, robust method of quantifying asymmetry”, Gait Posture 27, 622-627. doi: 10.1016 / j.gaitpost.2007.08.006); the Normalized Symmetry Index (Queen, R., Dickerson, L., Ranganathan, S., and Schmitt, D., 2020: “a novel method for measuring asymmetry in kinematic and kinetic variables: the normalized symmetry index”, J. Biomech. 99:109531. doi: 10.1016 / j.jbiomech.2019.109531); and the Weigthed Universal Symmetry Index (Alves SA, Ehrig RM, Raffalt PC, Bender A, Duda GN, Agres AN.: “Quantifying Asymmetry in Gait: The Weighted Universal Symmetry Index to Evaluate 3D Ground Reaction Forces”, Front Bioeng Biotechnol. 2020 Oct 23;8:579511. doi: 10.3389 / fbioe.2020.579511. PMID: 33195140; PMCID: PMC7644861).

[0005] However, significant technical limitations of these methods have been described, such as artificial inflation, which precludes the comparison of irregular signals, as is common in pathological gait, since many of the existing methods are not robust to small fluctuations in the amplitude of the measured movement or load and therefore have limited reliability.

[0006] Another challenge is the limited applicability of existing approaches with respect to the type of limb being assessed or the type of measurement devices used for data acquisition. Existing solutions are often designed as specialized solutions whose results are not comparable or integrable with those of other algorithms, measurement devices, or limbs.

[0007] The currently used approaches therefore have several disadvantages. In particular, their scope of application is often limited. It is an object to provide an improved system and method for quantifying motion or load asymmetries. The objects underlying the invention are achieved by the features of the independent claims. SUMMARY

[0008] In one aspect, the invention relates to a computer-implemented method for quantifying an asymmetry. The asymmetry is a movement asymmetry and / or a load asymmetry of a pair of limbs of a subject. The subject is a human or a non-human individual, e.g., a human or a non-human animal. The asymmetry can be quantified, for example, for a computer-assisted diagnosis and / or for assessing the effectiveness of a surgery or treatment, or for another application scenario. The method comprises: - providing a first pattern, the first pattern comprising a time series of first data points, the time series of the first data points indicating a movement pattern or a loading pattern of a first limb of the subject, the first limb being a limb or a joint on a first side of the subject's body; - providing a second pattern, the second pattern comprising a time series of second data points, the time series of the second data points indicating a movement pattern or a loading pattern of a second limb of the subject, the second limb being a limb or joint of a second side of the subject's body, the second side being opposite the first side, the second limb being of the same type as the first limb; - performing a linear transformation of the data values ​​of the first and second data points so as to obtain a scaled first pattern comprising a time series of scaled first data points and a scaled second pattern comprising a time series of scaled second data points, wherein the data values ​​of the scaled first and second data points have been transformed into a common data value range; - identifying pairs of first and second scaled data points representing the same time point within an observation time interval covered by the first and second patterns; - Calculating a normalized asymmetry score pattern comprising a plurality of normalized asymmetry scores, each asymmetry score being calculated from a respective one of the pairs of first and second data points as a function of the difference between the data value of the scaled first data point and the data value of the scaled second data point of that pair, each normalized asymmetry value indicating the magnitude of a movement and / or loading asymmetry between the first and second limbs at a time in the observation time interval represented by the pair of data points, and each normalized asymmetry value having been normalized to a predefined output value range; - Output of an indication of the presence and / or magnitude of the subject's asymmetry as a function of the normalized asymmetry score pattern to enable use of the indication for computer-assisted diagnosis and / or to enable use of the indication for evaluating the effectiveness of a surgical intervention or treatment to compensate for or reduce movement asymmetries.

[0009] The data value of a scaled data point is also called the “scaled data value”.

[0010] Calculating each asymmetry score as a function of the difference between the data values ​​of the first and second scaled data points of each pair may have the advantage of providing a mathematically robust method for quantifying side-to-side asymmetries that exhibits scale invariance and is also applicable to physical variables where the zero point (origin of the movement pattern) is not well defined.

[0011] The property of scale invariance means that changes in the scale of the input data do not lead to changes in the asymmetry result. In other words, the result remains unchanged when the units of measurement of the input data are transformed. This property is ensured in some existing approaches by calculating the asymmetry score as a function of the ratio of two signals. However, ratio-based approaches can only be applied to physical quantities for which the zero point is precisely and absolutely defined (e.g., ground reaction forces), since knowledge of the absolute origin of the motion is essential to eliminate the effect of different dimensions for a given measured physical quantity (e.g., knowledge of the "zero position" allows the conversion of Newtons to a percentage of body weight for ground reaction forces).Compared to ratio-based asymmetry score approaches, the calculation of the asymmetry score according to embodiments of the invention therefore does not require a well-defined zero point.

[0012] However, there are other physical variables where the situation is more complex, such as kinematic data (e.g., flexion-extension angle), where the zero point is defined based on the relative position between two segments, and different initial orientations of the knee joints can be used as the "zero position." Calculating the asymmetry score as a function of the difference between the data values ​​of the first and second data points allows the assessment of asymmetries even for kinematic data and other data types where the zero point is defined based on relative positions.

[0013] For example, the method can be used to assess the movement asymmetry of the knee joint angle of a person's left and right knee joints. The knee joint angle is typically defined by the long axis of the tibia relative to the long axis of the femur, with full extension defined as the reference (zero degrees) and movement in flexion as positive. The reference position is therefore achieved during calibration by a bipedal standing posture (N- or T-posture), which provides a reference orientation for the movement performed during the data acquisition phase. However, due to possible anatomical limitations, the posture may deviate from the reference position of bipedal stance and contribute to further artificial symmetry deviations. Furthermore, the reference position is only a convention, as alternative definitions exist, which also affects the symmetry calculation.The latter concept can be briefly illustrated by the following example: If the initial position of full extension is defined as 0 degrees and then the left knee moves to 10 degrees of flexion and the right knee to 20 degrees of flexion, the ratio between the left and right knees is 10 / 20=0.5. Conversely, if the initial full extension is defined as 180 degrees and the knees move into flexion with the same absolute values, the ratio between the left and right knees is 190 / 200=0.95. This demonstrates that for variables without a well-defined zero point, methods based on multiplicative invariance (functions of the ratio of two signals) are not suitable.

[0014] In contrast, the calculation of the asymmetry score according to embodiments of the invention as a function of the difference between two scaled signals allows the calculation of the difference, e.g., an absolute difference of 10 degrees, regardless of the definition of the relative reference point. Therefore, embodiments of the invention have a wide range of applications, since the calculation of an asymmetry score is not limited to physical quantities for which the zero point is universally defined (e.g., ground reaction forces), but is also applicable to physical quantities for which the zero point is not universally defined (e.g., joint angles, where the zero point is defined based on the relative position between two segments in one direction).

[0015] In some examples, the difference between the data value of the first data point and the data value of the second data point of each pair is an absolute amount of distance.

[0016] According to some examples, the difference between the data values ​​of the first and second data points of each of the pairs is calculated as a Euclidean distance between the data points in a two-dimensional coordinate system in which a first dimension represents the scaled data values ​​of the first data points and the second dimension represents the scaled data values ​​of the second data points.

[0017] According to some examples, the method comprises: selecting the common data value range into which the data values ​​are transformed as a measuring device-independent, limb type-independent, and load or movement parameter-independent value range; and / or selecting the predefined output value range into which the asymmetry values ​​are normalized as a measuring device-independent, limb type-independent, and load or movement parameter-independent value range. The method further comprises repeating the method according to any one of the preceding claims two or more times for one or more of the following: different limb types, different load or movement parameters, and different measuring device types.

[0018] Scaling the original data points to a predefined range of values ​​that is consistent across different limb types, different measurement device types, and different loading or movement parameters can have the advantage of allowing the same normalization factor and the same normalization process to be applied across many different device types, limb types, and parameters without adjusting the score calculation and without determining the variance of the original measurement data. Furthermore, the normalized asymmetry scores obtained across different limb types, parameters, and measurement devices are directly comparable.

[0019] Performing the linear transformation of the data values ​​of the first and second data points into a common data value range is also referred to herein as a "scaling step." The common data value range may be a predefined value range, e.g., the range [0 to 1]. Preferably, the common data value range is device-independent and limb-independent, meaning that the above-mentioned method can be performed for two or more different types of limbs and / or for two or more different types of measuring devices, using the same common data value range.

[0020] Performing a scaling step can have the advantage of reducing the dependence on the variance of the different measuring devices, limb types, and the type of load or movement parameter measured and investigated. A further advantage can be the possibility of defining and using a universal normalization factor capable of generating normalized asymmetry values, where "normalized" means that all asymmetry values ​​that can be calculated using the above-mentioned method lie within a predefined, known asymmetry value range. The use of a device-independent and / or a device-type-agnostic common value range during the scaling step enables the use of a device-independent and / or a device-type-agnostic, i.e., universal, normalization factor in the normalization step.This can make it possible to "abstract away" device-type-specific and / or limb-specific differences in variability (measured, for example, in standard deviations) of measurement data obtained for different device types and / or limb types and enables the comparison of asymmetry values ​​obtained for different limb types and / or using different device types. As a result of scaling the first and second data points to a predefined, device-type and / or limb-type-independent value range, the same normalization factor can be used to calculate normalized asymmetry values ​​for many different limb types and / or for data obtained with many different measurement device types.

[0021] Calculating asymmetry scores as normalized asymmetry scores can enable the comparison of asymmetry scores across different limb or device types. The applicant has determined that comparing asymmetry scores across different limb types can significantly improve the diagnostic value of the asymmetry scores, as asymmetries in one limb type often correlate with, and sometimes cause, asymmetries in other limb types. A combined asymmetry score that integrates asymmetries across different limb types so that asymmetries observed in different limbs have the same potential impact on the overall score can significantly improve the diagnosis, monitoring, or treatment of a disease or condition causing the asymmetries.

[0022] Therefore, embodiments of the invention may have the advantage of enabling the comparison of asymmetries observed in different types of limbs and / or the comparison of asymmetries based on data obtained with different types of measuring devices (e.g., different device types, different manufacturers, different firmware or firmware versions, etc.).

[0023] According to examples, the data value of each of the first data points represents an amplitude of a measured movement or load obtained for the first limb at a particular time in the observation time interval. Additionally or alternatively, the data value of each of the second data points represents an amplitude of a measured movement or load obtained for the second limb at a particular time in the observation time interval. For example, the value of a first or second data point may indicate the amplitude of the ground reaction force, the angular velocity of a rotational movement of a joint, etc.

[0024] In some examples, the first pattern consists of the time series of the first data points or a curve encompassing the series of the first data points. Additionally or alternatively, the second pattern consists of the time series of the second data points or comprises a curve encompassing the series of the second data points. For example, the first and second patterns may consist of continuous measurement curves or a series of discrete measurement values ​​collected over an observation period (time series measurements), optionally connected by a fitted curve.

[0025] Preferably, the first pattern is an average pattern obtained by averaging a plurality of patterns observed on the first limb, and the second pattern is an average pattern obtained by averaging a plurality of patterns observed on the second limb. For example, the first pattern may be an average movement pattern representing a single step of the left or right leg, obtained by averaging the observed movements of 10 steps of the left leg. The second pattern may be an average movement pattern representing a single step of the other leg, obtained by averaging the observed movements of 10 steps of the other leg.

[0026] As already mentioned, the normalization factor, which can be represented, for example, with the symbol "τ", is preferably chosen such that the asymmetry values ​​correlate approximately linearly with the absolute value of the difference between the data values ​​of the respective pairs of first and second data points. The absolute value of the normalization factor can depend on the common value range into which the data values ​​are transformed in the scaling step. For example, if the data values ​​are scaled to a common value range of [0-1], the normalization factor can be approximately 1 in some examples, e.g., 0.9 or higher. In other examples, where the data values ​​of the first and second data points, e.g.,are scaled to [0-10], the normalization factor can be another factor or function that ensures that the normalized asymmetry value grows approximately linearly with the absolute value of the difference between the data values ​​of a respective pair of data points. Thanks to the scaling step, which transforms all input values ​​into a predefined and known, common range of values, the same normalization factor can be used for different types of limbs and / or devices. Thanks to the combination of the scaling step, it is therefore usually not necessary to determine the variability of the measurement data obtained with a specific device type or with respect to a specific limb type.

[0027] However, the applicant has observed that in some cases, the scaled data values ​​may be very unevenly distributed. For example, essentially all scaled data values ​​may be concentrated in a range close to zero within the predefined common value range of [0-1.0], and there may be only a few outliers in the range 0.9-1.0. In this scenario, the following steps may enable the determination of a normalization factor capable of providing a normalized asymmetry score based on a computationally balanced distribution of the data values, and which indicates an asymmetry score that correlates linearly with the computationally balanced distribution of the data values.

[0028] According to some examples, the procedure includes: - Analysis of the distribution of the scaled values ​​of the first and second data points to determine whether a threshold for non-uniformity is exceeded; - if the determination is ‘true’, use of a normalisation factor τ in the range [0.1-0.5] to calculate the normalised asymmetry score (AS); and - if the determination is incorrect, to use a normalisation factor τ greater than 0.8, in particular a normalisation factor τ of 1 for the calculation of the normalised asymmetry score (AS) values.

[0029] As an example, analyzing the distribution and determining whether a threshold for unevenness is exceeded can be performed as follows: For each of the identified pairs of first and second data points, a preliminary asymmetry score (pAS) is calculated such that the pAS scores fall within a predefined pAS score range. For example, the pAS score range can be [-1.0;1.0], and the pAS scores can be calculated as follows: pAS=sign(x−y)∗(1−2τ2τ2+abs(x−y))

[0030] Determining whether a non-uniformity threshold is exceeded involves determining whether at least 80% of the calculated pAS values ​​are in the lowest or highest quarter of the predefined pAS value range.

[0031] The normalized asymmetry score (“normalized AS” or “AS”) can then be calculated using the following formula: AS(x,y,τ)=sign(x−y)∗(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1)

[0032] The above features can be advantageous because they allow for the determination of an accurate asymmetry score even when a highly uneven distribution of scaled measurements (corresponding to a highly uneven distribution of pAS scores) is obtained. Adjusting the normalization factor τ depending on the unevenness of the distribution of scaled data points / pAS values ​​increases the robustness of asymmetry quantification against outliers and skewed raw data.

[0033] According to some examples, the calculation of the normalized asymmetry values ​​is performed such that the normalized asymmetry values ​​increase approximately linearly with the magnitude of the difference between a respective pair of first and second data points used to calculate a respective normalized asymmetry value.

[0034] For example, an ‘approximately linear slope’ may mean that when the calculated normalised asymmetry score values ​​are plotted against the magnitude of the difference (xy) and a fit line is drawn through the normalised score values, at least 80% of the calculated normalised score values ​​deviate from the value of a corresponding point on the fit line by no more than 25% of the value on the fit line, preferably by no more than 20% or preferably 15%.

[0035] As already mentioned, the normalized asymmetry score calculated as described above may have the advantage of being independent of device type and limb type and of being universally applicable, even for movement patterns where the zero point of movement is not clearly defined and different clinicians may use different zero points at the beginning of the test.

[0036] In one example, the first and second data points are acquired using different devices, such as a first insole measuring the movements of a first foot, e.g., the left foot, and a second insole measuring the movements of the other foot. Typically, the different devices are of the same type. In another example, the first and second data points are acquired using the same device, such as a motion capture system or a force plate. For example, a motion capture system is only one global system that collects data.

[0037] The measurement data provided by a measuring device may be preprocessed by the measuring device, e.g., filtered, denoised, corrected, or otherwise modified, before being forwarded to the data processing device that performs the asymmetry calculation. It is also possible for the measuring device(s) to calculate a movement pattern or a series of pattern instances from the measured raw data. For example, joint movement patterns may be calculated by a motion capture system that has no direct physical contact with the joint and calculates the joint movement pattern indirectly using measurements taken from other joints or limbs.

[0038] In some examples, the method includes determining the affected side and the unaffected side of the person's body. For example, the method may include generating a graphical user interface (GUI) that allows a user to indicate whether the left or right side of the body is the affected side. The data values ​​and / or data points obtained for the affected side may be assigned as "x" and / or used as "first data points," and the data values ​​and / or data points obtained for the unaffected side may be assigned as "y" and / or used as "second data points."If the subject is a healthy individual or the affected side is unknown, the data values ​​and data points obtained from an extremity on the left side of the body are assigned as "x" and / or can be used as "first data points", and the data values ​​and / or data points obtained for the right side can be assigned as "y" and / or used as "second data points".

[0039] Specifying the affected and unaffected sides allows for a standardized output that reflects the direction of loading (e.g., the unaffected side is more heavily loaded than the affected side). This can increase the comparability of the asymmetry score pattern between different subjects.

[0040] The "affected" side is the side of the body where a condition, such as surgery or injury, has occurred or where a disease has manifested. The unaffected side is the side where this condition or disease has not occurred or is only observed to a lesser extent. In a healthy subject, the "unaffected" side is the dominant side, i.e., the side the subject uses when performing a particular action, such as kicking a ball, and the "affected" side is the "non-dominant side," i.e., the side the subject does not use when performing a particular action.

[0041] According to some examples, the calculation of the normalized asymmetry score pattern involves calculating each of the normalized asymmetry scores AS as a function of the formula: AS=(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1), ◯ where (xy) is the difference between the scaled data values ​​of a respective pair of first and second data points, ◯ where x is a numerical value indicating the scaled magnitude of the movement or load characterized in the one of the first data points measured at a given time within the observation time interval, ◯ where y is a numerical value indicating the scaled magnitude of the movement or load characterized in a second data point obtained at the same time within the observation time interval as the first data point, ◯ where τ is a predefined numerical value.

[0042] In some examples, the calculation of the normalized asymmetry score AS may be expressed with a variability margin of + / - 15% to account for the possibility of slight changes to the above formula. The normalized asymmetry score may therefore be a value within a range of values ​​calculated according to the following formula: AS=(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1)±15%

[0043] For example, the parameter τ can be the normalization factor. Typically, the three parameters are chosen such that the calculated normalized asymmetry value correlates approximately linearly with the magnitude of the difference xy and that the normalized asymmetry values ​​all lie within a predefined range of normalized asymmetry values, e.g., [-1.0, +1.0].

[0044] According to some examples, each of the normalized asymmetry score values ​​indicates a direction of asymmetry, and the output indication includes an indication of diagnosis or treatment success that also depends on the direction of asymmetry.

[0045] For example, if the method described here is used to assess load and weight symmetry in patients with total hip arthroplasty, negative asymmetry values ​​indicate less loading of the non-dominant or ipsilateral extremity.

[0046] In some examples, each of the normalized asymmetry scores AS is calculated as a function of the formula: AS=sign(x−y)∗(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1), ◯ where the sign of the normalized asymmetry value indicates the direction of the movement or load asymmetry, ◯ Where (xy) is the difference between the scaled data values ​​of a respective pair of first and second data points, ◯ where x is a numerical value indicating the scaled magnitude of the movement or load characterized in the one of the first data points measured at a given time within the observation time interval, ◯ where y is a numerical value indicating the scaled magnitude of the movement or load characterized in a second data point obtained at the same time within the observation time interval as the first data point, ◯ where τ is a predefined numerical value.

[0047] The normalized asymmetry score pattern includes an indication of the direction of movement asymmetry as a function of the sign of each of the multiple normalized asymmetry scores.

[0048] For example, a positive normalized asymmetry value may indicate higher values ​​on the affected side. Before calculating the asymmetry value, the affected / unaffected or dominant / non-dominant sides may be manually determined by a user or automatically determined by a software algorithm.

[0049] The above formula can be slightly modified, e.g., by adding constants or various correction factors to it. To account for the possibility of slight changes, the calculation of the normalized asymmetry score AS can be expressed with a variability margin of + / - 15%, according to: AS=sign(x−y)∗(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1)plus / minus 15% of(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1).

[0050] The applicant has found that modified formulas that use a slightly different normalization factor for a specific range of scaled values ​​still function within the permissible range of variation mentioned above.

[0051] The above-mentioned formula for calculating the normalized asymmetry value can be particularly advantageous because the sign of the asymmetry value indicates the direction of the movement or loading asymmetry. Determining the direction of the asymmetry can enable or facilitate the identification of the most affected side in cases where the affected side is not already known. Thus, the most affected side typically has lower values ​​in the input data (e.g., lower loading, more restricted movement).

[0052] According to some examples, τ is a normalization factor configured to normalize the asymmetry values ​​to a predefined normalized asymmetry value range. For example, τ may be a numerical value in the range of 0.1-1.1, in particular a number greater than 0.9, e.g., a number between 0.9-1.1, preferably 1.0.

[0053] In a preferred example, τ is equal to 1.0. For example, the predefined common range to which the data point values ​​are scaled is [0-1.0], τ is 1.0, and the asymmetry score values ​​are normalized to a permissible range of [-1.0; +1.0].

[0054] In some examples, the first limb is a limb on the left side of the body and the second limb is a limb on the right side of the body, or vice versa. This can be useful to ensure comparability of asymmetry score patterns between different healthy subjects or subjects where the affected site is not prima facie known, e.g., if a patient has had an accident that affected both sides of the body similarly.

[0055] According to other examples, the first limb is a limb affected by a movement or loading abnormality or impairment, in particular an injury, a disease, a deformity, and / or a surgical symptom, and the second limb is a limb on the opposite side of the body that is not or less affected by said abnormality or impairment, or vice versa. The method may, for example, comprise generating a graphical user interface (GUI) that allows a user, e.g., the patient or a medical expert, to input the affected side of the body, thereby determining which side of the body is considered the side with the first limb.

[0056] The output indication of the presence and / or degree of asymmetry of the subject shall be used for one or more purposes selected from a group consisting of the following: - computer-assisted diagnostics; - automatic computer-assisted diagnostics; - computer-aided surgical planning; - computer-aided design, manufacture or optimisation of products, in particular medical and non-medical aids for correcting or compensating movement asymmetries, in particular soles, insoles, shoes, joint supports, exoprostheses and walking aids; - Evaluation of the effectiveness of an operation or treatment, particularly to compensate or reduce movement asymmetries.

[0057] For example, using the indication to assess the effectiveness of a surgery or treatment may include determining the difference between pre- and post-therapy by determining the asymmetry score pattern for one or more different limbs and / or one or more predefined movement patterns at least once before initiating therapy and at least a second time during and / or after completion of therapy. Therapy may be surgery or conservative therapy such as physical therapy.

[0058] In some examples, the method is performed by a computer system, e.g., a server computer, and the first and second patterns are received from another computer, e.g., a client computer, over a network or calculated by the server from measurement data received from the client computer over the network. The step of outputting the indication comprises transmitting the indication of the presence and magnitude of the motion asymmetry over the network to the client computer that provided the measurement data or the first and / or second patterns, or to another client computer.

[0059] According to some examples, performing the linear transformation comprises scaling the sizes of the first pattern and the second pattern, wherein the scaling comprises: - Calculate a scaled data value x for each of the first data points according to : x=xori−min(DS1,DS2)max(DS1,DS2)−min(DS1,DS2), - Calculate a scaled data value y for each of the second data points according to: y=yori−min(DS1,DS2)max(DS1,DS2)−min(DS1,DS2), - where x org is the data value of the first data point before the transformation is performed; - where y ori is the data value of the second data point before the transformation is performed; - where DS 1 represents the time series of the first data points; - where DS 2 represents the time series of the second data points; - where min (DS 1 , DS 2 ) is the minimum data value that can be found in the combination of all the first data points that are in DS 1 and all second data points that are in DS 2 are included before the transformation is performed, - where max (DS 1 , DS 2 ) is the maximum data value that can be found in the combination of all the first data points contained in DS 1and all second data points that are in DS 2 are included before the transformation is performed, - where the asymmetry score for each of the pairs of first and second data points is used as a function of the difference between the scaled data value x of the first scaled data point and the scaled data value y of the second scaled data point of that pair.

[0060] This can have the advantage of bringing measured values, e.g., movement or load values, obtained from the first and second limbs onto a common, preferably limb- and / or device-type-independent scale. This allows the definition of a normalization factor τ and a normalization method capable of calculating normalized asymmetry values ​​that are comparable between different types of limbs and different types of measuring devices.

[0061] According to some examples, the procedure includes: - receiving a series of repeatedly measured instances of the first pattern, each instance of the first pattern describing the first movement or loading pattern observed for the subject's first limb during the subject's repeated performance of a particular movement or task by the first limb; - receiving a series of repeatedly measured instances of the second pattern, each instance of the second pattern describing the second movement or loading pattern observed for the subject's second limb during the subject's repeated performance of the particular movement or task by the second limb; - calculating an average first pattern from the series of instances of the first pattern, the average first pattern comprising a time series of first average data points, the data values ​​of which each indicate an amplitude mean of the movement or load observed for the first member at a particular time within the observation time interval in all measured instances of the first pattern; - calculating an average second pattern from the series of instances of the second pattern, the average second pattern comprising a time series of second average data points, each of whose data values ​​indicates an amplitude mean of the movement or load observed for the second limb at a particular time within the observation time interval in all measured instances of the second pattern; - Using the first average data points as the first data points and using the second average data points as the second data points to perform the identification of the pairs of the first and second data points, optionally also to perform the linear transformation and to calculate the normalized asymmetry score pattern.

[0062] For example, the series of repeatedly measured instances of the first pattern may be a series of 15 repeated executions of a particular movement, e.g., with the left leg. The first pattern may be, for example, "movement pattern of the left knee when performing a single step with the left leg." The series of repeatedly measured instances of the second pattern may be a series of 16 repetitions of the second pattern, "movement pattern of the right knee when performing a single step with the right leg." The first pattern used as input to the scaling step may be an average pattern calculated by averaging the 15 repetitions ("instances") of the first pattern, and the second pattern used as input to the scaling step may be an average pattern calculated by averaging the 16 repetitions (instances) of the second pattern.

[0063] Using first and second patterns calculated by averaging multiple measured patterns can increase the accuracy of the calculated asymmetry assessment. For example, patients post-surgery find it difficult to perform sitting-standing activities, so simply ignoring the first few movement cycles is not an option, as the patient may not be able to perform more than a few movement cycles in total. These patients' repeatedly performed and measured sit-stand cycles will exhibit great variability. Calculating the average signal pattern for each extremity based on multiple measured pattern instances avoids biases and distortions that can result from considering only a single movement cycle.

[0064] According to some examples, the procedure includes: - analyzing the series of repeatedly measured instances of the first pattern to determine whether a first stability criterion is met, wherein the first stability criterion indicates that a movement pattern or a loading pattern observed in each of the repeatedly performed movements or tasks of the first limb is stable; and - selectively using the instances of the first pattern that satisfy the first stability criterion to calculate the average first pattern; and / or - analyzing the series of repeatedly measured instances of the second pattern to determine whether a second stability criterion is met, wherein the second stability criterion indicates that a movement pattern or a loading pattern observed in each of the repeatedly performed movements or tasks of the second limb is stable; and - selectively using the instances of the second pattern that satisfy the second stability criterion to calculate the average second pattern.

[0065] Not using / suppressing the first few instances of the first and second patterns for calculating the asymmetry score may have the advantage of ensuring that the subject's movement causing the movement or loading pattern has stabilized before considering the measurement data. This may increase the accuracy of the asymmetry score calculation.

[0066] Determining whether a stability criterion is met may include determining whether a minimum number of repetitions have been performed and / or whether the variance of the movement pattern has fallen below a threshold. For example, when a patient stands up and begins to walk in a particular direction, the first few steps may have a higher variance and not correspond to the typical walking movement pattern: the patient's body, which was previously in a sitting position, may not yet be fully erect and thus have not yet assumed the normal walking posture. In addition, these first steps are often slower than the steps taken later and are therefore often referred to as 'acceleration steps' because the person is 'walking faster'. Eliminating the first few steps could therefore increase the quality of the metrics on which the asymmetry assessments are based.

[0067] In some examples, the method further comprises outputting feedback information indicating whether the stability criterion is met. For example, upon receiving measurement data from the measurement device, the server computer or the client computer may automatically determine whether a sufficient number of pattern instances have been obtained and / or may determine whether the observed variability of the pattern instances is below a predefined threshold. In this case, a message may be generated indicating to the user that the stability criterion is met. This may increase performance, as the user can then terminate the experiment. This may also be beneficial for the patient, as it allows them to terminate the repetition of a movement sooner, which may be painful or stressful, particularly for injured or recently surgical patients.

[0068] In some examples, the device type of the first and second measuring device is selected from a group consisting of: - Sensor / measurement devices for measuring forces, e.g., ground reaction forces (GRF), e.g., mediolateral and anteroposterior components of the GRF, wherein the device can in particular be integrated into an insole or a shoe and / or configured for measurement; this type of device can be used to measure plantar load; according to some examples, the devices can be load-sensing insoles for measuring plantar load, vertical ground reaction force, or plantar pressure; - Sensor / measuring device for measuring rotational movements, wherein the device can in particular be integrated into a wearable that can be worn around the knee and / or configured to measure knee angles; for example, IMUs (Inertial Measurement Units) can be used to measure knee flexion and extension angles - Sensor / measuring device for measuring three-dimensional movement, wherein the device can in particular be integrated into a wearable that can be worn around the ankles and / or configured to measure ankle angles; - Sensor / measuring device for measuring three-dimensional movement, wherein the device can in particular be integrated into a wearable that can be wrapped around the hips and / or configured to measure hip angles.

[0069] According to some examples, the first and second data points represent data selected from a group consisting of the following elements: - Kinetic data, in particular forces, moments or pressures measured on the subject's lower or upper limbs and joints; - Kinematic data, in particular translational or rotational data (e.g. change of an angle over time) obtained from the subject's lower or upper limbs and joints; - Spatiotemporal data, in particular the length, duration (e.g. stride length, phases of a gait cycle, etc.) or width originating from the subject's lower or upper limbs and joints; - Electromyography data, in particular the electrical activity generated by the muscles during movement.

[0070] The kinetic data may, for example, be or include one or more of the following elements: - a vertical force of the foot when walking, running, climbing stairs or jumping; - a mediolateral force of the foot when walking, running, climbing stairs or jumping; - an anteroposterior force of the foot when walking, running, climbing stairs or hopping.

[0071] In general, kinetic data refers to parameters such as force, power, etc., while motion data is kinematic data.

[0072] The kinematic data may be or include one or more of the following elements: - a relative movement of the upper and lower arm, which leads to a change in the angle of the elbow joint; - a relative movement of the upper and lower leg, which leads to a change in the knee joint angle; - a vertical movement of the upper arm relative to the torso, which leads to a change in the angle of the shoulder joint in the vertical direction; - a horizontal movement of the upper arm relative to the torso, which leads to a change in the angle of the shoulder joint in the horizontal direction; - a vertical movement of the thigh relative to the torso, which leads to a change in the angle of the hip joint in the vertical direction; - a horizontal movement of the thigh relative to the torso, which leads to a change in the angle of the hip joint in the horizontal direction.

[0073] In another aspect, the invention relates to a method for quantifying the success of a surgery or treatment to restore or reduce movement asymmetry in a person. The method comprises: - performing the method for quantifying an asymmetry of any of the examples and embodiments described herein using first and second patterns obtained from movement and / or loading data measured before or after surgery or the start of treatment to calculate and output an initial normalized asymmetry score pattern for the subject; this step may be performed to obtain reference values ​​and may be performed immediately or shortly after, but possibly also days or weeks after, the surgery or the start of treatment, depending on the type of treatment; - repeatedly performing the method for quantifying an asymmetry of any of the examples and embodiments described herein using first and second patterns obtained from movement and / or loading data measured (typically at predefined time intervals) after surgery or the start of treatment, thereby calculating and outputting a normalized asymmetry score pattern for the subject in each case; preferably, this step is performed after the acquisition of the reference data; - Analyzing changes in the calculated normalized asymmetry score patterns over time; and - Provide an indication of whether the surgery or treatment was successful, depending on the observed changes in the normalized asymmetry score patterns over time.

[0074] For example, if the analysis shows that asymmetry values ​​have decreased over time, the result may indicate that the treatment was successful. If asymmetry values ​​remain constant or increase over time, the output may be that the treatment was unsuccessful. In this way, the success of a treatment can be closely monitored, and the treatment can be quickly discontinued or modified if it fails to reduce the observed asymmetry. The effort invested in a treatment can be continued and justified if the asymmetry decreases during treatment.Because the asymmetry scores were scaled and normalized, the asymmetry score patterns of different patients are comparable, enabling a reliable assessment of the effectiveness of a new treatment approach by comparing the asymmetry score patterns of treated and untreated patient cohorts. Manufacturers of medical walking and mobility aids or providers of new physical therapy therapies can use the method described here to objectively and reproducibly demonstrate a health benefit for a patient group.

[0075] In another aspect, the invention relates to a method for quantifying the movement or loading asymmetry of two or more different types of limbs of a person. The method comprises: - performing the method for quantifying an asymmetry of one of the examples and embodiments described herein once or multiple times to quantify the movement asymmetry of a first limb type of the subject, thereby obtaining one or more first normalized asymmetry score patterns; - performing the method for quantifying an asymmetry of any of the examples and embodiments described herein one or more times to quantify the movement asymmetry of another type of limb of the subject, thereby obtaining one or more further normalized asymmetry score patterns, wherein all first and further normalized asymmetry score patterns are normalized to the same predefined output value range; - analyzing the one or more first normalized asymmetry score patterns and the one or more further normalized asymmetry score patterns to calculate a combined asymmetry score pattern as a function of the one or more first asymmetry score patterns and the one or more further asymmetry score patterns; and outputting a result of the analysis; and / or - analyzing the one or more first normalized asymmetry score patterns and the one or more further normalized asymmetry score patterns to predict whether an asymmetry observed in the first type of limb will cause or exacerbate an asymmetry of the other type of limb; and outputting a result of the analysis.

[0076] For example, a first asymmetry score pattern can be determined for the knees of a patient who has been in a car accident and has extensive leg and abdominal injuries. A second asymmetry score pattern can be determined for that patient's hips. Because the asymmetry scores have been scaled and normalized, the asymmetry score patterns of different limb types within a person are comparable, allowing for reliable determination of whether the knees are more severely affected than the hips, or vice versa, as a result of the accident.

[0077] Soft tissue injuries, i.e. injuries to muscles, tendons, nerves and connective tissue, are often not as easy to visualise using imaging techniques as bone fractures, so that in a complicated injury situation involving multiple soft tissue regions, it is often very difficult to assess the severity and focus of the current (motor) impairment. The patient himself is often unable to provide information about which of several injured body regions is most severely affected, since not all tissues contain pain receptors. However, embodiments of the method according to the invention make it possible to obtain additional important information about which body regions, and in particular soft tissues, are particularly severely affected by an injury based on the severity of movement and postural asymmetries expressed in the asymmetry score patterns and may also allow a faster and more targeted selection of a suitable therapy.Previous methods made it impossible to compare asymmetry between different limbs, as the asymmetry scores were often developed specifically for certain limb types and could not be used to quantify other limb types. And even if a method was, in principle, applicable to several different limb types, the resulting asymmetry values ​​were not comparable.

[0078] In another aspect, the invention relates to a method for quantifying the movement or loading asymmetry of two or more different types of limbs of a person. The method comprises: - performing the method for quantifying an asymmetry of any of the examples and embodiments described herein one or more times to quantify the movement asymmetry of the first and the corresponding second limbs of the subject and for a first type of movement or loading data selected from a group consisting of kinetic data, kinematic data and spatio-temporal data, thereby obtaining one or more first normalized asymmetry score patterns; - performing the method for quantifying an asymmetry of one of the examples and embodiments described herein one or more times to quantify the movement asymmetry of the first and second limbs and for a second type of movement or load data, wherein the second type of movement or load data is also selected from the group mentioned and differs from the first type of movement or load data, thereby obtaining one or more further normalized asymmetry score patterns, wherein all first and further normalized symmetry score patterns are normalized to the same predefined initial value range; - analyzing the one or more first normalized asymmetry score patterns and the one or more further normalized asymmetry score patterns to calculate a combined asymmetry score pattern for the first and second extremities as a function of the one or more first asymmetry score patterns and the one or more further asymmetry score patterns; and outputting a result of the analysis.

[0079] For example, the combined asymmetry score pattern may be a multidimensional asymmetry score pattern, where each individual asymmetry score pattern is represented by a corresponding dimension. In other examples, the combined asymmetry score pattern may be calculated by summing or averaging the many individual asymmetry score patterns.

[0080] The use of a combined asymmetry scoring scheme, which accounts for asymmetries of a limb determined by processing measurement data comprising at least two or more kinetic data, kinematic data, and spatiotemporal data, can have the advantage of enabling a better, more accurate, and more complete understanding of the nature of a limb's impairment. In particular, joints often have multiple degrees of freedom of movement, allowing both kinetic and kinematic data to be obtained from the joint or a body part that can be moved via that joint. By determining asymmetry in several different measurable aspects of that limb and providing this data in the form of a combined scoring scheme, a much better understanding of that limb is possible.Through normalization, the different individual asymmetry score patterns are also comparable with each other and all contribute equally to the overall combined asymmetry score pattern.

[0081] According to some examples, the calculated normalized asymmetry scores and / or the results calculated in the above-mentioned application scenarios, e.g., the efficacy of a treatment or surgery or the combined asymmetry score pattern or the predicted worsening, are output to a user via a user interface, e.g., a display.

[0082] Additionally or alternatively, the results and / or output are stored temporarily or permanently, e.g., in a volatile or non-volatile storage medium. The data may be stored in a computer- or human-readable data structure such as a text file, a binary file, a JSON file, an XML file, or another type of data structure.

[0083] Additionally or alternatively, the results and / or output are passed to a receiver software configured to use the normalized asymmetry results for further processing.

[0084] The receiving software may, for example, be software for planning a surgical procedure or software for creating physiotherapy plans by selecting various exercises suitable for treating the identified asymmetry and / or by selecting the intensity and / or duration of the physiotherapy. In another example, the receiving software may be software for the computer-aided design of all kinds of mobility aids, e.g., shoe soles, insoles, corsets, crutches, or prostheses. In a further embodiment, the receiving software is software for computer-aided diagnosis configured to assist a medical expert in assessing the nature and / or severity of an illness or injury and / or in selecting appropriate treatment options.In some embodiments, the receiving software is operatively coupled to an automated motion aid manufacturing facility and configured to control the manufacturing process in dependence upon the normalized asymmetry values ​​or derived results output to the receiving software.

[0085] In another aspect, the invention relates to a computer system configured to quantify an asymmetry. The asymmetry is a movement asymmetry and / or a loading asymmetry of a pair of limbs of a subject. The subject is a human or a non-human individual. The asymmetry can be quantified, for example, for computer-assisted diagnosis and / or for assessing the effectiveness of a surgery or treatment. The system comprises a network interface and a processor.

[0086] According to some examples, the computer system is operatively coupled to a display and configured to output the normalized asymmetry scores via the display, e.g., in the form of one or more plots showing the normalized asymmetry score pattern. In some embodiments, the computer system further comprises software configured to calculate a diagnosis or a probability for a diagnosis and / or for evaluating the effectiveness of a surgical procedure or treatment and output the result of that calculation, e.g., via the display.

[0087] In a further aspect, the invention relates to a system comprising the above-mentioned computer system, the display and / or the measuring devices.

[0088] In another aspect, the invention relates to a computer-readable data structure or a storage medium containing the same, the data structure comprising the normalized asymmetry score pattern or an indication of the presence or amount of asymmetry derived therefrom.

[0089] The network interface is configured to receive data characterizing movement or loading of a first and a second limb of the person, wherein the first limb is a limb or joint on the left side of the body.

[0090] The processor is configured for: - processing the received data to provide a first pattern, the first pattern comprising a time series of first data points, the time series of the first data points indicating a movement pattern or a loading pattern of a first extremity of the subject, the first extremity being an extremity or a joint of a first side of the subject's body; - processing the received data to provide a second pattern, the second pattern comprising a time series of second data points, the time series of the second data points indicative of a movement pattern or a loading pattern of a second limb of the subject, the second limb being a limb or joint of a second side of the subject's body, the second side being opposite the first side, the second limb being of the same type as the first limb; - performing a linear transformation of the data values ​​of the first and second data points to obtain a scaled first pattern comprising a time series of scaled first data points and a scaled second pattern comprising a time series of scaled second data points, wherein the data values ​​of the scaled first and second data points have been transformed into a common data value range; for example, the scaled first and second data points may have been transformed into a common data value range of, for example, [0 to 1.0]; - identifying pairs of first and second scaled data points representing the same time point within an observation time interval covered by the first and second patterns; - Calculating a normalized asymmetry score pattern comprising a plurality of normalized asymmetry scores, each asymmetry score being calculated from a respective one of the pairs of first and second data points as a function of the difference between the data value of the scaled first data point and the data value of the scaled second data point of that pair, each normalized asymmetry value indicating the magnitude of a movement and / or loading asymmetry between the first and second limbs at a time in the observation time interval represented by the pair of data points, and each normalized asymmetry value having been normalized to a predefined output value range; - Output of an indication of the presence and / or magnitude of the subject's asymmetry as a function of the normalized asymmetry score pattern to enable use of the indication for computer-assisted diagnosis and / or to enable use of the indication for evaluating the effectiveness of a surgical intervention or treatment to compensate for or reduce movement asymmetries.

[0091] A "normalized asymmetry value," as used herein, is an asymmetry value that has been normalized to a value within a predefined range of values. According to preferred embodiments, the predefined range of values ​​of the asymmetry score is limb-independent and / or device-independent, i.e., it is the same for asymmetry scores obtained for different types of limbs (which typically correspond to different movement or loading patterns) and / or for measurement data obtained from different types of measuring devices. Therefore, the range of the output asymmetry values ​​does not depend on the type of device used for the measurement or the type of limb, thus allowing the comparison of the asymmetry values ​​of different device types and / or limb types. The normalized asymmetry value can be determined using a normalization factor τ.Preferably, the normalization factor and the function used to normalize the asymmetry values ​​are chosen such that the asymmetry values ​​correlate linearly with the absolute value of the difference between the data values ​​of the respective pairs of first and second data points.

[0092] A "limb," as used here, is a member in the traditional sense of that term, or a joint. A limb, in the traditional sense, is one of the projecting paired appendages (such as wings, arms or legs, or individual fingers or toes) of an animal's body, used primarily for movement and grasping or for other purposes such as positional stabilization.

[0093] A "joint" is a point of contact between elements of an animal's skeleton and the parts that surround and support it. In biomechanics, these elements are often referred to as "segments." For simplicity, a joint is also referred to here as a form of limb.

[0094] It is understood that one or more of the embodiments and examples described herein may be combined, as long as the combined embodiments are not mutually exclusive. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Examples are described in more detail below with reference to the drawings in which: Fig. is a flowchart of a computer-implemented method for quantifying movement or loading asymmetry. Fig. is a flowchart illustrating a workflow for data acquisition and processing to quantify motion or load asymmetry. Fig. is a flowchart illustrating data quality control of movement or load data. Fig. is a flowchart that illustrates the collection of motion or load data. Fig.is a flowchart illustrating data filtering based on a constant minimum locomotor speed. Fig. is a flowchart illustrating a data acquisition process for providing first and second movement or loading patterns. Fig. is a flowchart illustrating the quantification of movement or loading asymmetry of the first and second patterns. Fig. is a collection of graphs showing asymmetry values ​​calculated using a state-of-the-art method. Fig. 9 is a collection of graphs showing asymmetry values ​​calculated according to embodiments of this invention. Fig. shows statistical parametric mapping trajectories in the two-sample t-test calculated for a previous method for calculating the asymmetry score. Fig.11 shows the trajectories of a statistical parametric mapping in the two-sample t-test calculated for a method for calculating the asymmetry score according to embodiments of the invention. Fig. shows the asymmetry scores calculated for ACLR patients and a control group using a previous method for calculating asymmetry scores. Fig. shows the asymmetry values ​​obtained for the same two groups as in Fig. were calculated using a method for calculating the asymmetry values ​​according to the embodiments of the invention. Fig. is a collection of graphs showing an asymmetry score calculated for different patient cohorts using a state-of-the-art algorithm. Fig.a collection of graphs illustrating an asymmetry score calculated for different patient cohorts using a method according to embodiments of the invention. Fig. is a diagram showing a sign function for calculating an unnormalized asymmetry score. Fig. is a heat map of the difference between a first sample data point value x and a corresponding second sample data point value y. Fig. is a graph showing the effect of the normalization factor τ on the asymmetry assessment without scaling the x and γ input values. Fig. is a graph showing the effects of the scaling step for different normalization factors τ on the asymmetry assessment. Fig.Figure 20 is a block diagram of a distributed computer system used to execute the computer-implemented asymmetry quantification method. DETAILED DESCRIPTION

[0096] In the following, similar elements are designated with the same numbers.

[0097] Fig. Figure 1 is a flowchart of a computer-implemented method for quantifying the movement or loading asymmetry of a pair of limbs of a person. The method may be implemented, for example, in a client-server computer system, as in Fig. Likewise, the method can be implemented in a single monolithic computer system or a network of computer systems, e.g., a distributed, cloud-based computer system.

[0098] The method can use and process preprocessed measurement data. For example, insoles with sensors or other types of measuring devices can continuously measure movement and / or load data while the test subject repeatedly performs a specific movement with the left and right extremities of a specific limb type, such as a foot, leg, knee, hand, or hip. Walking or running, for example, are types of movement that consist of the repeated execution of a movement pattern of the left and right legs corresponding to a "step" movement. Likewise, the subject can repeatedly stand up and sit down, rotate their shoulder, raise and lower an arm, etc.Preprocessing may also include removing outliers or noise, supplementing the measurement data with additional data such as metadata (time and location of measurement, patient information, indication of the affected limb) or movement patterns calculated from the measurement data.

[0099] In steps 102 and 104, patterns are provided that indicate a movement or loading pattern of the person's left and right limbs. The first pattern may, for example, represent the typical movement pattern of the person taking a single step with the left foot while walking. The second pattern may represent the typical movement pattern of the subject taking a single step with the right foot while walking. A "typical" pattern of a movement or loading pattern of the left and right limbs can be determined by identifying individual occurrences of this pattern in a series of recorded movements and calculating the average of these occurrences. This average pattern can be determined, for example, by averaging the movement pattern of at least 10 or even at least 20 individual steps, whereby the first few (1-4) steps can be ignored / filtered out.In some cases, it may not be possible or necessary to calculate the mean from a large number of repeatedly performed movements. In this case, the first and second patterns may represent a movement or loading pattern of the respective limb observed during a single movement of that limb by the subject.

[0100] The patterns provided in steps 102 and 104 may be time series, i.e., a series of data points comprising a data value indicating the magnitude of the load or the speed of a movement observed at a particular time. The pattern may, in some cases, be provided as a continuous curve, where the curve may be a continuously measured curve or a computationally fitted curve to a series of measurement points.

[0101] In step 106, the computer performs a linear transformation of the data values ​​of the first data points in the first pattern and the second data points of the second pattern to obtain a scaled first pattern comprising a time series of scaled first data points and a scaled second pattern comprising a time series of scaled second data points.

[0102] Next, in step 108, pairs of first and second scaled data points are identified that represent the same time point within an observation time interval covered by the first and second patterns. This step may have already been performed before the scaling step. This step represents a form of time alignment of the first and second patterns to ensure that the asymmetry score is calculated as a function of the data value difference of data points of the first and second patterns that correspond to substantially the same time point within an observation time interval. The "observation time interval" here refers to the duration of a single instance of a repeatedly performed movement of the subject, e.g., a single step during walking or a single movement cycle during standing and sitting down.

[0103] In the next step 110, the computer system calculates a normalized asymmetry score pattern comprising a plurality of normalized asymmetry scores. Each asymmetry value is calculated from one of the pairs of first and second data points as a function of the difference between the data value of the scaled first data point and the data value of the scaled second data point of that pair. The asymmetry scores are calculated such that they all fall within a predefined output value range and are therefore referred to as "normalized" asymmetry score values. Preferably, the predefined output value range of the asymmetry score is the same for movement or loading patterns obtained from different types of measuring devices and for different types of limbs to ensure that the asymmetry scores of different limbs and device types are comparable.

[0104] In step 112, an indication of the presence and / or extent of the subject's asymmetry is then output. The indication is calculated depending on the normalized asymmetry score pattern. Preferably, the indication and / or the result of the asymmetry analyses are visually presented on a display. The visual representation of the results provides the user with feedback about their current locomotion and asymmetry through visual representation, referred to herein as "biofeedback." Providing biofeedback through this approach can contribute to improving recovery or walking abilities and thus support the psychophysiological process of (re-)gaining control over one's own locomotion.

[0105] The output may be used for many different purposes, such as enabling the use of the indication for computer-assisted diagnosis 114, assessing the effectiveness of a surgical procedure or treatment to correct or reduce movement asymmetries 116, or for other purposes.

[0106] Additionally, the output data can be stored in a database. In some examples, the result of the asymmetry analysis is stored in a central database containing the results calculated for a variety of different subjects and limb types, based on data provided by a variety of different measurement devices. The way the asymmetry values ​​are calculated can ensure that the values ​​are comparable across many different patients, time points, limb types, movement patterns, and device types.

[0107] Fig. is a flowchart illustrating a data acquisition and processing workflow for quantifying motion or load asymmetry. According to the example shown, data quality control and / or calibration of the measuring device is performed in step 202. An example of this step is described with reference to Fig. described in more detail.

[0108] Next, in step 204, a bilateral data measurement is performed. A sequence of repeatedly executed instances of a first movement pattern from a first extremity of a first side of a person's body and another sequence of repeatedly executed instances of a second movement pattern from a corresponding second extremity of the opposite side of the person's body are obtained. This step is carried out, for example, using Fig. 4 described in more detail.

[0109] Next, in step 206, the sequence of pattern instances is analyzed to determine whether a constant movement speed is achieved. Pattern instances whose speed is lower than a minimum movement speed or which do not meet other quality criteria can be filtered out, i.e., they are not considered in the subsequent processing steps. This step is carried out, for example, using Fig.5. Step 206 may be performed after completion of step 204 or during the execution of step 204 and the receipt of measurement data. In some examples, for example, in a first step, a sequence of repeatedly executed and measured first and second movement or loading patterns may be acquired and then analyzed to filter out pattern instances that do not meet the quality criterion, e.g., pattern instances that were executed at too low a speed. According to other examples, the sequence of repeatedly executed and measured first and second movement or loading patterns is analyzed immediately after they are acquired to determine whether they meet the quality criterion (e.g., minimum speed), and they are filtered out / not saved and not used for further processing if the criterion is not met.

[0110] In the next step 208, the (possibly filtered) instances of the first and second patterns are analyzed to identify the start and end points of each pattern instance in the first and second sequence of pattern instances (see, for example, the description of a Fig. Next, in step 210, the sequence of repeatedly executed movement or loading instances, as originally measured or provided as a result of the quality filtering step 208, is used to identify and quantify movement or loading asymmetries between the first and second extremities. This step is described, for example, in the Fig. 1 and Fig. 6 described in more detail.

[0111] Finally, the result 212 is output to a user or a software application. The output may, for example, include a chart with one or more asymmetry score patterns, such as in Fig.and / or contain a suggestion for a possible diagnosis of a disease or an indication of the effectiveness of a surgical procedure or other form of treatment or physiological therapy.

[0112] Not all of steps 202-212 are mandatory. For example, in some situations, the calibration and data quality control step 202, filtering out sample instances that do not meet a quality criterion, may not be performed in step 206. Each of steps 202-212 may be performed on the same or different data processing devices. For example, the data quality control, data acquisition, and filtering 206 may be performed on a client computer operatively coupled to the measuring device, while the asymmetry quantification 210 and output of the results may be performed by a server computer, as in Fig.However, other system architectures are also possible, such as using a single local or remote computer to perform some or all of the Fig. tasks presented.

[0113] Fig. is a flowchart illustrating the data quality control of movement or load data. Fig. The method presented was developed primarily for insole devices. However, the method can also be applied to other devices and other types of limbs and movement patterns. Not all Fig. The steps shown are mandatory.

[0114] Depending on the type of asymmetry to be detected, different types of sensor hardware may be used. According to some examples, the initial measurement data 220 is collected and analyzed 222 to calculate whether an offset is detected to ensure that the hardware device (e.g., a load-sensing surface such as insoles or force plates) is properly calibrated (e.g., with a measurement accuracy of 0.5 kg or better). During calibration, the device settings may be adjusted so that the initial measurement offset is zero (i.e., so that the measured load or movement amplitude is zero in a situation where no load is applied and no movement is performed). The calibration step can help avoid misleading asymmetry results resulting from inaccurate calibration and offset values. In some examples, the calibration may be verified 224.The initial motion capture 220 and / or verification of the calibration of the measuring device may consist of asking a person (e.g., the person being examined) to perform a specific movement or task. For example, the person may be asked to offload each foot with a few steps (with assistance if necessary) and then stand upright without relying on external support. The walking steps allow offset detection during the offloading of each foot. The calibration steps may be performed with the aid of external support, such as crutches, which is particularly important for injured individuals. Load or motion data measured while the person is stationary provides feedback on the overall calibration of the measuring device.

[0115] If the device does not allow a new calibration or if individual physical limitations exist, the acquired information (offset and / or calibration result) can be automatically or manually saved 226 and forwarded, for example, via a network connection, to a software or software component that performs the asymmetry analysis. This can enable the analysis software to use the offset information to perform a quality control analysis of the output 226 and / or to calibrate the received load or motion measurement data before further processing to quantify the asymmetry.

[0116] Fig.is a flowchart illustrating the collection of movement or load data, such as bipedal locomotor activity. For example, a person may perform a series of repetitive movements with their left and right legs, such as a minimum number of repetitions of a particular movement (e.g., 4 steps per side or more).

[0117] If bipedal movement activity is to be measured and analyzed, only bilateral signals are considered, i.e. the person must be able to perform bipedal movement activities. Therefore, two separate signals (one from each side) must be generated independently and recorded by one (or more) load-dependent measuring device(s). The two separate signals could, for example, be the vertical ground reaction forces recorded by a person's right and left extremities during a forward movement. The external load measuring device is used to record the bilateral signals. The measured movement or load data can be stored either online or locally, e.g. in the form of one or more data files or a data stream. Preferably, a minimum of repetitions (e.g.4 steps per side or more) and the corresponding measurement data are collected and analyzed to ensure a reliable result, preferably filtering out one or more initial steps and not using them for further analysis, as they represent acceleration steps.

[0118] According to the Fig.In the example shown, the measurement of bipedal locomotor activity is started in step 232, and the measurement data is continuously analyzed to determine when a predefined number of initial pattern repetitions (instances) was observed. The predefined number of initial pattern instances is filtered out in step 234 or not used for further processing because these instances are influenced by movement or loading asymmetries resulting from the acceleration process. In step 236, the pattern iterations following the predefined number of initial pattern instances are then selected and used for further processing. Excluding the initial pattern iterations from further analysis can increase the accuracy of the asymmetry quantification.An example of a stability criterion used to increase the accuracy of asymmetry quantification is the fact that a measured pattern instance was obtained after the predefined number of initial pattern instances were observed. The stability criterion can also be referred to as a "quality criterion" because using this criterion during data acquisition or preprocessing can increase the accuracy and quality of asymmetry assessments.

[0119] Fig.is a flowchart illustrating data filtering based on a minimum constant locomotion speed. Including iterations of only constant locomotion patterns can be an alternative variant of a stability criterion used to filter out some patterns to ensure high quality asymmetry analyses. To reach this stage of steady locomotion, a minimum constant locomotion speed must be achieved. This stability criterion can also be used in combination with the stability criterion "predefined first pattern instances were observed." After movement initiation, the measured pattern iterations are checked for their respective stability or variability.Parameters such as stride length, stride time, speed, contact area or other relevant parameters can be calculated and used to support identification when constant locomotion activity has been detected.

[0120] To determine whether a constant movement speed has been achieved, one or more parameters can be selected in step 240 to perform the determination. Depending on the hardware used, the input signals are conditioned and post-processed for use. For example, if the hardware only provides vertical ground reaction forces, the "stride time" parameter can be analyzed for stability / variance. Depending on the type of hardware used, "stance time" (e.g., ground contact time) can be used as a substitute for "speed." If "stride length" and "cadence" are available, speed can be calculated by multiplying "stride length" (m) and "cadence" (steps / minute), divided by 60.

[0121] A constant level is reached when it is determined in step 242 that the fluctuation of the selected parameter is below a certain threshold (e.g., 10%). This point in time is automatically detected, and the individual motion analyses at constant speed are started.

[0122] Once a constant speed is reached, the subsequently measured patterns are saved or forwarded for further analysis. The hardware begins acquiring the input load data measured by the hardware system (one signal for each limb), which is mathematically processed to calculate the asymmetry. Alternatively, the measurement data may have already been acquired, and only the sufficiently stable (non-filtered) pattern instances are further processed.

[0123] In some examples, if a constant speed cannot be determined in the initial steps, the next step is performed based on the acquired measurement data without filtering. This option is intended for cases where movement data is being collected from acutely injured populations who cannot achieve a constant, uniform movement speed. A warning message may be displayed in step 246 indicating that the constant speed has not been achieved.

[0124] In cases where a constant speed is unlikely to be achieved, the assessment of whether and when the measured pattern instances meet a stability criterion can be omitted entirely.

[0125] Fig.is a flowchart illustrating a data acquisition process for providing first and second movement or loading patterns. The test subject can repeatedly perform a specific movement of their first and second extremities. The movement or loading measurement data obtained for the first limb is referred to as the first measurement data, and the movement or loading measurement data obtained for the second limb is referred to as the second measurement data.

[0126] The Fig.The process steps described can, for example, be performed by a local data processing device connected to the measuring device that originally measured and recorded the load or movement data. Alternatively, this process can also be performed on a remote computer system, such as a server computer system. The local computer system can be interoperable with the server computer system and act as a client, configured to send measurement data (e.g., filtered and processed measurement data) to the server and to receive the results of the asymmetry analyses performed by the server.

[0127] In step 250, the data processing system may obtain information about the calibration and offset determined at the beginning of or before the start of the measurements of the motion or load data. If an offset is present in the measuring device used to measure instances of the first pattern, it is removed from the measured sequence of repeatedly observed instances of the first pattern. If an offset is present in the measuring device used to measure instances of the second pattern, it is removed from the measured sequence of repeatedly observed instances of the second pattern. In some cases, this step includes performing further calibrations based on characteristics of the person (e.g., body weight or height).

[0128] In the next step 252, the method includes analyzing the first and second measurement data to determine the start and end times of each repeated movement of the first and second limbs, respectively. For example, the start time of each repeated movement of a foot may represent the time at which the foot hits the ground, and the end time of each repeated movement may represent the time of toe-off, i.e., the time at which the foot leaves the ground. Both the start time and the end time are defined as the time at which a threshold of 10% of the person's body weight is exceeded. When asymmetries of other limbs, e.g., a hand or an arm, are investigated, the repeatedly executed movement may be different, and the start and end times of each repeated movement instance may be defined differently.

[0129] Next, in step 254, a first initially identified sequence of repeatedly executed movements (also referred to as an initially identified sequence of first movement or loading pattern instances) is identified by analyzing the start times and end times identified in the first measurement data. A second initially identified sequence of repeatedly executed movements (also referred to as an initially identified sequence of second movement or loading pattern instances) is identified by analyzing the start times and end times identified in the second measurement data.

[0130] In step 256, the first initially identified sequence of repeatedly executed movements is then used as input to a dynamic time-warping algorithm. Dynamic time warping (DTW) is an algorithm for measuring the similarity between two temporal sequences of data points (e.g., measurements) that may differ in speed. For example, when a person takes several steps, accelerations and decelerations occur over the course of the observation, and no single step has exactly the same duration or speed. Nevertheless, DTW can be used to identify a repeatedly executed pattern or motif that represents the movement or load profile of a "typical" walking step.

[0131] DTW can be used to calculate an optimal match between multiple pattern instances, where each pattern instance can take the form of a time series of measured values. The DTW matching algorithm can, for example, be based on at least the following rules: Each index (measurement data point) from the first pattern instance (time series of measurement data points constituting a first pattern instance) must match one or more indices (measurement data points) from the other pattern instance (time series of measurement data points constituting another instance of the said pattern), and vice versa. A "pattern" can be a repeatedly executed movement. The first index of the first sequence must match the first index of the other sequence (but it does not have to be the only match). The last index of the first sequence must match the last index of the other sequence (but it does not have to be the only match).The optimal match is defined as the one that satisfies all constraints and rules and incurs the lowest cost, where the cost is calculated as the sum of the absolute differences between the values ​​of each matching index pair. The sequences are nonlinearly "warped" in the temporal dimension to determine a measure of their similarity independent of certain nonlinear variations in the temporal dimension. In this way, all movement or loading patterns observed while the first or second limb repeatedly performs a given movement or task are matched.

[0132] According to some examples, other algorithms may be used instead of the DTW algorithm to align the plurality of first pattern instances (or to align the plurality of second pattern instances). For example, some forms of adapted Needleman-Wunsch algorithms may also be used to perform the pattern instance alignments.

[0133] After the plurality of first pattern instances have been aligned with each other and after the plurality of second pattern instances have been aligned with each other, the aligned first and second pattern instances are used as a basis for performing an asymmetry in step 210.

[0134] Fig. is a flowchart illustrating the quantification of movement or loading asymmetry of the first and second patterns.

[0135] In step 260, the aligned first pattern instances are processed to determine the typical movement performed by the first limb when repeatedly performing a particular movement. The typical movement may, for example, be the average movement or loading pattern calculated as the average of all aligned first pattern instances. Furthermore, the aligned second pattern instances are processed to determine the typical movement performed by the second limb when repeatedly performing a particular movement.

[0136] The typical movement of the second limb may, for example, be the average movement or loading pattern calculated as the average of all aligned second pattern instances. Each first (or second) pattern instance may comprise or consist of a time series of measured movement or loading amplitude values. The average pattern of the first (or second) aligned first (or second) pattern instances may be obtained by calculating, for each of a plurality of time points within the observation time interval, the average value of all data values ​​in the plurality of aligned pattern instances measured at that time point. In this way, an average first pattern and an average second pattern are obtained.The average first pattern comprises a time series of data points, each representing the average amount of movement or loading observed for the first limb at a particular time point in the observation time interval. The average second pattern comprises a time series of data points, each representing the average amount of movement or loading observed for the second limb at a corresponding time point in the observation time interval. Typically, the observation time interval obtained by aligning the first pattern instances to each other is approximately identical or similar to the observation time interval obtained by aligning the second pattern instances.

[0137] Next, in step 262, the average first pattern and the average second pattern determined in step 260 are scaled to a common scale, e.g., a numerical value range between [0, 1, 0]. This step may be similar to step 106 in Fig. For the scaling process, the minimum value (xy min ) from both the averaged data points of the first pattern (each denoted as x) and the averaged data points of the second pattern (each denoted as y) and the corresponding maximum value (xy max ) are first determined. The scaled average data points x scal , y scal , which can then be used to calculate the normalized asymmetry values ​​for the respective data point pairs, are then given by: x scal = (x - xy min ) / (xy max - xy min ), y scal = (y - xy min ) / (xy max xy min). By this linear transformation, x scal and y scal limited to values ​​within a predefined range, e.g. between 0 and 1, resulting in (x - y) difference values ​​in the range from -1 to 1 when calculating the asymmetry score.

[0138] The scaling of the input data as shown allows the requirements for multiplicative invariance to be met, since a scalar multiplication (e.g., to convert degrees to radians) with the data points of the first and second pattern does not change the scaled values ​​of the data points of both patterns, x scal and y scal .

[0139] In step 264, a normalized asymmetry score is then calculated for each pair of aligned data points of the average scaled first pattern and the average scaled second pattern. In this way, a normalized asymmetry score pattern is generated that covers the observation time. A normalization factor τ can be used to perform the normalization. If the scaling step involves transforming the data values ​​to a range of values ​​between 0 and 1.0, the normalization factor can typically be 1.0. In some cases where the scaled data values ​​are very unevenly distributed, the normalization factor can have a different value. In this way, the artificial inflation often found in other methods can be mitigated without filtering out true features of the symmetry results.Since the normalized asymmetry value is calculated as a function of the difference between the data values ​​x and y of the respective data points in the first and second pattern, the value shown in . Fig. The distance map shown has a symmetry of x - y with respect to the diagonal.

[0140] In step 264, a normalized asymmetry value pattern is calculated, i.e., all possible asymmetry values ​​are calculated to lie within a predefined output value range, e.g., [-1, 1]. This allows the comparison of the calculated asymmetry between different subjects, devices, and limb types. In this step, a normalized asymmetry value is calculated for each pair of scaled first (x scal ) and second (y scal) data points representing the same time point in the observation period. The normalized asymmetry score AS using scaled input data can be determined as given in the following equation: AS(x,y,τ)=sign(x−y)∗(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1)

[0141] Where (x - y) is the difference between the scaled values ​​of the first and second data points of a pair, and τ is a predefined numerical value. According to one example, τ is the normalization factor and has a value between 0 and 1, where x is the scaled amplitude of a measured load or movement measured at a specific time for the first limb, and y is the scaled amplitude of a measured load or movement measured at the same time (within an observation time interval, i.e., the time interval covered by the first and second samples, e.g., the duration of execution of a single instance of a repeatedly performed movement).

[0142] The Fig. show the effect of different values ​​of the normalization factor on the normalized asymmetry value, and a comparison of the diagrams in the Fig. shows the effect of the scaling step.

[0143] Fig. shows a series of graphs with asymmetry values ​​calculated using a state-of-the-art method (wUSI, described in Alves et al. 2020). The series of graphs in the upper half of Fig.shows the original data in the form of measured forces (in % of body weight). The original data were measured while a subject performed several steps with the left and right foot under two different conditions (baseline and test). The average movement pattern during a single step with the left foot was calculated (black line). Likewise, the average movement pattern during a single step taken with the right foot was calculated (gray line). The baseline condition was normal walking, and the test condition was walking with crutch support. The forces were measured using insoles with ground reaction force (GRF) sensors for the three dimensions x, y, and z. The graph "Fx (Baseline)" therefore represents the ground reaction force in the x dimension measured during normal walking (without crutches).The "Fz (Test)" graph represents the ground reaction force in the z dimension measured during walking with crutches. Each graph covers an observation time of 100 time units, with the observation time including the time required to perform a single stance phase.

[0144] Below each graph containing the original measurement data, an asymmetry score pattern was calculated using a state-of-the-art algorithm (wUSI, Alves et al. 2020). The wUSI asymmetry score graphs have one axis representing the observation time (100% stance phase) and another axis indicating the wUSI asymmetry score in standard deviations. These graphs demonstrate that although wUSI is able to reduce artificial inflation, it can still be observed to some extent, especially when the amplitude of the force signal is zero or close to zero. Furthermore, as mentioned above, wUSI is limited to limbs and movement patterns where the zero position—that is, the position of a limb that serves as the starting point and reference when observing a relative movement of a limb—is generally defined.

[0145] Fig.shows a series of graphs showing asymmetry values ​​calculated according to embodiments of the claimed invention. The asymmetry values ​​were calculated based on the same raw data used in the graphs in Fig. However, before calculating the asymmetry values, the forces measured for the left and right foot for the different conditions and the different force directions x, y, and z were scaled to a predefined, common value range (here: 0-1.0). The scaled forces were used as input for calculating the normalized asymmetry score patterns according to the following formula: AS(x,y,τ)=sign(x−y)∗(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1) where the normalization factor τ = 1 was used.

[0146] Contrary to what was observed when using the wUSI algorithm for asymmetry calculation, the asymmetry score patterns calculated according to embodiments of the invention do not exhibit artificial inflation when the input signals are zero or close to zero. The problem of artificial inflation has thus been significantly reduced. This is very advantageous because artificial inflation can (at the individual level) mask existing asymmetries or (at the group level) yield asymmetries that are not present. This is due to the dominance of such individual artificial inflations when aggregated across the group. For this reason, a methodology that eliminates this effect is highly advantageous at both the individual and group levels.

[0147] The Fig.illustrate the comparability of the asymmetry values ​​determined for the three GRF components x, y and z using two different methods ( Fig. : a state-of-the-art algorithm; Fig. : the algorithm according to the embodiments of the invention).

[0148] Fig. shows the SPM results for condition comparisons using the original data as input for the wUSI method, as performed in the original publication (Alves et al., 2020, using σ = 0.5). The term "SPM" refers to the statistical parametric mapping trajectories in the two-sample t-test calculated for the respective algorithm for calculating the asymmetry score and for the respective GRF component. The horizontal dashed lines in the SPM plots indicate the critical thresholds (z-star values) for significance.

[0149] Fig.shows the corresponding SPM results for condition comparisons (baseline vs. test) using the scaled data as input for calculating the asymmetry scores according to one embodiment of the invention. For all GRF components, the absolute z-star value, the significant start and end points of the clusters, and the respective p-value for the wUSI method (original data, σ = 0.5) and the inventive algorithm for calculating the asymmetry score (scaled data, normalization factor = 1) are shown in the following table: method component z-star Area of ​​significance (% stance phase) p-value wUSI (state-of-the-art algorithm, original input data) Fx 3.8 33-42 p<0.001 Year 3.8 9-18 63-84 p=0.002 p<0.001 Fz 3.4 0-39 54-86 p<0.001 p<0.001 Normalized asymmetry assessment with the new algorithm (scaled input data) Fx 3.6 32-39 p=0.001 Year 3.6 8-19 63-86 p=0.002 p<0.0008 Fz 3.3 1-41 53-87 p<0.001 p<0.001

[0150] The SPM results for the three components of ground reaction force were consistent for both methods, as shown in the Fig.and the above-mentioned table. This shows that the inventive method can also be used with variables typically used as input in multiplicative symmetry methods. An important advantage of the new method is that its applicability is much broader, that calculation of is not required (as in the wUSI method), and the effect of artificial inflation is completely eliminated.

[0151] The Fig. illustrate the comparability of the asymmetry scores determined for a group of patients with an anterior cruciate ligament rupture using two different methods ( Fig. : a state-of-the-art algorithm; Fig. : the algorithm according to the embodiments of the invention).

[0152] Like the Fig.As can be seen, the new method has also been successfully applied to other datasets representing different limb types and movement patterns. For example, the new approach was successfully applied to a group of patients with an anterior cruciate ligament rupture (ACLR patients) and to healthy individuals (multiple time points) during walking (knee angles in the frontal, sagittal, and transverse planes of motion), and kinematic data were collected.

[0153] The asymmetry score patterns were determined and compared for the two groups. The data were collected at three different time points (A: baseline, corresponding to approximately 6 months postoperatively, B: 4 months postoperatively, C: 12 months postoperatively). The asymmetry score patterns of these groups at each time point were compared for each of the three components of the knee angle data.

[0154] Fig.shows a graph in which the asymmetry scores for the two groups at the three time points were calculated using the wUSI approach mentioned above (σ = 0.5). To perform the calculation, a corresponding value of σ was determined, which indicates the standard deviation of the measured data obtained by the measuring device with respect to a specific movement direction.

[0155] Fig. is a graph showing SPM results for group comparisons of the two groups using the original data as input. Fig. The representation shown uses the same input data as the representation in Fig. , but for Fig. The asymmetry score is calculated according to one embodiment of the invention. The mean and standard deviation of the waveforms in the sagittal plane of the ACLR patients for the three visits / time points A, B, and C are calculated and plotted in the graph (τ = 1).

[0156] Fig. illustrates that the approach to quantifying asymmetry according to embodiments of the invention is capable of identifying statistically significant differences between groups: It was observed that at time point A, which ranges from 6% to 12% of the gait cycle, the new method effectively mitigates the confounding influence of artificial inflation (inherent in the wUSI approach), thus facilitating the discrimination of group differences, in particular between the control group and ACLR patients.

[0157] It was found that the new approach can be implemented with significantly less effort than, for example, the previous wUSI approach, since the wUSI approach might require a value of σ for each of the three dimensions, which indicates the standard deviation of the measurement data determined with the measuring device with respect to a specific direction of movement, and a direct comparison of the wUSI values ​​determined for the different dimensions is usually not possible.

[0158] Fig. shows asymmetry score patterns of different patient cohorts using a state-of-the-art asymmetry score calculation method. The series of graphs at the top of Fig.shows vertical plantar loading patterns observed in a group of patients with total hip arthroplasty (THA group) and in healthy individuals during walking. The tests were performed multiple times (V1: 1 week before THA; V2: 1 week after THA; V3: 3-6 weeks after THA; V4: 6-12 weeks after THA). For the data from each of these plots, an asymmetry score plot was calculated using a state-of-the-art algorithm (Alves et al., 2020), with the σ value used being defined based on the minimum standard deviation of the waveforms, in this case = 0.9382). The dimension representing the wUSI in these plots is expressed in standard deviations. Fig. For the type of data presented, artificial inflation is not a major problem because the input signals are not close to or crossing the zero axis.

[0159] Fig.illustrates the scaling of the Fig. original data and shows the normalized asymmetry score patterns calculated for each of the five data plots according to an embodiment of the invention, using a normalization factor of 1. As can be seen from a comparison of the Fig. As can be seen, the performance of the new methodology based on scaled data is similar to that of the wUSI method when artificial inflation is not an issue, but the new method has wider applicability and improved robustness to artificial inflation and does not require the determination of the standard deviation of the variability of the original data, as is the case when using the wUSI approach.

[0160] Post-hoc tests revealed several significant differences between the THA groups and the healthy control groups at all visits. Significant group differences for wUSI using the original data ( Fig. ) were used for V1, V2 and V3 as well as for the new method ( Fig. ) using scaled data.

[0161] Fig. is a diagram that shows a function for calculating a not yet standardized asymmetry score ( ASnot_norm ) using the sign function. According to the diagram shown, ASnot_norm as the sign of the difference between x and y: ASnot_norm=sign (xy), where x represents the data value of a scaled first data point, γ represents the data value of a scaled second data point that represents the same time point within the aligned first and second patterns as the first scaled data point, and sign is the sign function. ASnot_norm, as calculated according to the illustration, can only distinguish between positive and negative values ​​(-1 or 1), as in Fig. can be seen. In order to enable symmetry output values ​​between the predefined output value range for the normalized asymmetry scores, e.g., -1 to 1, a more complex function for calculating the normalized asymmetry scores is used according to preferred embodiments of the invention.

[0162] Fig. is a heat map of the difference between a first sample data point value (denoted as "x") and a corresponding second sample data point value (denoted as "y"). "Corresponding" means that the two data points occur at the same time within the range determined by the first sample and the second sample. were observed within the covered observation period. The heatmap shows that the difference is diagonally symmetric. Calculating the asymmetry score as a function of the difference between pairs of corresponding scaled data point values ​​of the first pattern (x) and scaled data point values ​​of the second pattern (y) can have the advantage that the calculation of the asymmetry score is based on a diagonally symmetric function of x and y rather than a radially symmetric function, in contrast to state-of-the-art approaches that rely on multiplying x and y for score calculation. The advantage of this approach is that it is more robust to small variations of very small x and γ values, i.e., artificial inflation. Artificial inflation is prediction / extrapolation error that occurs when the data values ​​used as the basis for prediction or extrapolation are very small compared to the magnitude of the measurement errors.A further advantage may be that a normalization factor τ can be defined based on the standard deviation of the values ​​x and y. This also allows for mitigating artificial inflation without filtering out true features of the symmetry results.

[0163] Fig. is a representation of a more complex, difference-based function for calculating a normalized asymmetry score based on different normalization factors τ. The function used to calculate the Fig. The data used in the normalized asymmetry scores presented are original data, i.e. data that have not been scaled. The formula for calculating the Fig. The normalized asymmetry score values ​​AS shown are: AS(x,y,τ)=sign(x−y)∗(1−2τ2τ2+abs(x−y)),

[0164] Here, xy is the difference of the data values ​​between a respective pair of a first and a second data point and τ is a normalization factor.

[0165] As from Fig. As can be seen, the above-mentioned, comparatively complex formula for calculating the standardized asymmetry score provides a wider score range than, for example, the sign function from Fig.However, if the absolute differences of x - y are large (e.g., 10), a large value for the normalization factor τ is required to adequately differentiate the AS values. If the absolute differences of the values ​​for x - y are small, a small value of τ is required. This can be problematic because the absolute differences of x and y are not known in advance and therefore there is a risk of choosing a normalization factor that is unable to sufficiently distribute the score value over the entire available output value range. Second, the entire range of possible output score values, e.g., [-1 1], is only fully exploited at infinity. These disadvantages become more apparent when comparisons of symmetry values ​​from different variables (e.g., knee abduction angle and hip flexion angle) are sought.To calculate AS, a normalization factor τ would have to be estimated for each variable, potentially leading to different asymmetry score outputs and complicating direct comparisons. Furthermore, determining different normalization factors based on an estimate of the variable's variance is time-consuming and often requires performing additional tests similar to those required to determine the standard deviation needed to calculate the state-of-the-art wUSI asymmetry score. To overcome these problems, the input signals (x and y) are scaled to a common predefined range, e.g., [0 1], and used as input for calculating the normalized asymmetry scores.

[0166] For example, in a situation where no scaling is applied, the normalization factor τ may be chosen to represent the variance, e.g., the standard deviation (sd), observed in the movement or load data points measured by a particular type of measuring device for a particular type of limb. The applicant has noted that the variance in the amplitude of the load or movement measured for a particular limb can vary greatly depending on various factors such as the device type, device configuration, manufacturer, the nature of the movement or task being performed, and the type of limb being examined.The applicant has found that the accuracy of calculating asymmetry values ​​can be significantly increased by normalizing the asymmetry values ​​using a normalization factor that reflects a known or expected variance in the measurements obtained for a particular limb in a particular experimental environment. However, determining the standard deviation is time-consuming. Therefore, scaling the original data (first and second samples) and using a predefined normalization factor that is the same for all types of scaled input data (regardless of device type, movement type, or limb type) has been shown to also produce highly accurate results, but with significantly less experimental effort.

[0167] To illustrate the effect of the normalization factor τ (for unscaled data), Fig.Below are the curves of the normalized asymmetry values ​​determined from the original (unscaled) data for different normalization factor values ​​between 0 and 1. As can be seen from the graph, a very small normalization factor of, for example, 0.05 leads to a large scatter of the asymmetry scores, meaning that even very small differences between x and y result in an asymmetry score that is almost at the highest possible value (here: 1.0). Both small and large differences between x and y lead to an almost maximum asymmetry score. This effect is undesirable because the calculated asymmetry score is very sensitive to small fluctuations in the x and γ measurements and is therefore subject to artificial inflation, and because the asymmetry score may not accurately reflect the extent of asymmetry between the first and second extremities.

[0168] Setting the normalization factor to a higher value, e.g., 0.5 or more, preferably 1.0 or more, preferably 0.9 to 1.1, especially 1.0, results in a significantly better spread of the calculated asymmetry scores. This is advantageous because the calculated asymmetry score values ​​are more proportional to the difference between x and y and thus more proportional to the degree of observed asymmetry.

[0169] As in Fig.As can be seen, normalizing the asymmetry scores can have the further advantage that the generated output, the normalized asymmetry scores, is restricted to a predefined, fixed range of values, e.g., [-1,1]. This range of values ​​is preferably device-independent and limb-independent, i.e., the same range of values ​​for the normalized asymmetry scores is used for many different types of measuring devices and for many different types of limbs and measured movement or loading parameters. This can ensure that the asymmetry score calculated for a particular limb and movement type, e.g., a repeated change in knee abduction angle, is comparable to, and can therefore be integrated with, asymmetry scores calculated for a different limb type and movement, e.g., a repeated change in hip flexion angle.

[0170] As from Fig.As can be seen, although the use of the normalization factor to calculate the normalized asymmetry values ​​can enable a robust calculation of asymmetry, the resulting asymmetry values ​​may still not be able to accurately quantify movement or loading asymmetries in two corresponding limbs on the opposite side of a person's body in some situations: For example, when the absolute differences of x - y are large (e.g., 10), a large τ value is preferred to adequately differentiate the resulting asymmetry scores. When the absolute differences of the values ​​for x - y are small, a small τ value is advisable. Therefore, a normalization factor that is well suited to quantifying large movement or loading asymmetries in a person (corresponding to large x-γ differences) may be less suitable to adequately quantify small asymmetries, and vice versa.Second, the entire range of possible asymmetry score values ​​[-1, 1] calculated in the normalization step is only fully exploited at infinity.

[0171] To overcome these problems, the input signals (x and y) are scaled to a common range, e.g., [0 - 1], before being used to calculate a normalized asymmetry score (see Fig. ).

[0172] Fig. is a diagram illustrating the effect of different normalization factors τ on the calculated normalized asymmetry value, using the same formula as for calculating the Fig. In contrast to the results presented in Fig. The results presented were those in Fig. However, the normalized asymmetry values ​​presented are calculated using data scaled to a predefined range [0; 1,0]. Therefore, a comparison of the Fig.an evaluation of the effects of the scaling step for different normalization factors τ on the normalized asymmetry score.

[0173] Thanks to the scaling, the normalized asymmetry values ​​exploit Fig. better represent the entire range of possible outcome values, regardless of the variance in the original measurement data, which can depend on many different factors such as device type, limb type, measured load or movement parameter, movement pattern type, or similar. As in Fig.As shown, for input data scaled to the range [0 1], normalization factors of at least 0.9, preferably of approximately 1.0 or higher, seem to be ideal, since this normalization factor range leads to an approximately linear behavior of the normalized asymmetry score with respect to the difference in xy. For this reason, when using input data scaled to [0 1], τ = 1 can be chosen to allow a good comparison of several biomechanical variables.

[0174] If the input data are scaled to a different range, the optimal τ must be adjusted accordingly () to ensure an approximately linear correspondence between the normalized asymmetry value and the difference of the scaled differences of the values ​​of the respective pairs of first and second data points (xy values).

[0175] If the originally measured data values ​​x, y have an observed standard deviation of sd and no scaling were applied, the normalization value τ could be set to a value similar to sd. However, in the examples of the invention, the (preferably averaged) data points in the first pattern (each denoted as "x") and the (preferably averaged) data points in the second pattern (each denoted as "y") are scaled to a common range. In this case, the normalization factor τ of the scaled variable is chosen such that τ=2sd(xymax−xymin) where sd is the standard deviation of the measured amounts of movement or strain observed or expected to be observed when a particular measuring device is used to measure the strain or movement induced by a repeatedly performed movement or task in the first and second extremities, where xy maxis the maximum value observed in all (preferably averaged) data points included in the first and second patterns together, and where xy min is the minimum value observed in all (preferably averaged) data points contained in the first and second patterns together. Consequently, the xy max Value can be the value of a data point in the first pattern (an "x" data point) or the value of a data point in the second pattern (a "y" data point), depending on which of the two patterns contains the data point with the maximum value. Likewise, the xy min Value can be the value of a data point contained in the first pattern (an "x" data point) or the value of a data point contained in the second pattern (a "y" data point), depending on which of the two patterns contains the data point with the minimum value.

[0176] As an effect of the scaling, the normalization factor τ can be set to a value that does not necessarily reflect the standard deviation (sd) of the originally obtained measured values, but can be chosen as a number that correlates positively with the inverse of xy max - xy min . Furthermore, the normalization factor can be selected such that it correlates positively with the standard deviation (sd) of the originally measured data values ​​of the first and second sample instances. Essentially, the normalization factor is a value that is selected depending on the predefined value range used in the scaling step and is suitable for calculating a normalized score value based on the scaled first and second data points, which increases approximately linearly with the difference between the first and second data points of a respective data point pair.

[0177] The applicant has found that, in a situation where scaling is applied to a value range of [0,1.0], setting the normalization factor τ to a value of at least 0.8, in particular at least 0.9, and especially to a value in the range between 0.9 and 1.0, generally yields good results, e.g., accurate and comparable asymmetry values. A normalization factor value range of at least 0.8, and preferably about 1.0 or higher, is particularly useful in situations where the difference between the scaled x and y values ​​is approximately evenly distributed over a certain value range, e.g., -1 to 1. Choosing a normalization factor τ > 1 makes almost no difference to a normalization factor of 1.0; even τ = 100 results in only a straight diagonal line.The "difference of scaled x and y values ​​(x - y)," hereinafter also referred to as the "z value," refers to the difference between the scaled amount of movement observed for the first limb at a specific time point in an observation time interval and the scaled amount of movement observed for the second limb at that specific time point. The "scaled amount of movement or load" is the x or y value obtained in the scaling step.

[0178] According to some examples, the normalization factor τ is set to a value τ < 1, in particular to a value between 0.1 and 0.5, and especially to a value between 0.2 and 0.5 if the difference between the scaled x and y values ​​is very unevenly distributed. For example, in a situation where almost all z values ​​are between -0.1 and 0.1 and only very few are close to -1 and 1, a normalization factor τ = 1 would lead to asymmetry values ​​that would also all be close to -0.1 and 0.1 and only a few close to -1 or 1. However, if a small normalization factor is chosen, e.g. For example, if τ= 0.1, the resulting asymmetry values ​​for almost all z-values ​​are distributed over the interval between approximately -0.4 and 0.4, while the AS results for the few “outliers” would remain close to -1 or 1.Thus, the differences in the symmetry results calculated with a small τ would be more visible and distinct than the results obtained based on a normalization factor of about 1.0.

[0179] In summary, τ can be determined using one of the following two methods: a) choose τ as a value that is positively correlated with the inverse of τ. xy max - xy min correlated, as explained above. This can be considered a general approach for selecting τ, which works even when the scaling step yields scaled data values ​​within a range other than [0,1.0]; b) Choose τ as a value within the range of at least 0.8, e.g., a value in the range of 0.8 to 1, or in some cases 0.1 to 0.5. This may be applicable if the scaling step yields scaled data values ​​within a range of [0, 1, 0].

[0180] Fig. 20 is a block diagram of a distributed computer system 1800 used to execute the computer-implemented asymmetry quantification method.

[0181] According to some examples, all of the steps described herein for examples of the computer-implemented method for calculating asymmetry scores may be performed on a single data processing system, e.g., a data processing system 1802 communicatively coupled to one or more measurement devices 1806 configured to measure and transmit load and / or motion data of the subject to the data processing system. The measurement data may be transmitted to the computer 1802, for example, via a wired connection, e.g., USB, or via a wireless connection, e.g., Bluetooth or WIFI. According to other examples, some or most of the steps for calculating the asymmetry score pattern may be performed on a remote computer system 1812 operatively coupled to the local computer system 1802.In this case, the local computer system 1802 may act as a client computer system and the remote computer system 1812 may act as a corresponding server computer system.

[0182] For example, the client computer system may include software configured to perform one or more of the following steps: evaluating the data quality of the measuring device(s) 1806; calibrating the measuring device; receiving and processing measurement data from the measuring device 1806; analyzing the received measurement data, e.g., to determine whether one or more stability criteria are met, e.g.,, to filter out unstable or low-quality movement or loading pattern instances; performing time warping to align multiple instances of a repeatedly executed movement or loading pattern; the aligned pattern instances may in fact be time series of measured data points, and the alignment may be based on aligning start data points and end data points of the multiple time series with each other; using the aligned pattern instances to calculate the "typical" movement or loading pattern of the limb from which the movement or loading data was obtained. For example, the typical pattern may be calculated by calculating the average pattern of the entire matched pattern instances.

[0183] The client computer then sends the typical pattern observed for the first limb (“first pattern”) and the typical pattern observed for the second limb (“second pattern”) to the server computer via a network connection, e.g., the Internet. The server computer can then further process the received data: The server can scale the first and second patterns, calculate a normalized asymmetry score pattern, and send this normalized asymmetry score pattern back to the client computer, e.g., via a web service interface, by email, or via a web page displayed in a browser running on the client computer. Preferably, the server further comprises software configured to use the asymmetry score pattern for computer-assisted diagnosis or to determine whether a surgical procedure or treatment to treat an asymmetry or other condition was successful.

[0184] In some implementation variants, or for some types of limbs or devices or client computers, the server computer may perform one or more of the tasks described as being performed by the client computer, such as filtering, time warping, and averaging.

[0185] The server computer may be connected to one or more client computer systems 1802, 1818. The client computer systems may, for example, belong to a variety of different hospitals, doctor's offices, rehabilitation centers, or other healthcare facilities. Additionally or alternatively, some client computer systems may also belong to gyms, fitness clubs, shoe stores, or private individuals who may have an interest in being able to monitor the success of a treatment, such as physical therapy or the use of orthopedic shoes, shoe soles, or other orthopedic devices.

[0186] The server computer may include a user registry 1814 with a plurality of registered users. The server computer may be configured to generate a graphical user interface that allows each user to submit measured (and optionally processed) motion or loading data from at least one, preferably several different limb types (e.g., hip and knee and ankle and / or elbow, wherein for each limb type, motion or loading data obtained from the left and corresponding right limbs is used).

[0187] Although the invention has been shown and described in detail in the drawings and the foregoing description, these illustrations and descriptions are to be considered as illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments.

Claims

[1] A computer-implemented method for quantifying a movement asymmetry and / or a loading asymmetry of a pair of limbs of a subject, the subject being a human or non-human individual, the method comprising: - providing (102) a first pattern, the first pattern comprising a time series of first data points, the time series of the first data points indicating a movement pattern or a loading pattern of a first extremity of the subject, the first extremity being an extremity or a joint of a first side of the subject's body; - providing (104) a second pattern, the second pattern comprising a time series of second data points, the time series of the second data points indicating a movement pattern or a loading pattern of a second extremity of the subject, the second extremity being an extremity or a joint of a second side of the subject's body, the second side being opposite the first side, the second extremity being of the same type as the first extremity; - performing (106) a linear transformation of the data values of the first and second data points so as to obtain a scaled first pattern comprising a time series of scaled first data points and a scaled second pattern comprising a time series of scaled second data points, wherein the data values of the scaled first and second data points have been transformed into a common data value range; - identifying (108) pairs of first and second scaled data points representing the same time within an observation time interval covered by the first and second patterns; - calculating (110) a normalized asymmetry score pattern comprising a plurality of normalized asymmetry scores, wherein each asymmetry score is calculated from a respective one of the pairs of first and second data points as a function of the difference between the data value of the scaled first data point and the data value of the scaled second data point of that pair, wherein each normalized asymmetry value indicates the magnitude of a movement and / or loading asymmetry between the first and second limbs at a time in the observation time interval represented by the pair of data points, and wherein each normalized asymmetry value has been normalized to a predefined output value range, wherein the calculation of the normalized asymmetry score AS is performed with a variability margin of + / - 15%, wherein the normalized asymmetry score is a value in a value range,which is calculated according to the following formula: AS=(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1)±15%, ◯ where xy is the difference between the scaled data values of a respective pair of first and second data points, ◯ where x is a numerical value indicating the scaled magnitude of the movement or strain characterized in the one of the first data points measured at a given time within the observation time interval, ◯ where y is a numerical value indicating the scaled magnitude of the movement or strain characterized in a second data point obtained at the same time within the observation time interval as the first data point, ◯ where τ is a normalization factor with a predefined numerical value; - outputting (112) an indication of the presence and / or magnitude of the subject's asymmetry as a function of the normalized asymmetry score pattern to enable use of the indication for computer-assisted diagnosis and / or to enable use of the indication for evaluating the effectiveness of a surgery or treatment to compensate for or reduce movement asymmetries. [2] The computer-implemented method according to claim 1, - wherein the data value of each of the first data points represents an amplitude of a measured movement or load obtained for the first limb at a specific time in the observation time interval, and / or - wherein the data value of each of the second data points represents an amplitude of a measured movement or load obtained for the second limb at a specific time in the observation time interval, and / or - wherein the first pattern consists of the time series of the first data points or comprises a curve containing the series of the first data points; and / or - wherein the second pattern consists of the time series of the second data points or comprises a curve containing said series of second data points. [3] The computer-implemented method of any preceding claim, further comprising: - Selection of the common data value range into which the data values are transformed as a measuring device-independent, limb-independent and load or movement parameter-independent value range; and / or - Selection of the predefined initial value range to which the asymmetry scores are normalized, as a measuring device-independent, limb-independent and load or movement parameter-independent value range. - repeating the method according to any one of the preceding claims two or more times for one or more of the following: different types of limbs, different load or movement parameters and different types of measuring devices. [4] A computer-implemented method according to any one of the preceding claims, wherein the calculation of the normalized asymmetry values is performed such that the normalized asymmetry values increase approximately linearly with the magnitude of the difference between a respective pair of first and second data points used to calculate a respective normalized asymmetry value. [5] The computer-implemented method of any preceding claim, wherein calculating the normalized asymmetry score pattern comprises calculating each of the normalized asymmetry scores AS as a function of the formula: AS=(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1), ◯ where (xy) is the difference between the scaled data values of a respective pair of first and second data points, ◯ where x is a numerical value indicating the scaled magnitude of the movement or load characterized in the one of the first data points measured at a given time within the observation time interval, ◯ where y is a numerical value indicating the scaled magnitude of the movement or load characterized in a second data point obtained at the same time within the observation time interval as the first data point, ◯ where τ is a normalization factor with a predefined numerical value. [6] A computer-implemented method according to claim 5, wherein each of the normalized asymmetry values AS is calculated as a function of the formula: AS=sign(x−y)∗(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1), ◯ Where (xy) is the difference between the scaled data values of a respective pair of first and second data points, ◯ where the sign of the asymmetry rating indicates the direction of the movement or load asymmetry, ◯ where x is a numerical value indicating the magnitude of the movement or load characterized in the one of the first data points measured at a given time within the observation time interval, ◯ where y is a numerical value indicating the magnitude of the movement or strain characterized in a second data point obtained at the same time within the observation time interval as the first data point, ◯ where τ is a normalization factor with a predefined numerical value, ◯ and wherein the normalized asymmetry score pattern includes an indication of the direction of movement asymmetry as a function of the sign of each of the plurality of normalized asymmetry scores. [7] The computer-implemented method according to claim 5 or 6 further comprises: - Analysis of the distribution of the scaled first and second data points within the common scale; - Determine whether at least 80% of the not yet standardised score values lie in the lowest or highest quarter of the common data value range; - if the determination is ‘true’, use of a normalisation factor τ in the range [0.1-0.8] to calculate the normalised asymmetry score values; and - if the determination is incorrect, to use a normalisation factor τ greater than 0.8, in particular a normalisation factor τ of 1 for the calculation of the normalised asymmetry score values. [8] The computer-implemented method according to any one of the preceding claims, - where the first limb is a limb of the left side of the body and the second limb is a limb of the right side of the body, or vice versa; or - wherein the first limb is a limb affected by a movement or load anomaly or impairment, in particular an injury, a disease, a deformity and / or a surgical symptom, and the second limb is a limb on the opposite side of the body which is not or less severely affected by said anomaly or impairment, or vice versa. [9] A computer-implemented method according to any one of the preceding claims, wherein the method is performed by a server computer, and wherein the first and second patterns are received from a client computer over a network or are calculated by the server from measurement data received by the server from the client computer over the network, and wherein outputting comprises transmitting the indication of the presence and magnitude of the motion asymmetry over the network to the client computer or another client computer. [10] The computer-implemented method of any preceding claim, wherein performing the linear transformation comprises scaling the sizes of the first pattern and the second pattern, the scaling comprising: - Calculate a scaled data value x for each of the first data points according to : x=xori−min(DS1,DS2)max(DS1,DS2)−min(DS1,DS2)' - Calculate a scaled data value y for each of the second data points according to: y=yori−min(DS1,DS2)max(DS1,DS2)−min(DS1,DS2)' - where x ori is the data value of the first data point before the transformation is performed; - where y ori is the data value of the second data point before the transformation is performed; - where DS1 represents the time series of the first data points; - where DS2 represents the time series of the second data points; - where min (DS1, DS2) is the minimum data value that can be obtained by combining all the first data points contained in DS1 and all the second data points contained in DS2 before the transformation is performed, - where max (DS1, DS2) is the maximum data value contained in the combination of all first data points contained in DS1 and all second data points contained in DS2 before the transformation is performed, - where the asymmetry score for each of the pairs of first and second data points is used as a function of the difference between the scaled data value x of the first scaled data point and the scaled data value y of the second scaled data point of that pair. [11] The computer-implemented method according to any one of the preceding claims, the method comprising: - receiving a series of repeatedly measured instances of the first pattern, each instance of the first pattern describing the first movement or loading pattern observed for the subject's first limb during the subject's repeated performance of a particular movement or task by the first limb; - receiving a series of repeatedly measured instances of the second pattern, each instance of the second pattern describing the second movement or loading pattern observed for the subject's second limb during the subject's repeated performance of the particular movement or task by the second limb; - calculating an average first pattern from the series of instances of the first pattern, the average first pattern comprising a time series of first average data points, each of whose data values indicates an amplitude mean of the movement or load observed for the first member at a particular time within the observation time interval in all measured instances of the first pattern; - calculating an average second pattern from the series of instances of the second pattern, the average second pattern comprising a time series of second average data points, each of whose data values indicates an amplitude mean of the movement or load observed for the second member at a particular time within the observation time interval in all measured instances of the second pattern; - Using the first average data points as the first data points and using the second average data points as the second data points to perform the identification of the pairs of the first and second data points and to calculate the normalized asymmetry score pattern. [12] The computer-implemented method of any preceding claim, wherein the first and second data points represent data selected from a group consisting of: - Kinetic data, in particular forces, moments or pressures measured on the subject's lower or upper limbs and joints; - Kinematic data, in particular translational or rotational data, originating from the person's lower or upper limbs and joints; - Spatio-temporal data, in particular length, duration or width, measured at the person's lower or upper limbs and joints; - Electromyography data, in particular the electrical activity generated by the muscles during movement. [13] A method for quantifying the success of a surgical procedure or treatment to restore or reduce movement asymmetry in a person, comprising: - performing the method according to any one of the preceding claims 1-12 using first and second patterns obtained from movement and / or loading data measured before or after surgery or the start of treatment to calculate and output an initial normalized asymmetry score pattern for the subject; - repeatedly performing the method according to any one of the preceding claims 1-12 using first and second patterns obtained from movement and / or load data measured after the operation or the start of treatment, whereby a normalized asymmetry score pattern for the subject is calculated and output in each case; - Analyzing changes in the calculated normalized asymmetry score patterns over time; and - Provide an indication of whether the surgery or treatment was successful based on observed changes in the normalized asymmetry score patterns over time. [14] A method for quantifying the movement or loading asymmetry of two or more different types of limbs of a subject, comprising: - performing the method according to any one of the preceding claims 1-12 once or multiple times to quantify the movement asymmetry of the first and the corresponding second limb of the subject and for a first type of movement or load data selected from a group consisting of kinetic data, kinematic data and spatio-temporal data, thereby obtaining one or more first normalized asymmetry score patterns; - single or multiple performance of the method according to any one of the preceding claims 1-12 for quantifying the movement asymmetry of the first and second limbs and for a second type of movement or load data, wherein the second type of movement or load data is also selected from said group and differs from the first type of movement or load data, thereby obtaining one or more further normalized asymmetry score patterns, wherein all first and further normalized asymmetry score patterns are normalized to the same predefined initial value range; - analyzing the one or more first normalized asymmetry score patterns and the one or more further normalized asymmetry score patterns to calculate a combined asymmetry score pattern for the first and second extremities as a function of the one or more first asymmetry score patterns and the one or more further asymmetry score patterns; and outputting a result of the analysis. [15] A computer system (1800) configured to quantify a movement asymmetry and / or a load asymmetry of a pair of limbs of a subject, the subject being a human or a non-human individual, the system comprising: - a network interface (1801) for: ◯ Receiving data characterising a movement or load of a first and a second limb of the subject, wherein the first limb is a limb or joint of a first side of the body; - a processor (1803, 1805) configured to: ◯ processing the received data to provide (102) a first pattern, the first pattern comprising a time series of first data points, the time series of the first data points indicating a movement pattern or a loading pattern of a first extremity of the subject, the first extremity being an extremity or a joint of a first side of the subject's body; ◯ Processing the received data to provide (104) a second pattern, the second pattern comprising a time series of second data points, the time series of the second data points indicating a movement pattern or a loading pattern of a second extremity of the subject, the second extremity being an extremity or a joint of a second side of the subject's body, the second side being opposite the first side, the second extremity being of the same type as the first extremity; ◯ performing (106) a linear transformation of the data values of the first and second data points to obtain a scaled first pattern comprising a time series of scaled first data points and a scaled second pattern comprising a time series of scaled second data points, wherein the data values of the scaled first and second data points have been transformed into a common data value range; ◯ identifying (108) pairs of first and second scaled data points representing the same time within an observation time interval covered by the first and second patterns; ◯ Calculating (110) a normalized asymmetry score pattern comprising a plurality of normalized asymmetry scores, wherein each asymmetry score is calculated from a respective one of the pairs of first and second data points as a function of the difference between the data value of the scaled first data point and the data value of the scaled second data point of that pair, wherein each normalized asymmetry value indicates the magnitude of a movement and / or loading asymmetry between the first and second limbs at a time in the observation time interval represented by the pair of data points, and wherein each normalized asymmetry value has been normalized to a predefined output value range, wherein the calculation of the normalized asymmetry score AS is carried out with a variability margin of + / - 15%, wherein the normalized asymmetry score is a value in a value range calculated according to the following formula: AS=(1−2τ2τ2+abs(x−y)) / (1−2τ2τ2+1)±15%, - where xy is the difference between the scaled data values of a respective pair of first and second data points, - where x is a numerical value indicating the scaled magnitude of the movement or strain characterized in the one of the first data points measured at a given time within the observation time interval, - where y is a numerical value indicating the scaled magnitude of the movement or strain characterized in a second data point obtained at the same time within the observation time interval as the first data point, - where τ is a normalization factor with a predefined numerical value; ◯ Outputting (112) an indication of the presence and / or magnitude of the subject's asymmetry as a function of the normalized asymmetry score pattern to enable use of the indication for computer-assisted diagnosis and / or to enable use of the indication for evaluating the effectiveness of a surgery or treatment to compensate for or reduce movement asymmetries.

Images