DEVICE TO DETERMINE DYSKINESIA

Abstract

A device is disclosed to assist in the determination of the presence and type of dyskinesia in a patient. The device includes a sensor, removably attachable to a patient's body, such as on a limb or torso, the sensor being capable of detecting 3-D motion. Data generated by the sensor is transferred to and retained in a data retention means. A processing means is included to process the generated data, along with a look-up table of processed data for already known dyskinesia conditions for comparison, the processing means employs an evolutionary algorithm in the classification of the data. Output means display the diagnosed condition to a user.

Claims

1. A device to determine an extent and type of dyskinesia in a subject, the device comprising: a first motion detector configured to detect 3-D motion data; a data storage to store the 3-D motion data; and a processor configured to: preprocess the 3-D motion data in the data storage to generate speed data representing movement detected by the first motion detector; apply the speed data, with a known dyskinesia condition, to an evolutionary algorithm to generate a first classifier, wherein the evolutionary algorithm is executed to produce a plurality of classifiers and a maximally-diverse classifier is selected as the first classifier; apply the speed data, with a known dyskinesia condition, to an artificial biological network to generate a second classifier; combine the first classifier and the second classifier to generate an ensemble classifier, wherein the ensemble classifier is used to classify patient data to determine the known dyskinesia condition; and provide a diagnosed condition for display to a user.

2. The device according to claim 1, wherein the 3-D motion data is provided by the first motion detector having one or more accelerometers.

3. The device according to claim 1, further comprising a second motion detector configured to detect pitch, roll and yaw.

4. The device according to claim 1, wherein the first motion detector is further configured to measure position data.

5. The device according to claim 4, wherein the position data is determined with reference to Cartesian co-ordinates.

6. The device according to claim 4, wherein the processor is further configured to convert the position data to velocity and/or acceleration data.

7. The device according to claim 1, wherein the processor is further configured to transfer the 3-D motion data to a remote data storage using a wireless network.

8. The device according to claim 1, further comprising a plurality of sensors to capture the 3-D motion data.

9. The device according to claim 1, wherein the first motion detector obtains data readings at a rate of from 10-200 Hz.

10. The device according to claim 9, wherein the first motion detector obtains data readings at a rate of substantially 100 Hz.

Description

BRIEF DESCRIPTION OF THE DRAWING

[0021] The invention is described with reference to the accompanying drawing which shows by way of example only, one embodiment of a device. In the drawing:

[0022] FIG. 1 illustrates the sampling and processing of data from a subject.

DETAILED DESCRIPTION OF THE INVENTION

[0023] Although Parkinson's disease can be treated through the oral administration of the drug Levodopa, it is very important that the dose administered be the correct one. Both a deficiency and an excess of Levodopa can result in the patient suffering unwanted effects. The situation is complicated by a number of factors. Firstly, there is the general concern as to whether the patient, who may well be in a confused state of mind, is taking the prescribed medication correctly. Second, the effects of a wrong dosage can very often, especially to the untrained eye be indistinguishable from one another. It is certainly not unknown, for example, for individuals whose Levodopa levels were too high, to be misdiagnosed with the opposite problem and given additional doses of the drug and for patients with too low a level to have the drug withheld.

[0024] Often the only way of assessing accurately the state of a patient is for close monitoring in a hospital under the detailed supervision of a qualified physician.

[0025] The present invention provides a device for monitoring the movement of a patient which device includes a means for effectively determining the symptoms displayed by the patient and also of clarifying said symptoms to provide an assessment of the patient. This then enables the correct treatment to be carried out to address the second of the above problems and in many instances avoid the first problem.

[0026] The invention contemplates in its broadest aspect a device for determining dyskinesia, including the type of dyskinesia being exhibited by a patient. The type of dyskinesia is related to levels of Levodopa within the patient's body. A device, or plurality of devices records the movements of the subject under investigation. The movements are then analysed for given markers which are indicative of the particular dyskinesia type.

[0027] In one embodiment of the invention, the subject patient wears a number, typically eight sensors, about their limbs, and also attached to their head and trunk. These record the direction, velocity, or acceleration of the limb. Data can thus be collected on a continuous and automatic basis without discomfort to the subject, who can be in their normal environment, and also without the need for a medical practitioner to be present to observe the subject.

[0028] As illustrated with respect to FIG. 1, the sensors are wireless accelerometer gyroscopes which are small enough to be strapped to and worn by the subject. Attachment to the patient can be by conventional means, such as Velcro. The sensors include six measuring means to measure movement in three mutually perpendicular directions x, y, z along with pitch, roll and yaw, (i.e. rotation about each of those three directions). Data is sampled in accordance with a timing mechanism within the sensor at a rate of about 100 Hz. Other data collection rates can be contemplated such as 10, 20, 60 Hz, along with 120, 150, 200 Hz etc. The skilled person will make the decision as to the rate to use based on the desired accuracy and the memory capacity available for the data. Increasing the amount of data seat will usually require a greater data storage capacity.

[0029] Any data which needs to be transmitted can be sent via, for example, Bluetooth to a Smartphone. Alternative means can of course be employed. If analysis means are not included with each sensor then processing can be carried out remotely.

[0030] An example of processed data is as follows. Firstly, speed data on the movement of an individual sensor can be obtained from accelerometer derived data or other position data, from the application of geometry. So for example, the distance d moved during two different data sampling times (t1) and t, is d=sqr[(x.sub.tx.sub.t1).sup.2+(y.sub.ty.sub.t1).sup.2+(z.sub.tz.sub.t1).sup.2], where x, y and z are expressed in Cartesian co-ordinates. This translates to a speed of d/t where t is the time between t and t1. For a sampling rate of 100 Hz t will be 0.01 s.

[0031] In addition angular rotation can be assigned with value: abs(roll.sub.troll.sub.t1)+abs(pitch.sub.tpitch.sub.t1)+abs(yaws.sub.tyaw.sub.t1) where abs indicates the operator, take absolute value.

[0032] Once calculated the speed data is presented to an analyser means to determine the type of dyskinesia being exhibited by the subject.

[0033] The analyser means employs an evolutionary analysis methodology as disclosed in GB 1100794.5 in order to aid in the decision making process. As such an initial training stage for the decision making process is used in which data from each device is passed to and processed by the methodology to predict one of the five conditions previously set down in the methodology, and is detailed below. Again as further conditions are identified, these can be included in the methodology.

[0034] The algorithm underlying the evolutionary analysis is executed a number of times. Each execution produces one or more classifiers. An ensemble classifier is then created by selecting a subset of maximally-diverse classifiers from those found during all executions of the evolutionary algorithm. This selection of maximal-diversity can be achieved either by (i) carrying out different runs of the evolutionary algorithm on different subsets of the data, or

[0035] (ii) by post-hoc analysis, where the behaviour of each classifier is explicitly measured and those with minimal behavioural overlap are chosen for the ensemble. Behaviour, in this sense, can either be the differential response of the classifier to different subsets of the data, or the classifier's ability to recognise particular patterns within the data.

[0036] In addition, angular rotation data is similarly processed through a second evolutionary analysis methodology and again the results compared with previously set down conditions to provide an assessment of the subject.

[0037] The data from the speed and the angular rotation classifier can be combined together in an ensemble classifier analysis methodology to yield higher accuracy in reaching a conclusion.

[0038] In addition, an artificial biological network (ABN) can be employed in combination with the evolutionary algorithm and an ensemble classifier.

[0039] One means of producing classifications to which the results of analysed data can be fitted, and which in particular can be used in the training stage of the analysis methodology, involves using one or more trained clinicians to carry out the assessment. For example, the clinicians can assign a value of, for example, 1-4 against each of several types of different dyskinesia It will be recognised that a broader value range can be used, although the difficulty in assessing which of a then narrower value to assign would increase. For example, the four categories which could be used are 1) minimal, 2) mild, 3) moderate, 4) severe/incapacitating. As examples of overt dyskinesia which can be classified in this manner are a) choreiform dyskinesia, b) dystonic dyskinesia, c) other dyskinesia, d) tremor, e) bradykinesia. Again, it will be recognised that other dyskinesias can also be characterised as will be apparent to the skilled person. Moreover, the difference between dyskinesia (involving the side effect of the medication Levodopa) and the symptoms of Parkinson's Disease can be classified.

[0040] Once an evolutionary Algorithm has been tried to recognise dyskinesia, such as bradykinesia (the main symptom of Parkinson's Disease, a slowing of movement) and Parkinsonian rest tremor, the evolved expression is examined to identify those common aspects of all the subject patient's movement disorders that contributed most to the expression. A second evolutionary algorithm can be trained on those specific aspects to evolve a yet more discriminating expression.

[0041] It will of course be understood that the invention is not limited to the specific details described herein, which are given by way of example only, and that various modifications and alterations are possible within the scope of the invention.