WO2018146479A1 - Appareil et procédé d'évaluation d'une caractéristique d'un os - Google Patents

Appareil et procédé d'évaluation d'une caractéristique d'un os Download PDF

Info

Publication number
WO2018146479A1
WO2018146479A1 PCT/GB2018/050359 GB2018050359W WO2018146479A1 WO 2018146479 A1 WO2018146479 A1 WO 2018146479A1 GB 2018050359 W GB2018050359 W GB 2018050359W WO 2018146479 A1 WO2018146479 A1 WO 2018146479A1
Authority
WO
WIPO (PCT)
Prior art keywords
bone
vibration
tapper
subject
sensor
Prior art date
Application number
PCT/GB2018/050359
Other languages
English (en)
Inventor
Reza Saatchi
Hajar RAZAGHI
Original Assignee
Sheffield Hallam University
Sheffield Children's Nhs Foundation Trust
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sheffield Hallam University, Sheffield Children's Nhs Foundation Trust filed Critical Sheffield Hallam University
Publication of WO2018146479A1 publication Critical patent/WO2018146479A1/fr

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/45—For evaluating or diagnosing the musculoskeletal system or teeth
    • A61B5/4504—Bones
    • A61B5/4509—Bone density determination
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0048—Detecting, measuring or recording by applying mechanical forces or stimuli
    • A61B5/0051—Detecting, measuring or recording by applying mechanical forces or stimuli by applying vibrations
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0002—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network
    • A61B5/0015—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network characterised by features of the telemetry system
    • A61B5/0022—Monitoring a patient using a global network, e.g. telephone networks, internet
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the present invention relates to apparatus and method to facilitate determination of a physical characteristic of bone and in particular, although not exclusively, for the quantitative identification of bone density.
  • Bone Mineral Density is an important indicator of bone health as it correlates with bone strength and is a predictor of overall skeletal health and its fracture risk. It is well established that a reduction in bone mineral content (BMC) or bone mineral density (BMD) structurally weakens bone and thus predisposes it to fractures. This reduced bone mass can be a valuable predictor of increased fracture risk (Kanis, J.A., et al, (2005). Assessment of fracture risk. Osteoporosis International, 16(6), 581-589). The integrity of the skeleton depends on the balance between osteoblast and osteoclasts activities throughout the remodelling process.
  • Metabolic bone disease is a general term referring to diseases of bone in which bone remodelling is not normal and this abnormality results in a reduced volume of mineralized bone and/ or abnormal bone architecture (Sambrook, P., et al, (2001). The musculoskeletal system. Edinburgh, Churchill Livingstone).
  • BMD can be measured via a variety of methods, the most common method being Dual energy X-ray absorptiometry (DXA). Other methods include quantitative ultrasound (QUS) imaging and quantitative computed tomography (QCT).
  • DXA Dual energy X-ray absorptiometry
  • Other methods include quantitative ultrasound (QUS) imaging and quantitative computed tomography (QCT).
  • DXA is the most widely used method to measure BMD and fracture risk in adults and in paediatric patients (Binkley, T.L. et al., (2008). Methods for measurement of paediatric bone. Reviews in Endocrine and Metabolic Disorders, 9(2), 95-106) and is an enhanced form of X-ray technology which was commercially introduced in 1987 (Binkovitz, L.A. & Henwood, M.J. (2007). Paediatric DXA: Technique and interpretation. Pediatr radiol, 37, 21-31). DXA X-rays are typically generated with a fan beam or a pencil beam. Pencil- beam DXA machine uses small angled X-ray beam that moves through the patient's body in a linear direction.
  • the fan-beam machine uses a wider X-ray beam that decreases scan time but increases radiation dose to patients (Fewtrell, M.S. (2003) Bone densitometry in children assessed by dual X- ray absorptiometry: uses and pitfalls. Archives of disease in childhood, 88(9), 795-798).
  • the effective dose for whole body DXA is 1-5 ⁇ 8 ⁇ (National Osteoporosis Society.
  • DXA can quantify BMD in the total body or in a specific region of the skeleton such as the spine, hips, legs and arms and is an areal, rather than a volumetric density measurement.
  • bone mineral density derived from DXA is based on the two- dimensional area of a three-dimensional structure. The third- dimension, i.e. depth, cannot be determined directly, as it is in the same direction as the X- ray beam.
  • Vibration analysis includes the mechanical excitation of an object followed by the recording and analysis of the response signals.
  • the vibration response of a system to an excitation i.e. the energy causing the vibration
  • the vibration response of a system to an excitation depends on the nature and magnitude of excitation as well as the system properties, such as mass, density and physical structure.
  • By stimulating bone mechanically and analysing its response it is possible to monitor various conditions from fractures through to osteoporosis and e.g., artificial hip joint loosening (Nokes, L.D.M. (1999).
  • ICA independent component analysis
  • Reference within this specification to at least one bone characteristic includes for example the bone mineral density (BMD), the extent to which an implant is cemented, fused or otherwise attached to a bone, the extent of a fracture or fractures within a bone and any other bone characteristics including and between extreme bone status conditions from perfectly healthy bone to a bone status that gives rise to medical concerns and a need for possible therapeutic or surgical treatments.
  • BMD bone mineral density
  • the present apparatus and method provides a system to suitably induce bone vibration and record the resulting bone vibration responses.
  • Digital signal processing and pattern recognition techniques are utilised to extract relevant information from the vibration responses.
  • the present system is particularly advantageous to provide a reliable and rapid assessment method to determine a bone characteristic (such as BMD) and to assess bone status in subjects and in particular children with confirmed or suspected osteogenesis imperfector (01).
  • the vibration responses obtained from subjects according to the present apparatus and methods may be analysed and correlated with quantitative parameters obtained using other techniques with the analysis optionally comprising computer based analysis and modelling techniques.
  • the present invention provides a system to suitably induce bone vibration and record the resulting bone vibration responses.
  • computer controlled fully integrated assessment system is provided to induce bone vibration, display, store and process the vibration responses.
  • the system may comprise a handheld device comprising or connectable with a user interface to allow a user to set sample rate, the number of responses and their durations in addition to displaying the response both with regard to time and frequency domains.
  • the handheld device may be battery powered so as to be transportable and configured for placement in direct contact with the skin of a subject so as to induce vibration within the bone of the subject via a mechanical tapper the movement of which may be controlled quantitatively via the associated electronics.
  • the apparatus may be implemented as or forming a component part within a support structure adapted to positionally support a limp or body part of a subject during the recording of the induced vibration response signal.
  • the support is desk or table mounted and comprises a first region to support in touching contact a first region of the bone and a second portion in touching contact to support a second region of the bone with each respective portion of the support having components to induce vibration and to sense the induced vibration response signal, respectively.
  • Digital signal processing and pattern recognition techniques are employed to extract relevant information from the bone vibration responses induced by a tapper operated according to a repeating strike mode in which a discrete impulse strikes are applied to the bone. Accordingly, the present system via the induction of vibrations in bone, the associated signal sensing, processing and interpretation mechanism and methods provides a critical evaluation of a characteristic of a bone (such as BMD) capable of being used optionally to screen 01 particularly in children and also as a means of supplementing existing assessment techniques such as DXA.
  • a characteristic of a bone such as BMD
  • a device to facilitate an assessment of a characteristic of a bone of a subject comprising: a movable tapper to deliver a measured load force as an impact force to a first region of a bone of a subject; a load source to generate a motive force and arranged to transmit the motive force to the tapper; a quantitative load unit coupled to the load source to provide a magnitude of the impact force generated by the tapper is quantitatively pre-determined prior to delivery of the impact force to the bone via the tapper.
  • the magnitude of the impact force is predetermined prior to delivery in contrast to determination during or immediately after impact. Accordingly, the present device is configured to avoid damaging bone by an unnecessarily high impact force. As will be appreciated, an impact force of relative low magnitude may be sufficient to cause bone fracture in a subject with 01. Additionally, the present device, via the capability for variation of the known magnitude of the impact force is adapted to provide an induced vibration signal having characteristics that facilitate monitoring and recording of the signal in addition to subsequent processor based statistical analysis.
  • the load unit may comprise any one or a combination of the following set of: a load cell connected electronically to the load source and/or the tapper; a computer and/or a processor connected electronically to the load source and/or the tapper.
  • the load unit is configured to set and/or determine a supply voltage and/or a duration of the supply voltage at the load source to provide the quantitative pre-determined magnitude of the impact force prior to delivery by the tapper.
  • the load source comprises any one of: an electric motor; a solenoid.
  • the load source is configured to create a repeating predefined motive force to be delivered to the subject by the tapper as a series of impulse strikes.
  • the load unit is a load cell coupled to the load source and the tapper such that the load cell is positioned intermediate the load source and the tapper.
  • the load unit is an electronic component configured to identify a voltage applied to the load source.
  • the device further comprises a control module to control delivery of an operational parameter to the load source.
  • the control module may comprise a processor, a computer, software and a data storage utility.
  • the operational parameter comprises any one or a combination of the following set of: a voltage; an amplitude of excitation; a time period duration of a motive force pulse; a time period between respective motive force pulses.
  • the load source via the control module is configured to provide a variation of a magnitude of the motive force to selectively change a magnitude of the impact force to be delivered to the subject by the tapper.
  • the device may further comprise a tilt indicator to output an indication of an inclination of the device relative to a horizontal plane.
  • the device may further comprise a biased return actuator acting on at least a region of the tapper to provide a return force to move the tapper in a reverse direction relative to a forward direction of the tapper that delivers the impact force.
  • the software of the control module is configured to determine, identify, change and/or record a voltage applied by the power source to the load source.
  • the device may comprise at least one mechanical adjuster to change a position of the tapper at the device to provide variation of the magnitude of the impact force delivered to the bone by the tapper.
  • the apparatus comprising: a device as claimed herein; at least one vibration sensor positionable at at least one second region of the bone separated from the first region to identify a vibration signal induced in the bone and transmitted through the bone from the first region to the second region.
  • the vibration sensor comprises any one or a combination of the following set of: a microphonic based sensor; a piezoelectric based microphonic sensor; a ceramic based microphone; a capacitor based sensor; a film capacitor based sensor comprising a dielectric material; an air or glass dielectric based capacitor sensor; at least one piezoelectric sensor.
  • the apparatus may further comprises any one or a combination of: a processor, a computer, software, and a data storage facility.
  • the device may further comprise a vibration signal amplifier to amplify the vibration signal obtained from the vibration sensor.
  • the device may further comprise a signal frequency low pass and high pass filter to truncate a frequency of the vibration signal according to
  • the device may further comprise a signal analogue to digital converter to convert an analogue vibration signal to a digital vibration signal.
  • the analogue to the digital converter is configured for operation and data acquisition at a sample rate in the range 50 kHz to 150 kHz.
  • the device may further comprise vibration signal data analysis software to interrogate and/or process the vibration signal and output at least one said characteristic of the bone.
  • the apparatus may further comprise a plurality of vibration sensors positioned respectively at different regions of the bone relative to one another and the first region.
  • the apparatus may comprise at least one support having a first portion to support in touching contact the first region of the bone of the subject and a second portion to support in touching contact the second region of the bone of the subject, the tapper mounted at the support at the first portion and the vibration sensor mounted at the support at the second portion.
  • a method of inducing and recording vibration within a bone of a subject comprising: i) delivering an impact force to a first region of the bone using a tapper, a magnitude of the impact force being quantitatively pre-determined prior to delivery; ii) allowing an induced vibration signal to be transmitted through the bone; iii) recording a vibration response signal at at least one second region of the bone positionally spaced from the first region using at least one vibration sensor.
  • the method further comprises the step of conditioning the vibration response signal using an amplification and filtering device.
  • the method may further comprise converting the vibration response signal using an analogue to digital converter.
  • the method may further comprise storing data based on the vibration response signal within a data storage facility.
  • the magnitude of the impact force is quantitatively pre-determined prior to delivery using a quantitative load unit coupled electronically to a load source that is in turn coupled electronically to the tapper.
  • a method of estimating a characteristic of a bone of a test subject comprising: pre-processing data of at least one induced vibration response signal obtained at a region of the bone of the test subject via a vibration sensor to extract component part features of said signal; processing the component part features by principal component analysis to obtain a set of principal components of said signal; using a model to apply the principal components to a reference library containing a plurality of sets of principal components of induced vibration response signals obtained from a plurality of reference subjects, the reference library further comprising at least one characteristic of a bone and/or at least one characteristic of the subject of each of the respective reference subjects; obtaining from the model an estimate of the characteristic of the bone of the test subject.
  • the step of pre-processing the data comprises transforming the data by at least one or a combination of the following set of: time domain analysis; frequency domain analysis.
  • the frequency domain analysis comprises any one or a combination of the following set of: Fourier transformation decomposition; wavelet transformation decomposition; multi-resolution decomposition.
  • the at least one characteristic of the subject within the reference library comprises data relating to a set of physical characteristics of each reference subject including any one or a combination of the following set: age; gender; height; weight; body mass index; length of bone to be analysed.
  • the at least one characteristic of the bone of each of the reference subjects comprises a bone mineral density derived or obtained via a reference analytical measurement method.
  • the step of using the model to apply the principal components to the reference library comprises any one or a combination of the following set of: regression analysis; artificial neural network analysis; discriminant analysis; clustering analysis.
  • the step of pre-processing the data further comprises selecting a part of the data of the vibration response signal from which to extract the component part features.
  • the at least one characteristic of the bone of each of the reference subject comprises any one or a combination of the following set: a quantitative value
  • the method comprises estimating an extent to which: a prosthetic is secured to the bone of the test subject; or the bone of the test subject is fractured.
  • a method of assessing at least one characteristic of a bone of a subject comprising: at a first time period inducing and recording vibration within the bone via the steps of i), ii) and iii) as claimed herein; repeating the steps i), ii) and iii) as claimed herein at a second time period later than the first time period; comparing the vibration response signals obtained at the first and second time periods to identify differences in the respective vibration response signals.
  • a network to facilitate assessment of at least one characteristic of a bone of test subjects comprising: a central hub having a processor, software, at least one communication module and a data storage facility; a reference library containing sets of principal components of induced vibration response signals obtained from a plurality of reference subjects in addition to at least one characteristic of the bone and/or at least one characteristic of the subject of each of the respective reference subjects; a plurality of bone characteristic assessment devices, each device having a movable tapper, a load source and a quantitative load unit coupled to the load source to provide a magnitude of an impact force generated by the tapper is a quantitative pre-determined value prior to delivery of the impact force to the bone of each test subject by the tapper, each device further comprising
  • a device to facilitate an assessment of bone density of a body of a subject comprising: a movable tapper to deliver a measured load force as an impact to a first region of a body of a subject; a load source to generate a motive force and arranged to transmit the motive force to the tapper; a quantitative load cell to determine quantitatively a magnitude of the motive force generated by the load source and/or delivered to the subject by the tapper.
  • Reference within this specification to a 'measured' load force encompasses a force that is quantitatively identified, known or of predefined magnitude, prior to delivery to a bone to be contrasted with a force of unidentified magnitude prior to delivery.
  • the load unit is a cell that is an electronic component configured to identify a voltage applied to the load source.
  • the load cell is a transistor.
  • the load source is configured to provide a variation of a magnitude of the motive force to selectively change a magnitude of the impact by the tapper at the subject.
  • the load source is coupled to an electrical power source wherein a variation in a voltage applied by the power source to the load source provides said variation of the magnitude of the motive force.
  • the tilt indicator may comprise an analogue bobble-in-fluid type spirit level component or a digital inclinometer.
  • the return actuator may comprise a coil spring being positioned axially between at least one region of the tapper and a region of the housing and/or load cell.
  • the return actuator is capable of cooperative motion with the load source to provide a repeating (i.e., continuous or semi-continuous) linear extension and retraction of the tapper to provide multiple and repeating impacts at the bone preferably as discrete impulse strikes.
  • assessment apparatus to facilitate an assessment of bone density of a bone of a subject, the apparatus comprising: a device as claimed herein; at least one vibration sensor positionable at at least one second region of the bone separated from the first region to identify a vibration signal induced in the bone of the subject and transmitted through the bone from the first region to the second region.
  • the apparatus further comprises vibration signal data analysis software to interrogate and/or process the vibration signal and output at least one said physical characteristic value of the musculoskeletal body.
  • the software and/or data storage facility may be provided at the device or may be located remotely from the load source and the tapper.
  • the device comprises wired or wireless communication means to provide data transfer from the device to a remote device such as a personal digital assistant (PDA), tablet, phone, laptop, desktop or server utility.
  • the device may comprise Bluetooth or radiowave communication means including in particular ultrahigh frequency (UHF) or very high frequency (VHF) components.
  • UHF ultrahigh frequency
  • VHF very high frequency
  • a method of facilitating an assessment of bone density of a bone of a subject comprising: delivering a motive force as an impact to a first region of the bone of the subject using a load source and a tapper; allowing a vibration signal to be transmitted through the bone; identifying a vibration signal at at least one second region of the bone and transmitted through the done from the first region to the second region using at least one vibration sensor; interrogating and/or processing the vibration signal via a time domain and/or a frequency domain analysis.
  • the method further comprises amplifying a magnitude of the vibration signal.
  • the method comprises truncating a frequency of the vibration signal according to predetermined cut-off frequencies using a low pass and a high pass filter.
  • Figure 1 is a cross sectional view through a handheld device configured to induce vibrations within a bone a subject having a tapper, a solenoid and a load cell according to a specific implementation of the present invention
  • Figure 2 is a perspective view of the handheld device of figure 1 with selected components removed for illustrative purposes;
  • Figure 3 is an example architectural graph of a multilayer perceptron artificial network model type suitable for use with the subject invention;
  • Figure 4a are typical vibration signals obtained from ulna using a bone density assessment system according to one aspect of the present invention;
  • Figure 4b is an individual vibration response from two trials recorded from a subject
  • Figure 5a is a graph of 20 vibration responses obtained from a subject's ulna using the system according to one aspect of the present invention
  • Figure 5b is a graph of the corresponding frequency spectrums of the vibration responses of figure 5a;
  • Figure 6 is a typical magnitude frequency spectrum of the ulna within an impulse scheme resulting from the present system according a specific implementation
  • FIG. 7 is a schematic structure of a multilayer perceptron (MLP) artificial neural network suitable for use with the present invention in which input frequencies were less than 50 Hz;
  • MLP multilayer perceptron
  • Figure 8 is a graph of the MLP network performance for the impulse scheme according to one aspect of the present invention
  • Figure 9 are regression graphs for the MLP model suitable for use with the subject invention in which input frequencies were less than 50 Hz;
  • Figure 10 is an error histogram for the model of figure 9;
  • Figure 1 1 is obtained bone mineral density values from DXA scanning and calculated using the MLP model according to one aspect of the present invention;
  • Figure 12 is a graph of relative errors calculated by comparing the actual bone density values obtained from DXA scanning with those of the MLP model according to aspects of the present invention;
  • Figure 13 is a graph of a typical vibration signal obtained from the ulna using a continuous vibration mode according to the present system
  • Figure 14 is a typical vibration frequency spectrum obtained from the ulna using a continuous vibration mode according to aspects of the present invention.
  • Figure 15 is a schematic of a MLP model network suitable for use with the present invention but operated in a continuous vibration mode
  • Figure 16 is a MLP model network performance of the subject invention operated with a tapper in a continuous vibration mode
  • Figure 17 are regression graphs for a MLP model according to one aspect of the present invention having a tapper operated in a continuous vibration mode;
  • Figure 18 is an error histogram for the MLP model according to one aspect of the present invention with a tapper operated in a continuous vibration mode;
  • Figure 19 is a graph of bone density values obtained from DXA scanning and estimated according to a MLP model according to one aspect of the present invention having a tapper operated in a continuous vibration mode;
  • Figure 20 is a graph of relative errors calculated by comparing actual bone density values obtained from DXA scanning with those of the MLP model according to one aspect of the present invention having a tapper operated in a continuous vibration mode;
  • Figure 21 is a graph of participant's feedback on the vibration tests and DXA scans according to aspects of the present invention;
  • Figure 22 is a further graph of participant's feedback on the vibration tests and DXA scans according to aspects of the present invention.
  • Figure 23 is a schematic illustration of a bone vibration response recording system according to a specific implementation
  • Figure 24 is a schematic illustration of a part of bone vibration response apparatus to support an arm of a subject during assessment according to a specific implementation
  • Figure 25 is a graph of a distribution of DXA derived BMD values for subjects (children) included in a bone vibration response study
  • Figure 26 is a graph of a typical vibration response of an ulna of one of the subjects of figure 25;
  • Figure 27 is a magnitude frequency spectrum from a vibration response of an ulna of one of the subjects of figure 25;
  • Figure 28 is a scree plot of principal components extracted from the magnitude frequency vibration mode of the vibration response signals of the subjects of figure 25;
  • Figure 29 is a graph of a correlation between DXA derived and vibration analysis estimated BMD values using two analysis models referred to as (a) 'leave-one-ouf method and (b) 'partition' method;
  • Figure 30(a) and (b) are BMD bar charts for the (a) 'leave-one-ouf method, and the (b) 'partition' method of figure 29;
  • Figure 31 is a graph of differences between DXA derived and vibration analysis estimated bone mineral density values from the two (a) 'leave-one-out' and (b) 'partition' methods;
  • Figure 32(a) and (b) are box plots of BMD values obtained using the (a) 'leave-one-ou , and the (b) 'partition' methods where plot (a) is vibration analysis estimate BMD values and plot (b) is for DXA derived BMD values.
  • a handheld vibration induction device 10 comprises a main body or housing 1 1 having housing walls that define at least one or a set of internal chambers 12.
  • the internal chamber 12 may be considered to be divided into segments in the axial direction through the generally elongate device 10.
  • housing 11 comprises an external generally cylindrical configuration having a first forward end 11 a and a second rearward end 1 lb.
  • Device 10 comprises a solenoid 13 positioned towards rearward end 1 lb. Solenoid 13 is terminated at its forward end by a solenoid cap 14 from which extends a solenoid head 19 positioned towards forward end 1 la.
  • a load cell sleeve 15 extends axially forward from a region of the solenoid cap 14 towards first end 1 la with sleeve 15 positioned radially between the external housing 1 1 and a forward region of the internal chamber 12.
  • load cell sleeve 15 extends approximately half the axial length of the device 10.
  • a forward end of sleeve 15 abuts internally against an end cap 23 secured to the forward end 11a.
  • Cap 23 comprises a central circular aperture 24, formed as an opening.
  • the solenoid 13, solenoid cap 14, the load cell sleeve 15 and the cap 23 are rigidly mounted at the device 10 so as to be stationary in the axial direction.
  • a load cell 21 formed as a short cylindrical body is housed internally within chamber 12 and load cell sleeve 15.
  • the load cell 21 is capable of shuttling axially back and forth within sleeve 15 between end cap 23 and solenoid cap 14 by the action of axial
  • a coil spring 18 is mounted axially between load cell 21 and end cap 23 so as to provide a return bias force to the load cell 21 to force the cell 21 in the axially rearward direction towards solenoid 13 and away from the end cap 23.
  • a generally cylindrical tapper body 16 extends axially from and in contact with load cell 21 and is dimensioned so as to be capable of axial reciprocating motion to emerge and retract axially at aperture 24. Tapper body 16 comprises a forwardmost impact end 17, formed as a radially reduced part of the tapper body 16.
  • Tapper body 16 and the load cell 21 are coupled such that the rearward and forward motion of the load cell 21 provides a corresponding forward and rearward axial movement of the tapper body 16 and end 17.
  • Spring 18 provides a configuration for rapid reciprocating axial motion of the tapper body 16 such that the tapper end 17 is capable of extending and retracting rapidly from the opening 24.
  • Device 10 further comprises an inclination or levelling gauge 22.
  • Gauge 22 may be formed as a conventional bubble in liquid type spirit level.
  • gauge 22 may be an electronic or analogue inclinometer. This enables device 10 to be held horizontally and in contact with a region of a body of a subject under investigation.
  • device 10 is capable of operation according to an impulse strike mode or a continuous vibration mode.
  • the impulse strike mode voltage is supplied to solenoid 13 via a supply cable 20 to induce a corresponding axial displacement and retraction of tapper 16 to and from opening 24 so as to deliver a series of sequential impacts.
  • device 10 was configured to tap the bone (10 strikes for example) within a time interval of one second (as shown in Figure 4).
  • a vibration motor (not shown) is implemented in place of the solenoid 13.
  • a constant 3v DC signal is supplied to the device 10 (via LabVIEW, as described below) so the vibration motor continuously vibrates and as a result, the impact tapper end 17 is configured to continuously vibrate the bone.
  • a set time of 60 seconds was used to record the continuous vibration signal although only a few milliseconds of the signal was processed in this mode.
  • the device comprising load cell 21 (optionally implemented as a transistor) is configured with a means of accurately and quantitatively identifying a voltage applied directly to tapper 16 so as to quantitatively determine an impact strike force of the tapper end 17 at a bone of a subject.
  • the positioning of the load cell 21 axially intermediate the tapper 16 and the means to generate the motive force (i.e. solenoid 13) provides this specific quantitative evaluation of the force applied by the tapper 16 to the region of the bone as a function of time. Applying a known force to a region of the bone accordingly allows determination of a 'response function' so as to provide in turn normalisation of the applied force and a corresponding normalisation of the data obtained.
  • Such an arrangement is advantageous to provide bone density data analysis and values independently of the type of apparatus and method used for initial vibration signal data acquisition.
  • the position of the tapper end 17 of the device 10 may be adjusted relative to forward end 1 la to selectively change the magnitude of the impact force delivered by tapper end 17.
  • tapper 16 towards end 17 may be adjustable via screw threads or a sliding mechanism (not shown) to withdraw or extend tapper end 17 relative to forward end 1 1 a.
  • tapper 17 is mounted indirectly at housing 1 1 via a mount 31 that extends radially from housing 1 1 to sit over and about a mounting pin 30 connected to a rearward axial end of tapper 17 towards the rearward end 1 1 b of device 10.
  • Pin 30 and mount 31 may comprise operative screw threads (not shown) or the like to allow linear longitudinal adjustment of the pin 30 to extend or withdraw tapper end 17 relative to forward end 1 1a.
  • the mechanical adjustment of the position of tapper end 17 relative to forward end 1 la provides a device that is easily adjustable via an operator.
  • the magnitude of the impact force may be alternatively or additionally adjusted via control of at least one of electronic operational parameter such as a supply voltage to all or to selected components of device 10.
  • a further specific implementation of device 10 is illustrated to assess BMD of an ulna 38.
  • Device 10 is mounted within a first support 42a having a first portion 50 to support an elbow of an arm 37 and a second portion 51 to be placed upon a flat surface 41 such as a desk or table.
  • a second support 42b comprises a vibration sensor 36 and is shaped and dimensioned to support a wrist region of arm 37 via a first portion 53 whilst second portion 54 is positionable on flat surface 41.
  • the inputs for the neural network were the frequency parameters up to 2 kHz which were normalised so that they lay between +1 and -.1
  • the subjects' relevant parameters included gender, height, weight, age in days, dominant hand and length of the right hand were also used as inputs to the MLP after normalising between +1 and -1.
  • the gender was taken into account as +1 and -1 for females and males respectively.
  • the dominant hand was assigned as +1 and -1 for the right and left hand respectively.
  • the MATLAB function mapminmax was used for mapping the data between -1 and +1 which is formulated as y— (ymax — ⁇ ) * (. x ⁇ x min) / x max ⁇ x min) ⁇ (1) where y max and y min were +1 and -1 for x max and x min respectively.
  • the MLP had two hidden layers. There is no predefined method to exactly determine the number of nodes in the hidden layer without training networks and estimating the generalization error of each.
  • a recommended number of nodes in the hidden layer can be computed as:
  • This formula was used as the initial estimate to determine the number of nodes in the hidden layers. The exact number used was then determined through training and evaluating the MLPs with different number of nodes that different slightly from the calculated estimate. To optimise the network, the input data was randomly divided into the training, validation and test sets with the proportion of 45, 30 and 25 percentages respectively using the MATLAB function called dividerand. The transfer function for the network was the hyperbolic tangent for the two hidden layers and the linear for the output layer.
  • suitable values for the learning rate and momentum were selected as 0.01, 0.6 respectively.
  • a piezoelectric microphonic based vibration sensor (type CM-01B) was used to record vibration signals (CM-01 B, Measurement Specialties, 2015) according to both the continuous and impulsed vibration modes.
  • This sensor is lightweight with high sensitivity (typical 40 V/mm) designed to detect vibration signals (including acoustic signals), while minimizing external acoustic noise.
  • the sensor operating frequency may be between 8Hz and 2 kHz which is suitable for low frequency vibration analysis.
  • the design offers high sensitivity to vibration applied to the central rubber pad, whilst minimising sensitivity to other external movements.
  • the microphone sensor comprises a robust PBDF piezo film combined with a low-noise electronic amplifier to provide a unique sound or vibration pick-up with buffered output.
  • the design minimises external acoustic noise whilst offering high sensitivity to vibration applied to rubber pads.
  • the sensor is specifically configured for detecting sound having a broad bandwidth.
  • the use of such a sensor is advantageous to measure displacement of the bone exclusively in response to the application of the impulse force resultant from contact by tapper 16 and the device of figures 1 and 2.
  • This microphonic based sensor is advantageous over conventional accelerometer type sensors that would otherwise additionally sense movement of the body of the subject due to general movement, breathing and the pumping of blood around the body etc.
  • the induced vibration within the bone occurs at very small magnitudes and the present sensor arrangement is advantageous to monitor exclusively this vibration as a displacement within the bone and independently of any additional movement of the body part/limb containing the bone.
  • the electrical conductivity of body fluids requires a careful health and safety consideration when using medical electrical equipment to avoid electric shock.
  • a medical electrical device or sensor connected to body must be isolated to protect the body against electric shock.
  • the currents from, to or between body connections should not exceed 100 ⁇ (British Standard, 201 1).
  • the vibration sensor was insulated in a sealed plastic case. This ensured that no metal part made contact with the subject's bone.
  • Signal conditioning includes a set of operations performed on the sensor output to make the analogue signal suitable for reading by a computer.
  • the signal conditioning system designed for this study included signal amplification and filtering.
  • the resulting signal then was processed by an analogue to digital converter to allow computer read, display and store functionality.
  • the signal amplifier acts to increase the voltage range of the signal so that it can be processed by the analogue to digital convertor more accurately.
  • the wires that carry the signal from the sensor are susceptible to noise and this noise may have higher amplitude than the signal. Therefore to detect the sensor's signal, an amplifier with a high common mode rejection ratio (CMRR) is needed to amplify the signal while eliminating the common mode noise.
  • CMRR common mode rejection ratio
  • Most commonly used amplifiers with high CMRR are differential and instrumentation amplifiers.
  • AD623 instrumentation amplifier was used for this study. AD623 can operate with single supply and delivers a rail-to-rail output swing on 3V to 12V. This specification in addition to low power consumption makes AD623 suitable for battery powered applications.
  • the gain resistor was set to 1 1 kQ to provide the instrumentation amplifier with the gain of 5.
  • TLE242 (Texas Instrument, 1998) provided a constant reference voltage equal to one-half the main power supply voltage.
  • TLE2426 voltage regulator was selected because of its high performance and low power consumption. Unwanted signals include external disturbances and noise generated internally within instrumentation can degrade the performance of vibration monitoring and analysis system.
  • Signal filtering is required to reject the unwanted components of the signal while allowing the desirable components to pass through.
  • filtering is essential to restrict the signal band width to avoid the aliasing distortion that may cause unexpected frequency characteristics.
  • the vibration signal was low pass filtered using a 4th order Bessel filter with a cut off frequency of 2 kHz.
  • the filter's cut off frequency corresponded to the vibration sensor's reliable operating frequency of 2 kHz (above 2 kHz, the sensor attenuated the signal).
  • a Bessel filter was chosen as it exhibits approximately linear phase shift with frequency so its effect is to delay the output signal by a constant time period. Therefore it does not distort the signal.
  • a general Sallen-Key topology was used to design an analogue low pass 2-pole Bessel filter.
  • the 4th order low-pass Bessel filter was generated from two 2-pole filters.
  • a 14 pins LM324 operational amplifier was used in the filter circuit.
  • the LM324 (National Semiconductor, 2004) contains four independent op-amps and can be operated with a single voltage supply from 3 VDC to 30 VDC.
  • LM324 When LM324 works in a single supply mode, an offset voltage needs to be applied to shift the input signal to the positive voltage and hence to prevent the negative part of the signal to be clipped.
  • a voltage level shifter circuit was added to the filter circuit to apply a DC bias to the input signal before filtering.
  • the components' values chosen in shifter circuit were in order to have a DC offset equal to the half of the DC power supply voltage.
  • NI myDAQ national instrument data acquisition system board
  • NI myDAQ has two analogue input channels.
  • the device uses a 16- bit analogue to digital converter to sample both channels.
  • Analogue inputs can be recorded with the sample rate of up to 100,000 samples per second per channel.
  • the required sample rate was (e.g., maximum sample rate of 100,000 samples per second was used) set in the Labview software.
  • NI myDAQ can measure analogue signals up to ⁇ 10 V.
  • Laboratory Virtual Instrument Engineering Workbench (LabVIEW) software was used as interface with NI myDAQ to control the excitation system and the vibration signal acquisition.
  • the interface provided full control of both impulse and continuous vibration analysis methods.
  • the number of impacts, the time interval between successive impacts, the duty cycle of the square waveform that drove the tapper, sample rate, recording duration as well as file handing elements were controlled from this user interface.
  • the interface also displayed the recorded vibration responses in real time.
  • the vibration analysis system was powered from a single rechargeable battery that has been required by multiple regulator circuits to the required voltages, i.e. 6 volts for the tapper circuit, 5 volts for the vibration sensor, 10 volts for the amplifier and filter circuit and 3 volts for the vibrating motor.
  • a LM1085 regulator was used to perform voltage regulation for each part.
  • LM1085 is an adjustable low dropout positive voltage regulator which can set the output voltage with only two external resistors.
  • the computer controlled method to induce bone vibration was used to perform impulse and continuous vibration analysis on the children's ulna.
  • the tapper device 10 was used to induce the vibration.
  • the voltaged supplied to induce movement of the tapper 16 was 6 volts. This produced sufficient force to excite the ulna without hurting.
  • a square pulse with an amplitude of 5 volts and a duty cycle of 4% was generated using the LabVIEW. This pulse was sent through myDAQ to the MOSFET power transistor that drove the tapper 16.
  • the number of times the ulna was tapped was 10.
  • the time interval between the taps was 1 second.
  • the impulse scheme was repeated twice therefore two vibration signals were obtained. Each vibration signal included 10 vibration responses to the 10 impacts.
  • the continuous method of incusing bone vibration involved powering actuation of tapper 16 with a supply voltage of 3 volts. Using 3 volts, the vibration frequency and amplitude were 230 Hz and 6 g respectively (Precision Microdrives (b), 2013).
  • a CM-01B vibration sensor that was fully encapsulated in a plastic casing (to ensure electrical safety) was positioned on the skin above the ulnar head using Mefix self-adhesive fabric. The signal was fed to the signal conditioning system being amplified and low pass filtered with a cut off frequency of 2 kHz. The resulting signal was then digitised using the NI myDAQ and transferred into a laptop computer for display and storage. The signal sample rate for both the impulse and continuous vibration schemes was 100k and duration of recording was 10 seconds.
  • a neuron is an information-processing unit that is fundamental to the operation of a neural network.
  • the manner in which the neurons of a neural network are structured is intimately linked with the algorithm used to train the network.
  • three fundamentally different classes of neural network architectures are defined:
  • Single-layer feedforward network This is the simplest form of a layered network in which there is an input layer of source nodes that project onto an output layer of neuron, but not vice versa.
  • Multilayer feedforward network This feedforward neural network distinguishes itself by the presence of one or more hidden layers with hidden computation nodes. The function of hidden neurons is to intervene between the external input and the network output in some useful manner.
  • Recurrent network A recurrent neural network distinguishes itself from a feedforward neural network in that it has at least one feedback loop.
  • a significant property of a neural network is the ability of the network to learn from its environment, and to improve its performance through learning.
  • a neural network learns about its environment through an interactive process of adjustments applied to its synaptic weight and bias levels. Ideally, the network becomes more knowledgeable about its environment after each iteration of the learning process.
  • error- correction learning is the technique of comparing the system output to the desire output value and using that error to direct the training.
  • the most popular algorithm for use with the error-correction learning rule is the backpropagation algorithm. This algorithm uses the error to directly adjust the weights in such a way as to minimise the error at each training iteration.
  • a multilayer perceptron is a feedforward artificial neural network model using the backpropagation algorithm to solve difficult pattern recognition problems.
  • the error signal at the output of neuron j at iteration n (i.e., presentation of the nth training example) is defined by
  • Neuron j is an output node
  • dj(n) is the desire output
  • y/ (n) is the system output value.
  • the backpropagation algorithm changes the weight to minimise the error.
  • the correction ⁇ / ( ⁇ ) applied to the synaptic weight connecting neuron j is defined by the delta rule:
  • Awji (n) Awji (n - 1) + ⁇ 3 ⁇ 4 ⁇ (n) yi (n) (7)
  • the present bone density measurement system induces bone vibration to generate short (about 100 ms) signal responses that require appropriate signal processing and pattern recognition techniques to analyse and interpret.
  • signal and pattern recognition techniques were used to process the bone vibration responses according to the present invention, a summary of which is detailed below.
  • the vibration responses were represented by their average magnitudes, duration, zero crossings, trends, distribution and higher order statistics like skewness and kurtosis. These time domain features were used to characterise bone vibration responses.
  • a bone vibration response is finite in duration, typically lasting about 100 ms.
  • the signal decomposition is based on finite length basis function and so they can be more valuable than Fourier transform in analysing the bone vibration responses.
  • the inventors successfully used wavelet transform and its discrete version i.e., multiresolution analysis to identify suitable features to assess bone density.
  • the sensors that record the bone vibration responses are placed on the skin surface above the bone and thus they pick-up a mixture of vibration components that are sourced not only from the bone but also from the soft tissues such as muscles.
  • the inventors used signal source separation techniques such as independent component analysis to unmix the recorded bone vibration responses and extract components associated with bone vibration. Bone density feature interpretation and analysis
  • ANNs Artificial Neural network
  • ANNs Artificial Neural networks
  • the inventors successfully used ANNs to determine the bone density in 42 children.
  • the Anns were trained on features of bone vibration responses of a group of children and then the trained ANNs were evaluated on another group of children.
  • Clustering is a data analysis method that allows similarities in the extracted features to be explored.
  • the inventors used clustering methods such as kmeans and fuzzy kmeans to identify bones with similar densities.
  • Regression analysis allows models of data to be developed.
  • the inventors used regression to analyse bone vibration features and to produce models that assisted in their density assessment.
  • Finite element analysis allows the response of the bone to the induced vibration to be analysed.
  • Principal component analysis is a decorrelation method.
  • the inventors used principal component analysis method to decorrelate features of bone vibration responses and identify the one that provided best discrimination. Discriminate analysis
  • the inventors applied discrimination analysis to differentiate and classify bone vibration responses resultant from use of the present apparatus and methods. Results and Discussion (investigation 1)
  • Figure 4a shows the two trials of typical vibration signals obtained from a subject using the impulse excitation.
  • An individual vibration response for the 2 trials is shown in Figure 4b. Both trials have similar patterns indicating the consistency of the method to record the signals.
  • Figure 5 shows 20 vibration responses obtained from a subject (10 per each trial) and their corresponding magnitude frequency spectra. The magnitude frequency spectra were averaged and the resulting spectrum ( Figure 6) was used as input to the ANN.
  • Figure 6 shows that the main peaks are located below 200 Hz.
  • Frequency parameters were normalised between -1 and +1 using the formula (1) and used for the neural network analysis.
  • the network was trained for different sets of the frequency ranges. For instance frequency parameters up to 2 kHz were fed into the network and the results were compared to the cases when the input frequency ranges were up to 200 Hz and 50 Hz. The best results were obtained when the frequencies up to 50 Hz.
  • Magnitude frequency spectra of all subjects showed that the largest peaks are located below 50 Hz.
  • These frequency parameters i.e. 50 values
  • the subject's physical parameters i.e. 6 values indicating gender, age, height, weight, dominant hand and length of the right hand
  • Figure 8 indicates the performance of the MLP network
  • Figure 9 provides further performance indication expressed as regression graphs for the MLP model suitable for use with the subject invention in which input frequencies were less than 50 Hz
  • Figure 10 shows the error histograms for the training, validation and test sets obtained from the network.
  • the input values (i.e. 56 values) were fed into the trained network to estimate the corresponding BMD values.
  • Table 2 indicates the subjects' physiological details together with the actual BMD values obtained from DXA, the BMD values calculated by the ANN and the corresponding relative errors.
  • the relative error was calculated by dividing the absolute error by the magnitude of the exact value. Therefore
  • Figure 1 1 shows the comparison between the actual BMD values and the calculated obtained from the MLP using the frequencies up to 50 Hz as its inputs.
  • Figure 12 shows the percentages of relative errors for the BMD values calculated from the MLP using the frequencies up to 50 Hz as its inputs. The relative errors were calculated using Formula (8).
  • Figure 13 shows a small section of a typical vibration response obtained from the ulna using the continuous scheme.
  • the magnitude frequency spectrum of the signal was obtained and its parameters were mapped between -1 and +1.
  • a typical magnitude frequency spectrum obtained in the continuous scheme is shown in Figure 14.
  • Figure 19 shows the estimated BMD values obtained from the MLP in the continuous scheme in comparison with the exact BMD values.
  • the correlation coefficient between the exact BMD values with the calculated BMD values is 0.87.
  • Figure 20 shows the relative error for the BMD values calculated by the MLP in the continuous scheme.
  • the continuous scheme shows more improved results in comparison with the discrete impact ⁇ 'impulsed') scheme.
  • the network in the continuous scheme was tested with only 15% of the data sets whereas in the impulse scheme 25% of the data was used to test the network.
  • the impulse scheme of investigation 1 the best results obtained when the frequency parameters up to 50 Hz.
  • the estimated bone mineral density (BMD) values obtained from the neural network were in correlation with the exact BMD values obtained from the DXA scan with the correlation coefficient of 0.79.
  • the correlation coefficient between the estimated BMD values and the exact ones was 0.86.
  • the present inventors have identified that the present vibration analysis system (method and apparatus) performed either in the impulse scheme or the continuous form using neural network can be a useful tool to predict BMD.
  • a further bone density investigation was undertaken using the apparatus of figures 1 and 2 in addition to a vibration response recordal system illustrated in figure 23 that corresponds to the recordal system used to assess the bone density of the 41 children of the investigation discussed above.
  • a set of subjects (children and young people) were analysed to both construct a suitable model for estimating BMD values of test subjects. That is, some of the subjects were used to build the model and these may be referred to as 'reference subjects ' whilst a second set of the recruits were used to test the model by estimating their BMD, with the individuals of this second set being referred to as 'te t' subjects.
  • the apparatus and method of the further bone vibration analysis is discussed in detail below.
  • the ulna 38 was chosen for bone density assessment because of its relative ease for both inducing and measuring vibration.
  • the olecranon was avoided as it is a more sensitive part of the ulna.
  • Vibration was induced at a first region 39 by tapping the ulna two centimeters away from the olecranon. The vibration response was recorded from the head of the ulna at region 40, the point where the ulna is most prominent (closest to the skin surface).
  • the vibration tests were performed using a computer controlled system (of figure 23) that included (i) bone vibration inducing device 10, (ii) a circuit driver 32 for the tapper, (iii) a vibration response sensor 36, (iv) a signal conditioning (amplification and filtering) device 35, (v) an analogue to digital converter (myDAQO) 34 and (vi) a computer 33 for the graphic user interface, storage and analysis facilities.
  • a computer controlled system included (i) bone vibration inducing device 10, (ii) a circuit driver 32 for the tapper, (iii) a vibration response sensor 36, (iv) a signal conditioning (amplification and filtering) device 35, (v) an analogue to digital converter (myDAQO) 34 and (vi) a computer 33 for the graphic user interface, storage and analysis facilities.
  • a square pulse was sent from computer 33 through the National Instrument ® data acquisition device 34 to a CMS power transistor acting as interface to the tapper device 10.
  • the transistor ensured sufficient current to drive the device.
  • the device 10 was placed gently on the required test site (two centimetres from the olecranon) and the end 17 of the tapper 16 mildly tapped the skin surface of an arm 37 of a subject above the ulna 38 with a controlled force, inducing the required bone vibration.
  • the force magnitude and duration were controlled by varying the magnitude and duration of the supply voltage to device 10.
  • a user interface based on the National Instrument's Lab View ® fully controlled operation of the device 10.
  • the supply voltage to device 10 was 10 volts and each tap lasted for 50 ms.
  • CM-01B ® vibration sensor (CM-01B, 2017) 36 that was fully encapsulated in a customised plastic casing to ensure electrical safety for the purpose of this study.
  • sensor 36 is described as a contact microphone device that uses a sensitive but robust PVDF pizoelectric film combined with a low-noise electronic preamplifier for vibration detection. Its stated sensitivity is 40 volts/mm with flat frequency response in the range of 8 Hz to 2 kHz. Its electronic noise is 1 mV (peak-to-peak).
  • the inventors did not use inertia measurement units (e.g.
  • the microphonic sensor 36 is advantage as it is not sensitive (and does not detect) body movements.
  • Sensor 36 was fixed to the skin above the ulnar head using self-adhesive fabric.
  • the vibration signal was amplified by a factor of 6 and then lowpass filtered using 4 th order Butterworth filter with cutoff frequency set to 2 kHz. The cutoff frequency was the bandwidth limit of the vibration sensor.
  • the signal was then digitised using the National Instrument's myDAQ(TM) data acquisition device. The sample rate was 100,000 samples per second (the limit of myDAQ).
  • the myDAQ 44 was connected to the laptop computer 33 using a USB cable that displayed the bone vibration responses in realtime and stored them for off-line processing.
  • the graphic user interface was developed using the National Instrument's Lab View (TM) software. It enabled real time display of signals, allowed the user to adjust the amplitude and width of the square pulse that activated the tapper thus controlling the tapping force, the duration of each tap, time between successive taps and sample rate.
  • the vibration signals were recorded with the patient sitting on a chair of adjustable height with their dominant hand resting on a suitably located soft mat on a table.
  • the vibration was induced by gently holding the tapper on the skin, 2 cm from the olecranon.
  • the computer then by sending 10 pulses to the tapper facilitated the recording of 10
  • the vibration signals each consisting of 10 vibration responses had their mean removed, discrete Fourier transformed and their magnitude spectra were obtained.
  • the magnitudes of the peaks in the spectra declined sharply at around 300 Hz.
  • Twenty peaks gave sufficient spectral information without producing excessive number of variables.
  • the resulting magnitudes were normalised by initially dividing them by the value of the largest peak (this ensured maximum magnitude to be 1 and others were in relation to it) and then they were divided by their standard deviation. This normalisation ensured spectral features across the children could be compared.
  • the 20 selected normalised magnitudes were processed by principal component analysis (PCA) and the scree plot of the Eigen values was obtained.
  • PCA principal component analysis
  • the vibration signal for each child was represented by four main principal components.
  • age and sex are factors that influence BMD, the feature matrix also included age and sex for each child thus providing six features in total (in this matrix males and females were represented numerically by 0 and 1.
  • n is the number of children (i.e. 48)
  • pct k is the kf h principal component for child i (z-1...n)
  • agej and sext are age and sex for child i
  • ci...ce are the regression coefficients
  • BMDt is the DXA derived BMD for i th child. The regression modelling required the coefficients a...ce to be determined. As the data set was not large, two approaches were followed for comparison.
  • the feature matrix on the left hand side of (1)
  • the BMD matrix on its right hand side that indicated DXA derived BMD values
  • the distribution of whole-body BMD values provided by DXA for the children included in investigation 2 are shown in figure 25.
  • the largest proportion of BMD values is around 0.6 g/cm 2 .
  • density is conventionally measured as mass per volume with the unit of g/cm 3
  • DXA derived BMD is represented as mass per unit area, i.e. g/cm 2 , and so the unit of g/cm 2 is used for BMD representation throughout this paper.
  • Figure 26 shows a typical vibration response through the ulna 38. Its oscillation lasts about 70 ms. The amplitude of the response is initially much larger and decays very rapidly. The response has initially a narrower width (higher frequency) with duration about 5 ms.
  • Figure 27 shows the vibration response magnitude frequency spectrum recorded from a child's ulna 38.
  • the magnitude frequency spectra of the recorded vibration responses showed some variations from child to child with regards to the shape, magnitude and the frequency range but their main frequency components were below 300 Hz.
  • Figure 28 shows the scree plot of the Eigen values of the principal components used to decide on the number of components for regression analysis. The first four principal components were chosen as they contributed to 90.2% of overall Eigen value (latent).
  • Figure 29a and b show the relationship between the DXA derived BMD values and the BMD values estimated using vibration analysis using leave-one-out and partition methods respectively.
  • Figure 29a includes all children whilst Figure 29b contains the 24 children included in the evaluation matrix of regression model.
  • Figures 29a and 29b indicate that there is a relationship between the DXA derived and vibration analysis estimated BMD values, although there are significant deviations between the two measures for some children.
  • Figure 30a and b are bar charts for DXA derived and vibration analysis estimated BMD values using the leave-one-out and partition methods. The BMD differences between the DXA and vibration analysis are shown in Figure 31a and b. From Figure 30a, subjects 2, 8, 10, 15, 16, 28, 37, 39, 40, 43, 47 and 48 show a relatively larger deviation between the DXA derived and vibration estimated BMD values. The related information such as medication, previous history of bone fractures, height, weight for these children were considered.
  • the ratio of weight to height for the children showing a larger BMD difference between the DXA derived and vibration analysis estimated showed noticeable difference to the remaining children.
  • the 12 subjects with a larger deviation have similar average heights (average about 141 cm) to remaining 36 subjects but they are on average 8.1 kg heavier.
  • the percentage height to weight difference between the two group of children is 22% (i.e. (0.33-0.27)/0.27 x 100). This may indicate that the vibration analysis method may be less accurate in heavier children, probably due to more damping effect of soft tissues on recorded bone vibration responses.
  • FIGS 32a and b Box plots of DXA derived and vibration analysis estimated BMD values for leave-one-out and partition regression analysis methods are shown in figures 32a and b.
  • the vibration analysis estimated BMD values have a broader distribution than DXA derived BMD values but their medians are close.
  • Tables 7 provides a statistical comparison of DXA derived and vibration analysis estimated BMD values for the leave-one out regression analysis method.
  • Table 4 provides similar information for the partition method.
  • the mean and median BMD values for vibration analysis method (for both leave-one out and partition methods) are close to the DXA derived BMD values.
  • the range (maximum - minimum BMD values), standard deviation and interquartile range BMD values for vibration analysis are larger than those for DXA derived values.
  • Table 7 Comparison of BMD (g/cm 2 ) values obtained from dxa with those estimated from vibration analysis (VA) for the leave-one-out and partition methods
  • a summary of the statistics of a comparison of DXA derived and vibration analysis estimated BMD values for the leave-one-out and partition regression analysis methods are provided in table 8.
  • the correlation coefficients between the DXA derived BMD values and vibration analysis estimated BMDs for the leave-one-out and partition methods were 0.54 and 0.61 respectively.
  • di and Vi are DXA derived vibration analysis estimated BMD values for child i.
  • the value of ps indicate average closeness between the DXA derived and vibration analysis estimated BMD values.
  • the ps values for the leave-one out and partition methods were 80.2% and 78.6% respectively.
  • the present methods and apparatus may be regarded supplementary and not designed specifically to replace existing bone assessment systems and techniques but to be an effective, easy to use, non-invasive and completely harmless system that may complement other approaches.
  • the investigations detailed herein confirms the potential of vibration analysis in assessing bone characteristics such as BMD.
  • the present vibration analysis apparatus and methods may allow medical practitioners to have an effective tool for 'the quick screening' of those suspected of abnormal bone characteristics.
  • the present methods and apparatus used as a tool may allow subjects undergoing medically recovery to be monitored more regularly and reliably.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Artificial Intelligence (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Veterinary Medicine (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Rheumatology (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Orthopedic Medicine & Surgery (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)

Abstract

L'invention concerne un système pour faciliter la détermination d'une caractéristique physique d'un os, incluant en particulier la densité minérale osseuse (BMD), comprenant un dispositif de tapotement portatif ayant un solénoïde ou un moteur à vibration électrique configuré pour fournir une force d'impact quantitative en tant qu'un coup ou qu'une série de coups (à l'aide d'un solénoïde) ou sous la forme d'une vibration continue (à l'aide d'un moteur à vibration électrique) au niveau d'un os d'un sujet de manière à induire des vibrations dans l'os. Le signal ou les signaux de vibration sont identifiés et le signal est traité et analysé selon au moins un procédé d'analyse comprenant par exemple des réseaux neuronaux artificiels, un groupement, une analyse de régression, une analyse de composant principal et une analyse discriminante.
PCT/GB2018/050359 2017-02-08 2018-02-08 Appareil et procédé d'évaluation d'une caractéristique d'un os WO2018146479A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
GBGB1702080.1A GB201702080D0 (en) 2017-02-08 2017-02-08 Apparatus and method for quantitative determination of bone density
GB1702080.1 2017-02-08

Publications (1)

Publication Number Publication Date
WO2018146479A1 true WO2018146479A1 (fr) 2018-08-16

Family

ID=58462310

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/GB2018/050359 WO2018146479A1 (fr) 2017-02-08 2018-02-08 Appareil et procédé d'évaluation d'une caractéristique d'un os

Country Status (2)

Country Link
GB (1) GB201702080D0 (fr)
WO (1) WO2018146479A1 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109839384A (zh) * 2019-01-22 2019-06-04 四川安和精密电子电器有限公司 用于微型振动马达缺陷检测的视觉检测仪及检测方法
CN113223699A (zh) * 2021-04-01 2021-08-06 复旦大学附属华山医院 构建腰椎骨量减少和骨质疏松筛查模型的方法和系统
WO2022217997A1 (fr) * 2021-04-15 2022-10-20 四川千里倍益康医疗科技股份有限公司 Dispositif électrique de massage et son procédé de reconnaissance de squelette à base d'accélération
JP7283672B1 (ja) 2018-09-10 2023-05-30 京セラ株式会社 学習モデル生成方法、プログラム、記録媒体及び装置

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3477422A (en) * 1965-10-15 1969-11-11 John M Jurist Jr Vibratory bone density determination method and apparatus
WO2005023093A2 (fr) * 2003-09-05 2005-03-17 William Marsh Rice University Évaluation de tissu non invasive

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3477422A (en) * 1965-10-15 1969-11-11 John M Jurist Jr Vibratory bone density determination method and apparatus
WO2005023093A2 (fr) * 2003-09-05 2005-03-17 William Marsh Rice University Évaluation de tissu non invasive

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
-: "Contact Microphone CM-01 Datasheet", DURHAM INSTRUMENTS WEBSITE, 5 December 2009 (2009-12-05), pages 1 - 3, XP055468849, Retrieved from the Internet <URL:http://www.disensors.com/downloads/products/CM-01B%20Contact%20Microphone_165_292307.pdf> [retrieved on 20180419] *
ALI RIDITA ET AL: "Fuzzy C-Means clustering to analyze bone vibration as a method of screening fracture", 2016 10TH INTERNATIONAL SYMPOSIUM ON COMMUNICATION SYSTEMS, NETWORKS AND DIGITAL SIGNAL PROCESSING (CSNDSP), IEEE, 20 July 2016 (2016-07-20), pages 1 - 5, XP032968171, DOI: 10.1109/CSNDSP.2016.7573971 *
HAJAR RAZAGHI ET AL: "Vibration analysis as a tool for bone mineral density assessment in children", BRITISH SOCIETY OF PAEDIATRIC RADIOLOGY ANNUAL SCIENTIFIC MEETING, LONDON, 11 November 2016 (2016-11-11), London, XP055468470 *
RAZAGHI H ET AL: "Correlation analysis of bone vibration frequency and bone mineral density in children", 2014 9TH INTERNATIONAL SYMPOSIUM ON COMMUNICATION SYSTEMS, NETWORKS & DIGITAL SIGN (CSNDSP), IEEE, 23 July 2014 (2014-07-23), pages 188 - 192, XP032660711, DOI: 10.1109/CSNDSP.2014.6923822 *
RAZAGHI HAJAR ET AL: "An investigation of relationship between bone vibration frequency and its mass-volume ratio", 2014 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), IEEE, 4 May 2014 (2014-05-04), pages 3616 - 3620, XP032617797, DOI: 10.1109/ICASSP.2014.6854275 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP7283672B1 (ja) 2018-09-10 2023-05-30 京セラ株式会社 学習モデル生成方法、プログラム、記録媒体及び装置
JP7283673B1 (ja) 2018-09-10 2023-05-30 京セラ株式会社 推定装置、プログラム及び記録媒体
JP2023082022A (ja) * 2018-09-10 2023-06-13 京セラ株式会社 学習モデル生成方法、プログラム、記録媒体及び装置
JP2023085344A (ja) * 2018-09-10 2023-06-20 京セラ株式会社 推定装置、プログラム及び記録媒体
CN109839384A (zh) * 2019-01-22 2019-06-04 四川安和精密电子电器有限公司 用于微型振动马达缺陷检测的视觉检测仪及检测方法
CN113223699A (zh) * 2021-04-01 2021-08-06 复旦大学附属华山医院 构建腰椎骨量减少和骨质疏松筛查模型的方法和系统
WO2022217997A1 (fr) * 2021-04-15 2022-10-20 四川千里倍益康医疗科技股份有限公司 Dispositif électrique de massage et son procédé de reconnaissance de squelette à base d'accélération

Also Published As

Publication number Publication date
GB201702080D0 (en) 2017-03-22

Similar Documents

Publication Publication Date Title
CN106980746B (zh) 一种基于时序分析的通用无创血糖预测方法
WO2018146479A1 (fr) Appareil et procédé d&#39;évaluation d&#39;une caractéristique d&#39;un os
US11992335B2 (en) System for health monitoring on prosthetic and fixation devices
US20210145608A1 (en) Quantitative Design And Manufacturing Framework For A Biomechanical Interface Contacting A Biological Body Segment
Khalifa et al. Upper esophageal sphincter opening segmentation with convolutional recurrent neural networks in high resolution cervical auscultation
Doheny et al. Feature-based evaluation of a wearable surface EMG sensor against laboratory standard EMG during force-varying and fatiguing contractions
EP1454586A1 (fr) Appareil de diagnostic d&#39;une fonction chaologique du cerveau
Shafiq et al. Multimodal chest surface motion data for respiratory and cardiovascular monitoring applications
Abbod et al. Survey on the use of smart and adaptive engineering systems in medicine
Baker et al. Wearable nanocomposite sensor system for motion phenotyping chronic low back pain: a BACPAC Technology Research Site
Klum et al. Minimally spaced electrode positions for multi-functional chest sensors: ECG and respiratory signal estimation
WO2021055520A1 (fr) Plate-forme de surveillance de santé interactive pour systèmes de capteurs sans fil pouvant être portés
Tigges et al. Classification of morphologic changes in photoplethysmographic waveforms
CN108324283B (zh) 对于疼痛的记录方法
KR100955716B1 (ko) 생체데이터의 문자열 생성 방법 및 장치와, 이를 이용한 생체 파형 검출 방법 및 장치와 맥파 분류 방법 및 장치
Bilgin et al. Evaluation of vibration signals measured by 3-Axis MEMS accelerometer on human face using wavelet transform and classifications
JP7301275B2 (ja) 睡眠状態推定システム
Mauck et al. Numerical simulation of mechanically stimulated bone remodelling
Hoffmann et al. Radiopacity assessment of neurovascular implants
Laidig et al. Event-based sampling for reducing communication load in realtime human motion analysis by wireless inertial sensor networks
Pecolt et al. Conversion of bioelectric sEMG signals into analog form for the BLDC motors control
Razaghi et al. Evaluation of vibration analysis to assess bone mineral density in children
Masum et al. Measurement of walking speed from EMG signal using kurtosis of approximate coefficients
Hoffmann et al. Development of a skull phantom for the assessment of implant X-ray visibility
Bahr et al. High speed digital interfacing for a neural data acquisition system: Enabling of live display functionality for a neural recording system

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18705472

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18705472

Country of ref document: EP

Kind code of ref document: A1