3.2.2 Kinematic analysis
There are many types of kinematic sensors available like accelerometer, gyroscopes, tracking systems using infrared and ordinary cameras. By attaching a sensor to a limb the total mass of the limb changes and thus the dynamics (see chapter 2.***) and it is therefore important to have sensors with low masses. Accelerometers are the most commonly used kinematic sensors (Grimaldi&Manto 2008 93blz***why).
There are different types of accelerometers which work according with different components like piezoelectric, piezoresistive en capacitive components. The working principle is the same for all sensors: there is a mass attached to a component acting as a spring and damper. Because of an acceleration the mass moves, causing a deformation which is measured. An example of a piezoelectric accelerometer is depicted in Figure 6.
Figure 6 Piezoelectric accelerometer (Lecture slides WB2303: Measurement in engineering)
When the piezoelectric accelerometer is attached to the limb of a subject, the acceleration of the limb will result in a force on the piezoelectric crystal because of the inertia of the mass on top of it. The crystal will then deform, resulting in a voltage over the crystal which is proportional with the applied force. By using Newton's second law the acceleration can be calculated because the mass and the force are known. Accelerometers are nowadays often MEMS (microelectromechanical systems) so they have a very low mass.
Since the accelerometer uses a mass, it is also sensitive to gravity. This feature is sometimes desired, for example in digital cameras to switch between portrait and landscape or even in kinematics to determine the orientation of a limb. However, when used to record tremor the signal is corrupted by gravity and the recording shows a bias or even modulations depending on the amount of rotation involved in the tremor. These alterations would obviously also result in a affected power spectrum, so proper measures must be taken to prevent or correct for influences of gravity.
HofH
Low G accelerometers/piezoelectric detectors (Harish, Venkateswara Rao, Borgohain, Sairam, & Abhilash, 2009)
Results in (Harish, Venkateswara Rao, Borgohain, Sairam, & Abhilash, 2009)
Surface EMG and piezoresistive accelerometer (Raethjen, et al., 2004)
(Grimaldi & Manto, Neurological Tremor: Sensors, Signal Processing and Emerging Applications, 2010)
"Accelerometers are also used for intraoperative assessment of the best position to implant electrodes for deep brain stimulation (DBS) and neurophysiological monitoring of stereotactic intervention of movement disorders [28]"
***accuracy emg/accelerometer
3.2.3 Signal processing
All the steps in signal processing, from acquiring the signal to presenting the results, are susceptible of errors.
There are several factors influencing the quality of the signal to be measured i.e. the signal to noise ratio. Noise can be picked up from the environment, like the 50Hz from the mains electricity or electromagnetic fields. Poor skin contact in EMG also leads to a low signal to noise ratio. Noise can be reduced by averaging the signal (in time or frequency domain). Artifacts due to movements and other undesired phenomena like the gravitational component of an accelerometer are also pitfalls.
When the signal is converted from analog to digital, the sample rate must fulfill the Nyquist criterion to prevent aliasing. Filtering techniques like low or high pass filters can be applied to eliminated unwanted signals or noises from the measured signal. After measuring a subject's tremor, the signal is often edited through visual inspection by an expert, who looks at a graphical representation of the signal (Grimaldi & Manto, Tremor: From Pathogenesis to Treatment, 2008).
Before applying a Fourier transformation the signal is often windowed to reduce spectral leakage. Spectral leakage is caused by signals which do not have an integer number of periods in a time sample, which corrupts the frequency spectrum by showing power at frequencies which are not in the signal. There are several window functions available like Blackman, Hamming and Hanning window functions.
The frequency resolution of the power spectrum depends on the observation time. Since tremor is often non periodic, the length of the observation time is limited because techniques like Fourier transforms can only be applied to time invariant measurements. This impairment limits the frequency resolution. ***guestimation of achievable resolution
***wavelets also possible page 104
***
To investigate the relation between two signals, for example the EMG signals of two antagonistic muscles, the cross-spectral density can be calculated. By scaling the square of the cross-spectrum with the auto-spectra of the two signals, the coherence is obtained. The coherence is a value between 0 and 1 and gives an indication of a possible linear relation between the two signals. If the relationship is a nonlinear one or is there is much measurement noise, the coherence will drop.
3.3 Other methods
EEG/EMG
DBS, Local field potential
fMRI/PET
PA Studies?
***
3.4 Haptic devices
To increase the subjectivity and repeatability of a measurement, haptic devices can be used (***figure). The same sensors can be used as discussed in section 3.1 and 3.2***.
*** (Grimaldi & Manto, Tremor: From Pathogenesis to Treatment, 2008)
A subject's limb is placed on the manipulator and the subject can perform different tasks like making certain motions, applying certain forces or resisting perturbations applied by the manipulator. The manipulator has the ability to create certain environments by changing the inertia, stiffness and damping of the manipulator. By creating different conditions for the subject's limb, different characteristics can be derived from the measurement, which can help understand the functioning of the human body and may give insight into a pathology.
