Limits...
The rates of change of the stochastic trajectories of acceleration variability are a good predictor of normal aging and of the stage of Parkinson's disease.

Torres EB - Front Integr Neurosci (2013)

Bottom Line: Yet this trend breaks down in PD where the statistical signatures seem to be a more relevant predictor of the stage of the disease.Those patients at a later stage of the disease have more random and noisier patterns than those in the earlier stages, whose statistics resemble those of the older NC.Overall the peak rates of change of the stochastic trajectories of the accelerometer are a good predictor of the stage of PD and of the age of a "normally" aging individual.

View Article: PubMed Central - PubMed

Affiliation: Psychology Department, Computer Science, Cognitive Science, Sensory Motor Integration, Rutgers University Piscataway, NJ, USA.

ABSTRACT
The accelerometer data from mobile smart phones provide stochastic trajectories that change over time. This rate of change is unique to each person and can be well-characterized by the continuous two-parameter family of Gamma probability distributions. Accordingly, on the Gamma plane each participant can be uniquely localized by the shape and the scale parameters of the Gamma probability distribution. The scatter of such points contains information that can unambiguously separate the normal controls (NC) from those patients with Parkinson's disease (PD) that are at a later stage of the disease. In general normal aging seems conducive of more predictable patterns of variation in the accelerometer data. Yet this trend breaks down in PD where the statistical signatures seem to be a more relevant predictor of the stage of the disease. Those patients at a later stage of the disease have more random and noisier patterns than those in the earlier stages, whose statistics resemble those of the older NC. Overall the peak rates of change of the stochastic trajectories of the accelerometer are a good predictor of the stage of PD and of the age of a "normally" aging individual.

No MeSH data available.


Related in: MedlinePlus

Schematics to explain the first step of the methods using data from one patient. (A) Examples of frequency histograms from the accelerometer data taken from 5 consecutive entries across sessions from 2 day readings from a single patient (PD patient Cherry, entries 1–3 are the last from 1 day, entries 4–5 are the first from the next day, for example). For the 5 entries shown in (A), the left hand subplot contains histograms of the mean acceleration. The right hand subplot contains histogram of the max acceleration deviation relative to the mean acceleration. The number of points per entry that went into each histogram is specified in each case. The curves are from probability density functions from the estimated parameters of the continuous two-parameter Gamma probability distribution family  where a is the shape and b is the scale parameter, and Γ is the Gamma function. These were fit to the frequency histograms in (A). (B) The estimated a-shape and b-scale parameters are plotted on the Gamma plane with 95% confidence intervals for each one of the estimated sets of values in (A) using the same color code as in (A). (C) Sample segment of the stochastic trajectory using the 5 measurements with the arrows indicating the flow in the order in which these 5 measurements were obtained.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3713394&req=5

Figure 1: Schematics to explain the first step of the methods using data from one patient. (A) Examples of frequency histograms from the accelerometer data taken from 5 consecutive entries across sessions from 2 day readings from a single patient (PD patient Cherry, entries 1–3 are the last from 1 day, entries 4–5 are the first from the next day, for example). For the 5 entries shown in (A), the left hand subplot contains histograms of the mean acceleration. The right hand subplot contains histogram of the max acceleration deviation relative to the mean acceleration. The number of points per entry that went into each histogram is specified in each case. The curves are from probability density functions from the estimated parameters of the continuous two-parameter Gamma probability distribution family where a is the shape and b is the scale parameter, and Γ is the Gamma function. These were fit to the frequency histograms in (A). (B) The estimated a-shape and b-scale parameters are plotted on the Gamma plane with 95% confidence intervals for each one of the estimated sets of values in (A) using the same color code as in (A). (C) Sample segment of the stochastic trajectory using the 5 measurements with the arrows indicating the flow in the order in which these 5 measurements were obtained.

Mentions: The 100-value sliding window for data entries proved stable across the set of participants. When hourly daily sessions were very dense we could also obtain wider sampling windows without affecting the final outcome of the analyses. For example, Figure 1 shows selected histograms from 2 consecutive days for a PD patient (Cherry) with 1000+ points used in each histogram for the estimation of several points of a segment of the overall stochastic trajectory. Using instead the 100-basic unit size increased the density of points in a given segment of the overall trajectory (e.g., the small segment shown in Figure 1C would have more points but would keep the general trend). Sampling 1000+ points did not change the overall trend of the patterns in the longer stochastic trajectory of a day (e.g., as the one shown in Figure 2A), but it would improve the estimation by lowering the errors and tightening the confidence intervals.


The rates of change of the stochastic trajectories of acceleration variability are a good predictor of normal aging and of the stage of Parkinson's disease.

Torres EB - Front Integr Neurosci (2013)

Schematics to explain the first step of the methods using data from one patient. (A) Examples of frequency histograms from the accelerometer data taken from 5 consecutive entries across sessions from 2 day readings from a single patient (PD patient Cherry, entries 1–3 are the last from 1 day, entries 4–5 are the first from the next day, for example). For the 5 entries shown in (A), the left hand subplot contains histograms of the mean acceleration. The right hand subplot contains histogram of the max acceleration deviation relative to the mean acceleration. The number of points per entry that went into each histogram is specified in each case. The curves are from probability density functions from the estimated parameters of the continuous two-parameter Gamma probability distribution family  where a is the shape and b is the scale parameter, and Γ is the Gamma function. These were fit to the frequency histograms in (A). (B) The estimated a-shape and b-scale parameters are plotted on the Gamma plane with 95% confidence intervals for each one of the estimated sets of values in (A) using the same color code as in (A). (C) Sample segment of the stochastic trajectory using the 5 measurements with the arrows indicating the flow in the order in which these 5 measurements were obtained.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3713394&req=5

Figure 1: Schematics to explain the first step of the methods using data from one patient. (A) Examples of frequency histograms from the accelerometer data taken from 5 consecutive entries across sessions from 2 day readings from a single patient (PD patient Cherry, entries 1–3 are the last from 1 day, entries 4–5 are the first from the next day, for example). For the 5 entries shown in (A), the left hand subplot contains histograms of the mean acceleration. The right hand subplot contains histogram of the max acceleration deviation relative to the mean acceleration. The number of points per entry that went into each histogram is specified in each case. The curves are from probability density functions from the estimated parameters of the continuous two-parameter Gamma probability distribution family where a is the shape and b is the scale parameter, and Γ is the Gamma function. These were fit to the frequency histograms in (A). (B) The estimated a-shape and b-scale parameters are plotted on the Gamma plane with 95% confidence intervals for each one of the estimated sets of values in (A) using the same color code as in (A). (C) Sample segment of the stochastic trajectory using the 5 measurements with the arrows indicating the flow in the order in which these 5 measurements were obtained.
Mentions: The 100-value sliding window for data entries proved stable across the set of participants. When hourly daily sessions were very dense we could also obtain wider sampling windows without affecting the final outcome of the analyses. For example, Figure 1 shows selected histograms from 2 consecutive days for a PD patient (Cherry) with 1000+ points used in each histogram for the estimation of several points of a segment of the overall stochastic trajectory. Using instead the 100-basic unit size increased the density of points in a given segment of the overall trajectory (e.g., the small segment shown in Figure 1C would have more points but would keep the general trend). Sampling 1000+ points did not change the overall trend of the patterns in the longer stochastic trajectory of a day (e.g., as the one shown in Figure 2A), but it would improve the estimation by lowering the errors and tightening the confidence intervals.

Bottom Line: Yet this trend breaks down in PD where the statistical signatures seem to be a more relevant predictor of the stage of the disease.Those patients at a later stage of the disease have more random and noisier patterns than those in the earlier stages, whose statistics resemble those of the older NC.Overall the peak rates of change of the stochastic trajectories of the accelerometer are a good predictor of the stage of PD and of the age of a "normally" aging individual.

View Article: PubMed Central - PubMed

Affiliation: Psychology Department, Computer Science, Cognitive Science, Sensory Motor Integration, Rutgers University Piscataway, NJ, USA.

ABSTRACT
The accelerometer data from mobile smart phones provide stochastic trajectories that change over time. This rate of change is unique to each person and can be well-characterized by the continuous two-parameter family of Gamma probability distributions. Accordingly, on the Gamma plane each participant can be uniquely localized by the shape and the scale parameters of the Gamma probability distribution. The scatter of such points contains information that can unambiguously separate the normal controls (NC) from those patients with Parkinson's disease (PD) that are at a later stage of the disease. In general normal aging seems conducive of more predictable patterns of variation in the accelerometer data. Yet this trend breaks down in PD where the statistical signatures seem to be a more relevant predictor of the stage of the disease. Those patients at a later stage of the disease have more random and noisier patterns than those in the earlier stages, whose statistics resemble those of the older NC. Overall the peak rates of change of the stochastic trajectories of the accelerometer are a good predictor of the stage of PD and of the age of a "normally" aging individual.

No MeSH data available.


Related in: MedlinePlus