P933: Quantification of movement disorders symptoms employing attitude sensors

Distinguishing Patients With a Coordination Disorder From Healthy Controls Using Local Features of Movement Trajectories During the Finger-to-Nose Test

“…Our accuracy results are consistent with other studies [17], [31], [32] that tried to distinguish between healthy and pathological FNT trials. Table I displays characteristics of earlier studies and ours using different techniques to distinguish patients from healthy participants, allowing a direct comparison.…”

“…First, because we combined participants of two different studies, our study includes the largest number of healthy participants and patients, so far. Second, our method involved only two variables whereas the other methods [17], there was no need to apply such techniques allowing for straightforward visualization and interpretation of the results. The other studies apply different classification algorithms such as random forest and k-means.…”

“…This study was performed using data acquired in the context of the project Quantification of symptoms of movement disorders employing motion sensors [17]. Part of the data in [17] were previously used to investigate whether a random forest classifier employing 14 features derived from 3D movement trajectories during the FNT could classify children with coordination problems and age-matched healthy control children. In the present study we only use two features, which are different from those used in [17] and include additional participants.…”

“…In the present study we included two groups of participants. The first group concerns the data used in the study mentioned above, consisting of 34 children: 16 By including patients and controls over a wider age range, we increased the complexity of the data compared to the previous study (see [17]). All parents of the children and all adult participants provided written informed consent.…”

“…During task execution, participants wore three inertial measurement units (IMUs Shimmer3, Shimmer, Dublin, Ireland-based Realtime Technologies) on the upper arm, fore arm, and index finger. The data collected by the IMUs at 51.2 Hz were used to estimate 3D trajectories of the participants' index finger using an upper limb model [17] implemented in Labview (Austin, Texas, USA). We subsequently applied a moving-average filter of the 3D trajectory data using a window of 15 samples to smooth the signals.…”