Table 6 Signals constructed from key-points and used for feature extraction for each of the five bradykinesia items.

From: Computer vision quantification of whole-body Parkinsonian bradykinesia using a large multi-site population

MDS-UPDRS item

Time-series signal

Formula

Finger Tapping

Euclidean distance between the thumb-tip key-point and the index finger tip key-point, measured in units of estimated-standing height.

\({\textstyle{\overrightarrow{\left| {{\left. {{{{\mathbf{P}}}}_{h(4)}{{{\mathbf{P}}}}_{h(8)}} \right)} } \right|} \over {{{\mathrm{H}}}}}}\)

Hand Movement

The area of the convex hull (ACH) of the four finger tips key-points and the palm key-point, measured in units of estimated-standing-height squared (H2).

\({\textstyle{{{{{\mathrm{ACH}}}}\left( {{{{\mathbf{P}}}}_{h(i)}:i \in \left( {0,8,12,16,20} \right)} \right)} \over {{{{\mathrm{H}}}}^2}}}\)

Pronation-Supination

The angular velocity of the vector from the thumb-tip key-point to the little-finger-tip key-points, measured in degrees per frame.

\(\angle \left( {{{{\mathbf{V}}}}_{h(4,20)}^{t - 1},{{{\mathbf{V}}}}_{h(4,20)}^t} \right)\)

Toe Tapping

The vertical distance between the small toe and the neck, measured in units of estimated-standing height.

\({\textstyle{{\left( {{{{\mathbf{y}}}}_{b(i)} - {{{\mathbf{y}}}}_{b(1)}} \right)} \over {{{\mathrm{H}}}}}},\quad i = \left\{ {\begin{array}{*{20}{l}} {20} \hfill & {{{{\mathrm{left}}}}} \hfill \\ {23} \hfill & {{{{\mathrm{right}}}}} \hfill \end{array}} \right.\)

Leg Agility

The Euclidean distance between the knee key-point and the neck key-point, measured in units of estimated-standing height.

\({\textstyle{\overrightarrow{| {{{{\mathbf{P}}}}_{b(i)}{{{\mathbf{P}}}}_{b(1)}} |} \over {{{\mathrm{H}}}}}},\quad i = \left\{ {\begin{array}{*{20}{l}} {13} \hfill & {{{{\mathrm{left}}}}} \hfill \\ {10} \hfill & {{{{\mathrm{right}}}}} \hfill \end{array}} \right.\)