AI-based Model for Rehabilitation Engagement and Motor Performance Evaluation in Pediatric Patients

This measure is intended to capture the therapist's observation of patient engagement. PRIME-SP is a validated self-reported questionnaire that is composed of three main parts: Part A, where the therapist can perform an overall evaluation of patient engagement according to a 5-point Likert scale (from 0 to 4, with higher values corresponding to positive engagement); Part B, where the therapist can perform a domain-dependent (affective, cognitive, behavioral domains) evaluation of patient engagement according to a 5-point Likert scale (from 0 to 4, with higher values corresponding to positive engagement); Part C, where the therapist can take free notes about factors and circumstances that he/she believes may have affected patient engagement in the session.

The Self-Assessment Manikin is a self-reported questionnaire that evaluates the emotional valence, arousal, and dominance of the subject in relation to a specific context. It is composed of 3 items, namely valence, arousal, and dominance, to which the therapist has to answer according to its feelings regarding the participant's emotional state. Each item is measured through a 9-point Likert scale, from 1 to 9, where:

• valence: lower values stand for greater unpleasantness; higher values for higher pleasantness.
• arousal: lower values stand for lower affective activation; higher values for higher affective activation.
• dominance: lower values stand for lower control over the felt emotion; higher values for control over the felt emotion.

PANAS is a self-reported questionnaire that is used to assess the positive and negative affective states. It is composed of two distinct scales: Positive Affect (PA) and Negative Affect (NA), with 20 items in total. Each item is structured with a 5-point Likert scale, from 1 to 5. The items are the following: interested, distressed, excited, upset, strong, guilty, scared, hostile, enthusiastic, proud, irritable, alert, ashamed, inspired, nervous, determined, attentive, jittery, active, and afraid. For each item, higher values on the Likert scale represent higher agreement with that specific word describing different feelings and emotions. Higer values for the items interested, excited, strong, enthusiastic, proud, alert, inspired, determined, attentive, active represent higher levels of affective engagement; conversely, higher values for the items distressed, upset, guilty, scared, hostile, irritable, ashamed, nervous, jittery, afraid represent lower levels of affective engagement.

It measures the number of heartbeats per minute. It is computed as the average number of R-peaks that occur within a minute. A moderately elevated HR may indicate physical or emotional engagement. Extremely high or low values might reflect discomfort or disengagement. A slightly increased HR, compared to rest condition, is generally interpreted as a positive sign of engagement during active tasks.

RMSSD is a measure of heart rate variability (HRV) that reflects short-term variations in the time between heartbeats. It is calculated as the square root of the mean squared differences between successive RR intervals (e.g., intervals between successive R-peaks). RMSSD is commonly associated with parasympathetic (rest-and-digest) activity. Higher RMSSD values are typically interpreted as positive, as they indicate a more regulated autonomic state and lower distress, potentially supporting better engagement.

HRV LF refers to the power of HRV in the low frequency range (typically 0.04-0.15 Hz). It is calculated using spectral analysis of RR intervals. Moderate LF values with respect to rest condition may suggest adaptive physiological arousal, but excessively high LF can reflect distress or discomfort.

HRV HF is the high-frequency component of HRV (0.15-0.4 Hz), associated with parasympathetic nervous system activity and respiratory influences. It is calculated using spectral analysis of RR intervals. In the context of rehabilitation, higher HF values typically indicate calm, focused engagement and are thus seen as positive when participants are actively involved without distress.

AvSCL is the average level of the tonic (slow-changing) component of skin conductance over time. It reflects general autonomic arousal. It is calculated from electrodermal activity (EDA) after separating the tonic and phasic components as the mean value over time of the tonic component. Moderately higher AvSCL values suggest increased physiological arousal, which is typically positive if not excessive, as it indicates increased engagement into therapy.

NSSCR represents the frequency of phasic skin conductance responses (rapid changes of skin conductance level, SCR) per minute. It is calculated by counting the number of over-threshold SCR peaks. A higher NSSCR generally indicates greater arousal and cognitive engagement and is considered a positive marker when aligned with task demands.

SCR LF measures the power of the skin conductance phasic signal in the low frequency band (0.045-0.15 Hz), focusing on slow oscillations in the phasic component. It is obtained using spectral analysis after SCR decomposition. An increased SCR LF may indicate sustained physiological activation and is generally interpreted as a positive indicator of cognitive and emotional engagement.

Walk Cadence is the number of steps a person takes per minute. It is calculated using inertial sensors placed on L5. Higher cadence typically reflects greater motor activation and is considered a positive outcome, especially when aligned with therapeutic goals.

Step time is the average time taken to complete one step. It is calculated as the time between heel strikes of alternating feet. Lower step times (faster stepping) can reflect better motor coordination and task engagement. A decrease in step time, if associated with smooth and safe gait, is considered a positive sign of engagement and motor activation.

The exergame score represents the participant's performance level in gamified motor rehabilitation exercises. It is calculated from metrics such as task accuracy, reaction time, and completion rate, depending on the game design. Higher scores generally indicate better motor performance and sustained attention and are also interpreted as a positive measure of both physical and cognitive engagement during the session. A unique normalized scale across different exergames is not available.