Objective evaluation of attention-deficit/hyperactivity disorder medication effects using a load-cell-embedded chair in a simulated classroom

Abstract

Background

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental condition in children. Treatment outcomes are assessed using established tools such as the Swanson, Nolan, and Pelham (SNAP-IV) scale and the Vanderbilt ADHD Diagnostic Rating Scale, which primarily rely on subjective reporting. In a previous pilot study, we used load cells to evaluate the effects of ADHD medication in patients receiving Ritalin immediate-release (IR). In the present study, we analyzed the effects of both Ritalin IR and ORADUR-Methylphenidate on the basis of changes in the average trajectory length (ATL) at different time points during a 15-min learning session in a simulated classroom.

Methods

A simulated classroom was constructed to replicate a real-world educational environment, featuring a desk and chair facing a large display screen. Load cells measuring mechanical force through material deformation were installed on each chair leg to monitor body movement. A total of 56 children with ADHD (boys: 49) were enrolled. They received Ritalin IR (n = 37; 10 mg/day orally) or ORADUR-Methylphenidate (n = 19; 22 mg/day orally) for 1 month. During assessment, participants were instructed to remain seated and watch an age-appropriate mathematics video. The ATL, derived through continuous load cell recordings, was calculated to quantify participants’ postural movement while seated as a representation of the mean distance of body movement over time. Parents and teachers also completed the SNAP-IV scale before and after participants began treatment.

Results

In the Ritalin IR group, posttreatment ATL values decreased significantly from 0.0318 ± 0.0191 to 0.0180 ± 0.0126 (43.40% reduction, p < 0.0001). Parent-reported SNAP-IV scores decreased from 37.75 ± 14.24 to 28.08 ± 16.94 (25.62% reduction, p < 0.0001), and teacher-reported scores decreased from 40.68 ± 17.66 to 24.36 ± 18.87 (40.12% reduction, p < 0.0001). In the ORADUR-Methylphenidate group, posttreatment ATL values decreased from 0.0431 ± 0.0277 to 0.0304 ± 0.0258 (29.47% reduction, p = 0.0140). Parent-reported SNAP-IV scores decreased from 40.75 ± 16.44 to 29.17 ± 17.80 (24.42% reduction, p = 0.0053), and teacher-reported scores decreased from 42.91 ± 18.04 to 25.64 ± 17.44 (40.22% reduction, p = 0.0082). Pretreatment ATL values progressively increased over time, particularly during the latter part of the video, whereas posttreatment values remained consistently low throughout the session.

Conclusion

The load cell–embedded smart chair may serve as an objective tool for assessing pharmacological treatment efficacy in children with ADHD, allowing researchers to effectively quantify hyperactivity and behavioral changes across time points.

Clinical trial number

Not applicable.

Introduction

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental condition that affects approximately 8–10% of all children globally [1]. Without appropriate intervention, ADHD may impair functional capabilities during early development and impose social and academic challenges later in life [2]. ADHD has been neurobiologically associated with atypical neural activity in the basal ganglia (BG), particularly in the striatum—the main input center of the BG [3]. The striatum is functionally divided into the dorsolateral, dorsomedial, and ventral (nucleus accumbens) regions, which direct motor control, executive function, and emotional regulation, respectively [4, 5]. Brain regions involved in movement—for example, the supplementary motor area and the motor, premotor, and somatosensory cortices—provide excitatory glutamatergic inputs to the dorsolateral striatum [6, 7], which transmits signals to the dorsal portions of the globus pallidus and substantia nigra [8]. Clinical and neuroimaging studies have supported the vital role of the cortico-BG circuitry in ADHD pathogenesis [9–11]. Experimental models have further corroborated this connection, demonstrating that increased activity in the nucleus accumbens, induced by glutamatergic, cholinergic, or dopaminergic agonists, can cause hyperactivity in animals [12–15]. Additionally, focal disinhibition through γ-aminobutyric acid type A receptor blockade in various striatal regions can induce hyperactivity and stereotypic behaviors in both primates and rodents [16, 17]. Given that hyperactivity is a core symptom of ADHD, precise hyperactivity monitoring is necessary to evaluate treatment outcomes.

Timely diagnosis and treatment are crucial for achieving favorable long-term outcomes in ADHD. Treatment strategies typically involve a pharmacological, nonpharmacological, or combined approach. Pharmacotherapy, particularly with stimulant medications, has demonstrated greater neurochemical efficacy than did behavioral interventions alone [18]. However, approximately 20% of all pediatric patients do not respond adequately to stimulant therapy [19]. To evaluate the effectiveness of ADHD treatments, researchers commonly use the Swanson, Nolan, and Pelham-IV (SNAP-IV) scale [20], the Vanderbilt ADHD Diagnostic Rating Scale [21], and visual analog scales [22]. However, these tools depend heavily on subjective reports from caregivers or educators, which may introduce bias and variability. Accordingly, an objective, quantitative assessment method for therapeutic outcomes is required.

Recent advancements in sensor technologies have enabled the development and use of compact, energy-efficient wireless systems in research on various medical conditions, including osteoarthritis, cerebral palsy, Parkinson’s disease, and stroke [23–26]. In parallel, research has consistently demonstrated that children with ADHD exhibit greater physical activity than do their neurotypical peers [27, 28]. In our previous pilot study, we used a load cell to evaluate the effects of Ritalin immediate-release (IR) pharmacotherapy [29]. In the present study, we extended this work by further including ORADUR-Methylphenidate in the analysis and by characterizing the changes in average trajectory length (ATL) across multiple time points during a 15-min simulated classroom session. Accordingly, we developed an objective measurement system that comprised a “smart chair” embedded with load cells. This system was deployed in a simulated classroom environment designed to replicate real-life educational settings. Children with ADHD watched an educational mathematics video while seated on the chair, and their body movements were recorded both before and after treatment with either immediate-release medication (Ritalin IR) or extended-release medication (ORADUR-Methylphenidate).

To replicate the proposed system in real-world settings, load cells can be embedded in actual classroom chairs to detect subtle movements without attracting the user’s attention. The system can collect and transmit data wirelessly through mobile devices. The intuitive user interface was designed to allow nonspecialists to operate the system with ease. This system has broad potential for application in educational settings and can serve as an effective tool for the objective evaluation of behavioral responses to ADHD treatments and interventions [30].

Methods

Study cohort and ethics

The study cohort comprised 56 children who received an ADHD diagnosis on the basis of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria between November 2022 and February 2025. All diagnoses were confirmed by a pediatric neurologist. Individuals with a history of epilepsy, intellectual disability, genetic syndromes, metabolic disorders, substance abuse, traumatic brain injury, or psychotic disorders were excluded from this study. All participants were drug naïve.

This study adhered to the Declaration of Helsinki. Informed consent was obtained from a parent or legal guardian for all participants. The study protocol was approved by the Institutional Review Board of Kaohsiung Medical University Hospital, Taiwan (approval number: KMUIRB-SV(I)-20190060).

Study procedure

After ADHD diagnosis, 37 patients received 10 mg Ritalin IR once daily for 1 month, whereas the remaining 19 received 22 mg ORADUR-Methylphenidate once daily for the same duration. Of the 37 patients who received Ritalin IR, 13 were from our previous pilot study [29]. Medication adherence was tracked through pill counts.

Each participant was required to sit on a specially designed chair embedded with load cells in a simulated classroom setting and watch a 15-min age-appropriate educational video on mathematics. All participants completed two simulated classroom sessions. The first session was conducted before participants began pharmacological treatment, and the second session was conducted after approximately 1 month of continuous medication usage under stable dosage. To mitigate potential familiarization or learning effects, different mathematics videos were played in the pretreatment and posttreatment sessions. To ensure comparability, both sessions followed the same experimental protocol. In addition, the same testing environment, recording duration, and task conditions were maintained. All sessions were conducted in the morning.

Symptom severity was determined using the SNAP-IV scale. Parents and teachers also completed the SNAP-IV scale before and after participants began treatment. All patients took their medication within 3 h before data collection. On the basis of predominant symptom cluster, ADHD was classified as the inattentive (ADHD-I), hyperactive/impulsive (ADHD-H), or combined (ADHD-C) type.

Study instrument

The SNAP-IV scale comprises 26 items—18 on core ADHD symptoms (equally divided between inattention and hyperactivity/impulsivity) and 8 on oppositional defiant disorder symptoms, measured on the basis of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria. Each item is rated on a 4-point Likert scale with end points ranging from 0 (not at all) to 3 (very much). The SNAP-IV scale has demonstrated high test–retest reliability and strong internal consistency among diverse populations in both school-based and clinical settings [31–35].

Experimental setup

To ensure ecological validity, a simulated classroom environment was created to mimic a typical elementary school setting, including a desk and chair facing a large screen [29]. The smart chair used in the experiment replicated standard classroom furniture but was enhanced with sensing capabilities. The sensing system comprised four DECENT DS-928DS full-bridge load cells (capacity: 50 kg), each installed beneath a chair leg. The load cells were connected to an analog-to-digital converter, a microcontroller, and a battery. The converter model (ADS1115) converted analog signals into 16-bit digital signals, offering a maximum sampling rate of 860 samples per second across four supported analog input channels.

An ESP32 microcontroller was used to process the sensor data, timestamp each entry, and store it within 3 Mb of onboard memory at a sampling frequency of 5 Hz, allowing for > 5 h of continuous operation. The ESP32 microcontroller also served as a local Web server, which users could connect to through mobile browsers to initiate recordings, manage data, and download files. The user interface was designed as follows:

1. Web page access: Users enter the ESP32 IP address into a mobile Web browser to access the main Web page.

1. Start recording: When the “start recording” button is pressed, the microcontroller begins recording measurement data, including time tags and the four load cell values.

2. Stop recording: When the “stop recording” button is pressed, the detector stops recording.

3. Recorded data management: Users press the “recorded data” button to manage the data recorded by the detector.

• (2)Data management Web page: Users can view, store, or delete recorded data on this Web page.
• (3)Data viewing Web page: Users can verify the correctness and completeness of the recorded data on this Web page.

Movement analysis

Children with ADHD often have difficulty remaining seated, and their frequent shifts in body posture repeatedly change their center of gravity. Accordingly, we hypothesized that these movements would produce variation in the load distribution across the four sensors embedded in the smart chair. We therefore used the weight ratios recorded by each load cell to calculate the center of gravity [29]. The four load cells were labeled according to the chair leg as follows: left front (W_LF), left back (W_LB), right front (W_RF), and right back (W_RB). W_LF, W_LB, W_RF, and W_RB were calculated as follows:

1

2

3

4

G (G_FB and G_RL) denotes the position of the center of gravity:

5

6

where the front–back (G_FB) and right–left (G_RL) gravity values range from 0 to 2. When G_FB = 0, the center of gravity is at the back of the chair, whereas when G_FB = 2, it is at the front of the chair. Similarly, when G_RL = 0, the center of gravity is on the left side of the chair, whereas when G_RL = 2, it is on the right side of the chair. If G is (1, 1), the center of gravity is located at the center of the chair.

After a recording session, the microprocessor captures the time-domain waveforms recorded by the four load cells. Figure 1 presents sample waveforms corresponding to an empty chair, a motionless seated position, a back-and-forth swaying motion, and a side-to-side swaying motion. Each movement type generates a distinct pattern in the sensor outputs.

Fig. 1.

Trajectory of the center of gravity during different movements. When the chair is empty, the center of gravity remains in the same position (orange dot). When the participant sways left and right, the trajectory shifts accordingly (yellow line). Swaying back and forth causes the trajectory to move in the same direction (green line). When the participant sits still, the trajectory changes only slightly (blue line)

At each moment, the load cell readings correspond to a specific center-of-gravity position. Therefore, the time-domain waveforms of the four sensors delineate the movement trajectory of the chair user’s center of gravity over time. As depicted in Fig. 1, when the chair user sways back and forth, the center-of-gravity trajectory shifts along the front–back axis, whereas when the user sways from side to side, the trajectory shifts along the left–right axis. When the user sits still, the trajectory exhibits only minor fluctuations over time. An empty chair is associated with minimal movement in the center-of-gravity trajectory. In this study, to quantify participants’ movements while seated in the chair, the ATL was calculated. ATL represents the mean distance of body movement over time, which was derived from continuous load cell recordings. A higher ATL value indicates greater postural movement or restlessness, whereas a lower ATL value indicates stable sitting behavior and less hyperactivity. We used ATL as an objective indicator of hyperactivity and compared measured values before and after pharmacological treatment to evaluate changes in motor activity levels. The ATL was calculated over a specified sequence of frames reflecting the average spatial movement within a defined time window as follows:

7

where N is the number of data points within the specified time window.

To evaluate the seated movements of participants with ADHD over the full 15-min recording session, two types of ATL analyses were performed. First, a global ATL was calculated throughout the full session according to Eq. (7), which provided a summary measure of overall movement magnitude. Second, to assess temporal variations in movement, the session was divided into 15 nonoverlapping 1-min intervals. The ATL was then calculated for each interval on the basis of the same definition. This calculation yielded the ATL change sequence A, defined as follows:

8

Each corresponds to the ATL computed from the th 1-min interval, which contains samples acquired at a sampling rate of 5 Hz, to capture average movement behavior during those 60 s of the session.

To capture temporal movement patterns, we analyzed the slopes of the ATL change sequence through a nonoverlapping sliding window approach. Each window of size is defined as follows:

where

9

denotes the number of complete nonoverlapping sliding windows, and denotes the window size. Each window represents a temporal segment of ATL values.

Within each window, a linear regression was performed using the least squares method to compute the slope of the best-fit line ). The resulting slope sequence () was defined as follows:

where

10

where represents the time index within , and represents the corresponding ATL value. A positive slope indicates a progressive increase in movement over time, whereas a negative slope reflects decreasing activity or behavioral stabilization.

To evaluate treatment effects, we compared pretreatment and posttreatment values to calculate the percentage reduction in ATL, , as follows:

11

where and denote the ATL values measured before and after treatment, respectively.

Percentage changes in parent- and teacher-reported SNAP-IV scores were likewise calculated. The percentage reduction, defined as , was calculated as follows:

12

where and denote the total parent- or teacher-reported SNAP-IV scores before and after treatment, respectively. These percentage reduction indices provide normalized measures of improvement, enabling direct comparison between participants with different symptom severity levels at baseline.

Statistical analysis

All statistical analyses were performed using SAS (version 9.3; SAS Institute, Cary, NC, USA). Descriptive statistics are presented as mean ± standard deviation values. Paired t tests were performed to evaluate treatment effects. Associations between the objective ATL metrics and subjective SNAP-IV scores were determined using Spearman’s rank correlation coefficients. To evaluate the magnitude of ATL differences, Cohen’s d was calculated as a measure of effect size. A p value of < 0.05 was considered statistically significant.

Results

The mean age was 6.9 ± 1.1 years in the Ritalin IR group and 8.75 ± 1.75 years in the ORADUR group. In total, 11, 1, and 44 patients had ADHD-I, ADHD-H, and ADHD-C, respectively, indicating that hyperactivity-related symptoms predominated in the cohort. In the Ritalin IR group (n = 37), parent-reported SNAP-IV scores decreased by 25.26% from 37.75 ± 14.24 to 28.08 ± 16.94 (p < 0.0001). Teacher-reported SNAP-IV scores similarly decreased by 40.12% from 40.68 ± 17.66 to 24.36 ± 18.87 (p < 0.0001; Table 1). In the ORADUR-Methylphenidate group (n = 19), parent-reported SNAP-IV scores decreased by 24.42% from 40.75 ± 16.44 to 29.17 ± 17.80 (p = 0.0053), and teacher-reported scores decreased by 40.22% from 42.91 ± 18.04 to 25.64 ± 17.44 (p = 0.0082; Table 2).

Table 1.

Average trajectory length and SNAP scores before and after Ritalin 10 mg (N = 37)

	Before	After	Mean diff	Std diff	df	t-value	P value	Cohen’s d	Hedges’ g
Average trajectory length	0.0318 ± 0.0191	0.0180 ± 0.0126	−0.0138	0.018	36.000	−4.710	< 0.001*	−0.774	−0.758
SNAP total score (P)	37.75 ± 14.24	28.08 ± 16.94	−11.3333	11.832	26.000	−4.977	< 0.001*	−0.958	−0.930
Inattention score (P)	15.33 ± 5.92	11.11 ± 5.24	−4.2222	4.635	26.000	−4.733	< 0.001*	−0.911	−0.884
Hyperactivity-impulsivity score (P)	14.04 ± 6.48	10.41 ± 7.48	−3.6296	4.869	26.000	−3.874	0.001*	−0.746	−0.724
Oppositional score (P)	11.33 ± 6.91	7.85 ± 8.01	−3.4815	5.760	26.000	−3.140	0.004*	−0.604	−0.587
SNAP total score (T)	40.68 ± 17.66	24.36 ± 18.87	−16.3200	17.373	24.000	−4.697	< 0.001*	−0.939	−0.910
Inattention score (T)	17.84 ± 6.50	11.28 ± 6.74	−6.5600	7.725	24.000	−4.246	< 0.001*	−0.849	−0.822
Hyperactivity-impulsivity score (T)	14.80 ± 8.39	8.32 ± 7.81	−6.0800	7.199	24.000	−4.223	< 0.001*	−0.845	−0.818
Oppositional score (T)	8.04 ± 7.16	4.76 ± 6.75	−3.2800	4.971	24.000	−3.299	0.003*	−0.660	−0.639

Note: P, parents; T, teacher; SNAP, Swanson, Nolan and Pelham Rating Scale; diff, difference; Std, standardized deviation; df, degree of freedom

p < 0.05

Table 2.

Average trajectory length and SNAP scores before and after Methydur SR 22 mg (N = 19)

	Before	After	Mean diff	Std diff	df	t-value	P value	Cohen’s d	Hedges’ g
Average trajectory length	0.0318 ± 0.0191	0.0180 ± 0.0126	−0.0127	0.020	18.000	−2.723	0.014	−0.625	−0.598
SNAP total score (P)	37.75 ± 14.24	28.08 ± 16.94	−10.8718	11.779	38.000	−5.764	< 0.001	−0.923	−0.905
Inattention score (P)	15.33 ± 5.92	11.11 ± 5.24	−4.2051	4.911	38.000	−5.348	< 0.001	−0.856	−0.839
Hyperactivity-impulsivity score (P)	14.04 ± 6.48	10.41 ± 7.48	−3.8205	4.936	38.000	−4.834	< 0.001	−0.774	−0.759
Oppositional score (P)	11.33 ± 6.91	7.85 ± 8.01	−2.9211	5.334	37.000	−3.376	0.002	−0.548	−0.536
SNAP total score (T)	40.68 ± 17.66	24.36 ± 18.87	−16.2778	17.209	35.000	−5.675	< 0.001	−0.946	−0.925
Inattention score (T)	17.84 ± 6.50	11.28 ± 6.74	−6.2778	7.041	35.000	−5.350	< 0.001	−0.892	−0.872
Hyperactivity-impulsivity score (T)	14.80 ± 8.39	8.32 ± 7.81	−6.1389	7.251	35.000	−5.080	< 0.001	−0.847	−0.828
Oppositional score (T)	8.04 ± 7.16	4.76 ± 6.75	−3.5833	5.628	35.000	−3.820	0.001	−0.637	−0.623

Note: P, parents; T, teacher; SNAP, Swanson, Nolan and Pelham Rating Scale; diff, difference; Std, standardized deviation; df, degree of freedom

p < 0.05

ATL, calculated as an objective behavioral indicator of seated movement, also decreased significantly after treatment. In the Ritalin IR group, ATL values decreased after treatment by 43.40% from 0.0318 ± 0.0191 to 0.0180 ± 0.0126 (p < 0.0001; Table 1). Similarly, in the ORADUR group, ATL values decreased by 29.47% from 0.0431 ± 0.0277 to 0.0304 ± 0.0258 (p = 0.0140; Table 2). Pooled data from both treatment groups revealed a large effect size for ATL reduction (Cohen’s d = 0.8739), indicating that methylphenidate-based interventions exerted a consistently large effect on hyperactivity-related movement, as indicated by ATL values.

Figure 2 depicts the movement trajectory of a representative participant before and after treatment. The participant’s fidgeting behavior markedly reduced after pharmacological intervention. Posttreatment ATL values remained relatively stable throughout the 15-min video session, whereas pretreatment ATL values revealed a progressive increase in movement, particularly during the latter part of the session. To further explore these temporal patterns, we analyzed the slope of the ATL sequence over time. The slope analysis (Fig. 3) revealed a significant difference between pretreatment and posttreatment sessions, particularly during the final minutes of the session, indicating improved movement stability after medication. To complement these findings, histograms were generated to visualize the distribution of ATL values across both local temporal segments and the overall global measure (Fig. 4). ATL values significantly decreased across all 5-min intervals (0–5, 5–10, and 10–15 min) and in the overall global comparison between pretreatment and posttreatment recordings. In addition, Fig. 5 demonstrated the raw ATL data to illustrate the progression between the pretreatment and posttreatment sessions across different time intervals.

Fig. 2.

Trajectory of the center of gravity. Movement signals were measured before (blue) and after (orange) treatment in one patient with ADHD. Before treatment, the trajectory is dispersed. After treatment, the trajectory converges within a smaller area

Fig. 3.

Comparison of temporal movement patterns before and after treatment by using slope-based analysis. The slope remains stable after treatment at different time intervals. By contrast, the slope increases significantly toward the end of the video session before treatment. (a) 2-minute, (b) 3-minute, (c) 4-minute, and (d) 5-minute time interval analyses

Fig. 4.

Distribution of atl values across local and global segments shown by histogram. In both analyses, atl values significantly decreased after treatment, indicating reduced hyperactive movements and improved behavioral stability. (a) atl values aggregated in 5-minute intervals (0–5, 5–10, and 10–15 min) before and after treatment. (b) Global atl values comparing pre- and post-treatment sessions

Fig. 5.

Comparison of temporal movement patterns before and after treatment using raw atl. atl values significantly decreased after treatment across different time intervals: (a) 2-minute, (b) 3-minute, (c) 4-minute, and (d) 5-minute analyses

To identify the correlation between subjective symptom ratings and objective movement data, Spearman’s rank correlation analyses were performed. No significant correlation was observed between the percentage reduction in ATL and that in parent- or teacher-reported SNAP-IV scores (Fig. 6). This finding suggests that the ATL reflects distinct aspects of hyperactivity that questionnaire-based assessments cannot fully capture.

Fig. 6.

Comparison of relationship between percentage reduction in atl and percentage change in SNAP-IV scores. Scatter plots show that no significant correlations were observed in either analysis, suggesting that atl reduction reflects aspects of hyperactivity not fully captured by questionnaire-based ratings. (a) atl reduction and parent-reported SNAP-IV. (b) atl reduction and teacher-reported SNAP-IV

Discussion

In children with ADHD, treatment with Ritalin IR or ORADUR-Methylphenidate significantly reduced the ATL, an objective measure of seated movement. Corresponding reductions in parent- and teacher-reported SNAP-IV scores were also observed after 1 month of treatment. These results support the utility of the ATL as an objective and quantifiable marker of treatment efficacy in children with a predominantly hyperactive ADHD presentation, which was measured in this study through load cell–equipped smart chairs.

Although the SNAP-IV scale is among the most widely used tools for ADHD assessment and has been validated across cultures and settings [33, 36, 37], concerns remain regarding interrater variability, particularly between parent and teacher ratings [18]. A study indicated that although parent ratings are valuable in research settings, they are less reliable for clinical diagnosis. By contrast, teacher assessments tend to offer greater clinical utility, particularly regarding hyperactivity and impulsivity assessment [37]. Discrepancies between parent and teacher ratings can create diagnostic ambiguity, underscoring the value of complementary objective assessments. The reliability of the SNAP-IV scale may be further compromised when raters lack sufficient observation time. The scale depends heavily on subjective evaluations by parents and teachers, who must observe the child consistently throughout a set assessment period. However, in practice, variability in the duration and context of these observations because of factors such as the brief and sporadic nature of classroom interactions may affect the accuracy of ratings. Teachers may lack adequate opportunities to observe students in diverse situations, which leads to underreporting or misinterpretation of key symptoms, such as inattention and impulsivity [38]. These limitations highlight the need to integrate objective tools, such as motion-tracking or load cell systems, into subjective assessments to improve diagnostic precision.

In our previous study involving a smaller sample (N = 13), we demonstrated the feasibility and sensitivity of a load cell system combined with a SNAP-IV-based assessment for the evaluation of treatment effects. We observed substantial improvements in both ATL and hyperactivity scores after Ritalin IR treatment [29]. In the present study, we extended this analysis by including the 13 participants from our previous pilot study in a cohort of children who received Ritalin IR (total = 37) and enrolling an additional 19 children who received ORADUR-Methylphenidate. This expanded sample allowed us to more comprehensively evaluate the effects of different stimulant formulations on patient movements, which were quantified on the basis of ATL values obtained using load cells. Consistent with our pilot findings, ATL values significantly decreased after stimulant administration, reflecting improvements in motor regulation. The consistency of ATL reduction between both immediate-release and extended-release methylphenidate formulations supports the robustness of this parameter as an objective biomarker of treatment response in ADHD. These results also provide a complimentary layer of quantitative evidence, supporting findings from the SNAP-IV and other behavioral rating scales that medication exerts measurable effects on hyperactivity-related movement. Nevertheless, although the larger sample than that of our pilot study strengthens the generalizability of our findings, future longitudinal studies involving even larger cohorts are required to confirm the clinical applicability of ATL as a reliable indicator of the effectiveness of personalized ADHD treatment.

Unlike behavioral rating scales, which are inherently subjective and thus influenced by parent or teacher perceptions, the ATL is derived from direct biomechanical data, which we captured by embedding load cells in a classroom chair. This metric reflects the spatial magnitude of postural movements over time, serving as a proxy for psychomotor agitation and hyperactive behavior. The ATL demonstrated high sensitivity even within the smaller cohort, in which it decreased significantly (p = 0.0055), although scores for several SNAP-IV domains did reach statistical significance because of the limited sample size. In the larger cohort, ATL reductions remained highly significant (p < 0.0001), further validating the potential of this metric as a treatment-responsive parameter. The ATL captured subtle behavioral changes in our sample that aligned closely with improvements in teacher-rated hyperactivity scores, reinforcing its ecological validity in real-world classroom-like settings. Thus, the ATL can be integrated into ADHD assessment protocols as an objective measure that complements subjective scales. This measure provides a continuous, high-resolution behavioral signal that may enable earlier or more precise detection of treatment effects than do questionnaire-based assessments alone. As a noninvasive and easy to implement approach, the ATL represents a promising tool for longitudinal monitoring in both clinical trials and educational settings.

In this study, no significant correlation was identified between reductions in parent- and teacher-reported SNAP-IV scores and those in the ATL, as measured by the load cell system. This discrepancy may stem from differences in the constructs under assessment. SNAP-IV scores reflect subjective perceptions of inattention, hyperactivity/impulsivity, and oppositionality in daily life. We used the ATL to quantify subtle postural instability and restlessness in a simulated classroom setting. Both measures therefore capture related but distinct dimensions of ADHD. Moreover, parental and teacher ratings are influenced by contextual factors, rater expectations, and observer bias, resulting in moderate interrater reliability in prior studies [31]. By contrast, the ATL provides a continuous and objective signal with high temporal resolution, but its scope is limited to short-term motor activity. Different measurement timescales may also contribute to the observed discrepancy: the ATL reflects the acute motor effects of medication during a 15-min task, whereas SNAP-IV scores summarize behaviors observed over multiple weeks. Furthermore, ADHD medication treats symptom domains beyond motor restlessness—for example, attention and impulse control, which may be prominent to parents and teachers but comparatively less detectable through load cell–based monitoring. Finally, heterogeneity in ADHD presentations and sample size limitations may have further reduced the power to detect correlations. Collectively, these proposed reasons for the lack of significant association between subjective and objective symptom measures suggest that these treatment response indices should be regarded as complementary rather than interchangeable.

Several sensors have been used to evaluate the effects of ADHD medications. Smartwatch and actigraphy devices worn on the wrist or ankle allow monitoring at home or school rather than in a laboratory [39]. However, these devices must be attached to the patient’s body to function, which limits their ecological validity. Additionally, these sensors only record the movement of body parts to which they have been attached. Smartwatches capture patient movements during normal daily activities and have limited battery life [40, 41], whereas actigraphy devices are used to study patients’ sleep efficiency and are limited by a low sampling rate [42, 43]. Electrooculography targeting oculomotor control, which is closely tied to attention, holds promise for objective ADHD markers and biofeedback [44], but this approach typically requires electrode use within controlled tasks and is vulnerable to motion and environmental interference artifacts. Furthermore, evidence for its use in medication response monitoring remains limited.

In our previous study, we embedded piezoelectric sensors in the seat cushions of consulting chairs to assess participants’ movement patterns in a clinical setting. The study revealed substantial reductions in movement-related features, including the signal variance, zero-crossing rate, and high energy ratio, among 22 children with ADHD after 1 month of methylphenidate treatment [41]. Although both piezoelectric sensors and load cells can detect physical activity, their contextual suitability and technological characteristics differ. Our piezoelectric sensor system was designed for use in rotatable consulting chairs and captured movement during physician–patient interactions. By contrast, the load cell system was embedded in a four-legged chair placed in a simulated classroom environment, enabling us to track physical activity within structured, classroom-like conditions. Load cells provide several advantages over piezoelectric sensors. For example, load cells exhibit relatively low susceptibility to electromagnetic interference and high stability in detecting static postures, given their direct current (DC) response capabilities. By contrast, piezoelectric sensors are more susceptible to electromagnetic noise and monitor stationary behaviors less effectively because of their limited DC response. In the present study, we introduced the ATL as an alternative primary metric of interest to signal variance or the zero-crossing rate. The ATL was selected for its practical advantages and compatibility with both the hardware and software of the load cell system. Given that load cell outputs contain a DC component associated with posture, traditional signal variance measures may cause these components to be misinterpreted as noise. Conversely, the ATL captures meaningful movement dynamics while maintaining sensitivity to posture-related data.

We chose a 15-min task to evaluate posttreatment changes in seat stability, as measured using the ATL, in children with ADHD. This duration was considered appropriate for both ecological validity and experimental control. The 15-min video simulated the attentional demands of a typical classroom session within a manageable observation period. Our clinical observations indicate that longer sessions tend to increase the likelihood of participants leaving their seats or standing up, thereby reducing the validity of ATL measurements. By contrast, shorter sessions allow most participants to remain seated throughout the task. Therefore, the 15-min duration provides an optimal balance between ecological relevance and measurement reliability for objective assessment of attention in individuals with ADHD. In our study, participants rarely stood up or left the chair. All such events that transpired were clearly identifiable in the load cell recordings as abrupt drops in pressure across all sensors. These segments were excluded from the analysis to ensure that the calculated ATL values reflected seated movements only. Therefore, the ATL represents continuous in-seat movement rather than gross motor actions, such as standing or leaving the chair.

This study has several strengths. To our knowledge, this is the first study in which a smart chair system incorporating load cells was used to objectively evaluate the therapeutic effects of both immediate- and extended-release methylphenidate formulations in children with ADHD. The system is unobtrusive and easy to operate; thus, it minimizes participant burden and mitigates behavioral distortion due to monitoring awareness. The use of a simulated classroom environment further strengthened ecological validity by providing a realistic setting for the evaluation of behavioral responses.

This study also has several limitations. First, most participants had ADHD-C, which may limit the generalizability of our findings to other ADHD subtypes, such as ADHD-I or ADHD-H. Future studies should aim for a more balanced distribution of ADHD subtypes to investigate subtype-specific treatment effects. Second, our study did not include a nontreatment control group. Future research involving a control group can improve causal inference and reproducibility. Third, unmeasured factors, such as sleep quality, emotional state, and recent dietary intake, may have affected participants’ movement behaviors during assessment. We thus recommend that scholars develop and apply a standardized questionnaire to account for these potential confounders in future protocols. Finally, each participant was assessed only twice (pretreatment and posttreatment), which complicated efforts to distinguish medication effects from potential habituation, day-to-day variability, or regression to the mean. In future studies, multiple baseline and follow-up assessments should be conducted to strengthen reliability and causal inference.

Conclusion

Hyperactivity is a core symptom of ADHD-H and ADHD-C. To improve hyperactivity assessment, we embedded load cells in a smart chair to objectively quantify seated movements in children with ADHD by using the ATL metric. ATL values significantly decreased after 1 month of methylphenidate treatment, indicating that treatment measurably improved hyperactive behaviors. These results highlight the potential of load cell–embedded smart chair systems as a practical and objective tool for evaluating treatment response in children with ADHD, particularly those exhibiting hyperactivity. Compared with traditional subjective rating scales, such as the SNAP-IV scale, the ATL provides continuous, real-time data reflecting behavioral changes in classrooms and other ecologically valid settings. This approach may complement current evaluation frameworks to increase the precision of treatment response monitoring in both clinical and educational contexts.

Abbreviations

ADHD

Attention-deficit/hyperactivity disorder

ATL

Average trajectory length

BG

Basal ganglia

SNAP-IV scale

Swanson, Nolan, and Pelham-IV rating scale

Author contributions

CO and LL: conceptualization. RY, RW, and LL: study design and methodology. CO, RY, and LL: manuscript writing, review, and editing. CO and YC: quantitative analysis. CO: original draft preparation. CO and LL: funding acquisition. All authors have read and agreed to the submitted version of the manuscript.

Funding

This study was supported by Kaohsiung Medical University Hospital (grant numbers: KMUH113-3R43, KMUH-SI11308, and KMUH-SA11304), a cooperation project between Kaohsiung Medical University and National Kaohsiung University of Science and Technology (grant number: 114KK019), and Taiwan’s National Science and Technology Council (grant numbers: NSTC 111-2314-B-037077, NSTC113-2314-B-037-095, NSTC 111-2221-E-992-090-MY3, and NSTC 112-2221-E-214-002-MY2).

Data availability

No data sets were generated or analyzed in this study.

Declarations

Ethics approval and consent to participate

All experimental procedures adhered to the Declaration of Helsinki. Informed consent regarding participation and data publication was obtained from a parent or legal guardian for all participants. The study protocol was approved by the Institutional Review Board of Kaohsiung Medical University Hospital (approval number: KMUIRB-SV(I)-20190060).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.