Assessing rhythm reproducibility during gait after internal rhythm-learning tasks: a pilot study in healthy adults and patients with Parkinson’s disease

Abstract

[Purpose] To investigate the feasibility of a new rhythm-learning task using internal cues in healthy individuals and patients with Parkinson’s disease. [Participants and Methods] This study included 27 healthy individuals and six patients with Parkinson’s disease. All participants first learned rhythms through either stepping or tapping tasks. Subsequently, they attempted to reproduce the learned rhythms while walking. The relative rhythm reproducibility error was calculated by comparing the target rhythm with the rhythm generated during walking. Experiment 1 examined the effects of task type and rhythm condition in healthy individuals, whereas Experiment 2 focused on the effects of stepping-based rhythm-learning task in patients with Parkinson’s disease. [Results] In healthy participants, rhythm reproducibility errors differed significantly depending on task type and rhythm condition. Stepping produced more accurate walking rhythms than tapping. Patients with Parkinson’s disease who had lower scores on the Frontal Assessment Battery and longer completion times on the Trail Making Test Part B tended to show reduced rhythm reproducibility. [Conclusions] This rhythm-learning task using internal cues could be applied to individuals with Parkinson’s disease. The findings suggest a possible association between the ability to learn rhythmic patterns internally and frontal lobe cognitive function.

INTRODUCTION

In humans, rhythmic movement is triggered by external stimuli or spontaneously initiated by internal cues¹⁾. Externally guided movements are centered around the prefrontal cortex, premotor cortex, parietal lobe, and cerebellum^{1, 2)}. In contrast, internally initiated movements are centered around the prefrontal cortex, basal ganglia, and supplementary motor cortex^{1, 3)}.

Strategies using these two pathways are adapted in patients with Parkinson’s disease (PD). External cues to induce walking in patients with PD include auditory stimuli, such as metronome sounds, and visual stimuli, such as a line drawn on the floor to be crossed⁴⁾. These external cues are compensatory when there are deficiencies within the internally guided system, which is influenced by frontal lobe function⁵⁾.

Although the effectiveness of external cues has been established, less is known about internal cueing strategies. The traditional internal cueing paradigm is referred to as the synchronous continuation task (SCT). Participants start by tapping their fingers in a series of equally spaced auditory beats, known as the synchronization phase (SP)^6‒9). After a period of synchronous tapping, the auditory beat stimulus is removed, and participants are asked to continue tapping at the same rhythmic beat for a specific period (continuation phase [CP]). This type of internal cueing activates the internally guided system.

A novel internal cueing method has recently been implemented whereby participants sing a song verse to music. The music is then stopped, and they are asked to sing while walking in time to their own voice¹⁰⁾. This internal cueing method improved gait characteristics, including cadence, the coefficient of variation of stride length, stride time, and single support time, compared to external cueing in healthy young adults, older adults, and patients with PD. However, this method involves walking while singing a nursery rhyme, which might be impractical in routine situations. Additionally, Harrison’s study¹⁰⁾ found significant differences in movement variability, suggesting that stepping, tapping, and walking might not be completely uniform mechanisms. Although Harrison’s method was shown to be effective, internal cueing methods have not been fully explored for patients with PD with impaired internally initiated motor pathways.

Indicators of the effectiveness of cueing techniques include the Freezing of Gait Questionnaire (FOGQ) score, walking speed, and gait rhythm variability¹¹⁾. In addition, gait rhythm variability is an important indicator of the risk of falls in older adults¹²⁾. However, the extent to which internal cueing methods are feasible, depending on the severity of PD and older adults, remains unclear. Therefore, it is important to understand the extent to which patients with PD and older adults are trained using internal cueing. This understanding helps to increase the variety of internal cue intervention methods.

Rhythm tasks using the upper extremities include finger tapping¹³⁾ and flexion/extension movements of the elbow and wrist^{8, 14)}. Those using the lower extremities include stepping¹⁵⁾ and ankle joint tapping movements¹³⁾. The use of internal rather than external cueing for walking, tapping, and stepping movements results in less movement variability¹⁶⁾. From a motor learning perspective¹⁷⁾, different tasks involving the upper and lower limbs may also affect gait rhythmicity. However, it remains unclear whether internal rhythm learning using the upper or lower extremities is more transferable to gait rhythm.

Therefore, we investigated a novel method of internal rhythm learning in healthy adults (Experiment 1). Specifically, we assessed whether using the upper or lower limbs was more conducive to generalized walking tasks implemented in healthy older and younger adults. Furthermore, we then assessed the feasibility of the method applied in Experiment 1 to patients with PD (Experiment 2).

PARTICIPANTS AND METHODS

Experiment 1 involved 27 participants, including 14 young adults (age, 20.7 ± 0.5 years), 13 older adults (age, 66.2 ± 2.0 years), and seven females. The average Mini-Mental State Examination (MMSE) score was 29.8 ± 0.4. None had neurological or orthopedic concerns, and all were able to understand and execute the tasks. All participants provided written informed consent.

At the outset, the participants completed the Trail Making Tests A and B (TMT-A, B) to evaluate frontal lobe function, specifically attention. Test B posed a greater cognitive demand than Test A. The baseline walking rhythm of the participants was also determined twice using the 10-meter walking test (10MWT). All participants were evaluated under conditions that were 10% faster and 10% slower compared with the average walking rhythm (bpm) calculated based on the 10MWT tests (control).

We applied tapping and stepping tasks to assess the impact of upper limb (tapping on the desk surface) and lower limb (stepping on the floor) rhythm learning on gait rhythmicity (Fig. 1). Both tasks followed the synchronization–continuation paradigm. Participants sat in a quiet room and attempted to synchronize their tapping or stepping to the beat of an MA-1 electronic metronome (Korg Inc., Tokyo, Japan) for 10 s (SP), followed by 20 s without the metronome beat (CP). Each rhythm–learning task included a series of 30-s practice sessions, repeated four times.

Fig. 1.

Flowchart of the study.

10MWT: 10-meter walking test; slow (bpm): walking 10% slower than the control; fast (bpm): walking 10% faster than the control.

We randomized the order of the tapping and stepping tasks, considering that it may influence the results. The participants performed the tapping and stepping tasks on different days, with intervals of at least 2 days.

We captured the timing of limb movements by attaching an MVP-RF8-HC-2000 accelerometer (Micro Stone Inc., Nagano, Japan) to the ankle (for walking and stepping) and to the wrist (for tapping). Acceleration of the lower limbs, feet, and arms while walking, stepping, and tapping was assessed at a frequency of 200 Hz.

Acceleration data were processed to derive composite acceleration vectors (x, y, z) and identify key events, such as foot–ground and hand–desk contact. Rhythms (bpm) while walking and during the rhythm-learning tasks were computed based on peak-to-peak intervals of the composite acceleration vectors (Fig. 2). Given that each rhythm condition was applied twice, we assessed gait reproducibility using results from four 10MWT trials. We calculated the Relative Rhythm Reproducibility Error % (RRE%) during walking relative to the presented rhythm using the following equation:

$$\mathrm{RRE}\%=\left(\mathrm{measured} \mathrm{bpm} / \mathrm{presented} \mathrm{bpm}\times 100\right)-100$$

Fig. 2.

Examples and accelerometry while walking.

The red and blue lines indicate the left and right feet, respectively. Red and blue circles indicate peak times of left and right feet, respectively. Rhythm (bpm) while walking and during each rhythm–learning task was calculated by excluding the first and last two peak times, then calculating the time between the right and left peaks, from which average bpm per interval was derived.

Positive and negative values indicated faster and slower rhythms relative to the presented rhythm, respectively. We distinguished between different rhythm-learning tasks using a repeated two-way analysis of variance (ANOVA) of the RRE% during gait (factors: task and rhythm conditions). We applied the Bonferroni correction for multiple comparisons. Data were statistically analyzed using IBM SPSS 25 (IBM Corp., Armonk, NY, USA). Values with p<0.05 were considered statistically significant. The same statistical methods were applied to compare differences in rhythm reproducibility during the CP to confirm that each rhythm-learning task was appropriately executed without the metronome beat.

Experiment 2 included six patients with PD (female participants, n=2; Hoehn and Yahr disability scale [H–Y], 3¹⁸⁾: age, 74.7 ± 4.6 years; MMSE: 27.8 ± 3.0; Frontal Assessment Battery [FAB]¹⁹⁾: 14 ± 2.2; Unified Parkinson’s Disease Rating Scale [UPDRS] [Part III]: 28.2 ± 11.4; FOGQ¹⁴⁾: 12.7 ± 11.4). All participants provided written informed consent to participate in the experiment, which proceeded for at least 1 h after the administration of PD medication. Table 1 presents the clinical information of the six participants.

Table 1. Clinical data of the patients.

Number	Sex	Hoehn–Yahr	Age (years)	MMSE	FAB	UPDRS	TMT-A (s)	TMT-B (s)	FOGQ
1	F	3	68	25	15	19	71	86	8
2	F	3	71	30	12	30	69	565	10
3	M	3	75	30	15	20	65	97	12
4	M	3	80	29	14	50	56	229	20
5	M	3	79	23	11	24	97	234	18
6	M	3	75	30	17	26	42	72	8

F: female; M: male; MMSE: mini-mental state examination; FAB: frontal assessment battery; UPDRS: unified Parkinson’s disease rating scale; TMT: trail making test; FOGQ: freezing of gait questionnaire.

Based on the findings from Experiment 1, stepping was selected as the rhythm-learning task for Experiment 2. The experimental procedure and RRE% calculations were performed as described in Experiment 1. Given that the RRE% during walking after the stepping task in the control condition was low in Experiment 1, we introduced 10% faster (fast) and slower (slow) rhythms to reduce the patient burden. Each rhythmic condition was measured once; the experiment was conducted in a clinical rehabilitation setting for approximately 20 min. The Research Ethics Committee at Kawasaki University of Medical Welfare approved this study (approval ID: 16-070).

RESULTS

Table 2 lists the mean RRE% while walking after the tapping and stepping tasks performed in Experiment 1. The two-way ANOVA revealed significant main effects for task (F(1, 26)=5.60, p=0.03, ηp2=0.18) and rhythm (F(2, 52)=46.80, p<0.001, ηp2=0.64). Interactions did not significantly differ (task × rhythm: F(2, 52)=4.41, p=0.145, ηp2=0.15). Multiple comparisons showed that the tapping task generated significantly larger errors than the stepping task, specifically under the fast condition (F(1, 26)=16.78, p<0.001, ηp2=0.39). This indicated that the fast condition was more reproducible in the stepping than in the tapping task.

Table 2. Relative rhythm reproducibility error (RRE%) during walking under control, slow, and fast conditions after the tapping and stepping tasks in younger and older healthy participants.

Task	Control			Slow			Fast
	Younger	Older	Average	Younger	Older	Average	Younger	Older	Average
After tapping	–0.4 ± 1.8	–0.2 ± 2.4	–0.3 ± 2.1	3.4 ± 4.9	4.2 ± 4.5	3.8 ± 4.6	–4.7 ± 3.3	–4.1 ± 3.7	–4.4 ± 3.4
After stepping	–0.6 ± 4.0	2.1 ± 2.9	0.7 ± 3.7	3.8 ± 4.6	3.8 ± 3.3	3.8 ± 4.6	–3.7 ± 2.7	–0.45 ± 3.6	–2.1 ± 3.5

RRE: relative rhythm reproducibility error; Fast (bpm): walking 10% faster than the control; Slow (bpm): walking 10% slower than the control.

Table 3 lists the mean RRE% during the rhythm learning tasks performed in Experiment 1 during the CP. The two-way ANOVA revealed no significant effects for task (F(1, 26)=3.06, p=0.09, ηp2=0.11) or rhythm (F(2, 52)=1.79, p=0.177, ηp2=0.06). Interactions did not significantly differ (task × rhythm: F(2, 52)=2.80, p=0.07, ηp2=0.10). The results indicated similar reproducibility across the three rhythm conditions and no difference between the stepping and tapping tasks. The RRE% during the stepping task under slow and fast conditions during the SP was 0.6 ± 3.6 and 0.04 ± 1.6, respectively.

Table 3. Relative rhythm reproducibility error (RRE%) during rhythm learning tasks during the continuation phase in younger and older healthy participants.

Task	Control			Slow			Fast
	Younger	Older	Average	Younger	Older	Average	Younger	Older	Average
Tapping	–0.2 ± 1.3	0.1 ± 1.9	–0.1 ± 1.6	1.2 ± 5.8	–1.0 ± 2.2	0.1 ± 4.5	0.4 ± 1.4	1.0 ± 2.7	0.7 ± 2.1
Stepping	–0.3 ± 1.6	–1.1 ± 5.4	–0.7 ± 3.8	0.01 ± 5.8	–3.4 ± 7.0	–1.6 ± 6.5	0.6 ± 1.7	0.1 ± 3.4	0.4 ± 2.6

RRE: relative rhythm reproducibility error; Fast (bpm): walking 10% faster than the control; Slow (bpm): walking 10% slower than the control.

Table 4 summarizes the RRE% during the walking and stepping tasks in the CP and SP in each patient with PD under slow and fast conditions (Experiment 2). The mean RRE% in the patients while walking was 6.8 ± 6.9 and −3.4 ± 1.6 under the slow and fast conditions, respectively. The mean RRE% during the stepping task under slow and fast conditions during the SP was 5.1 ± 7.6 and 6.4 ± 9.6, respectively, and during the CP was 0.7 ± 1.0 and 2.7 ± 9.3, respectively.

Table 4. Relative rhythm reproducibility error while walking and stepping tasks in six patients with Parkinson’s disease.

Patient number	Slow				Fast
	Gait		Stepping task		Gait		Stepping task
	10MWT (s)	RRE%	RRE% in SP	RRE% in CP	10MWT (s)	RRE%	RRE% in SP	RRE% in CP
1	23.2	–2.1	–2.2	–0.2	19.6	–0.2	3.3	–3.2
2	17.7	15.9	14.5	2.2	17.2	–1.5	19.6	21.6
3	14.2	–0.2	–0.6	0.1	10.8	–4.7	3.5	–0.1
4	12.9	6.7	Failure	31.0	11.6	–5.9	–3.2	0.1
5	18.9	9.9	13.5	1.4	23.3	–2.7	16.7	–0.7
6	12.6	2.3	0.5	0.0	14.1	–2.4	–1.8	–1.2

10MWT: 10-meter walking test; CP: continuation phase; Fast (bpm): walking 10% faster than the control; RRE: relative rhythm reproducibility error; Slow (bpm): walking 10% slower than the control; SP: synchronization phase.

Patient 1 was the youngest (68 years) and had the lowest UPDRS score. The reproducibility did not differ from that observed in healthy participants. More time was required to complete the 10MWT under slow conditions compared to fast conditions (RRE%: −2.1 and −0.2, respectively). The RRE% of the patients was the most accurate under the fast condition.

Patient 2 had an FAB score of 12, which was below the cutoff. The patient also required the most time to complete the TMT-B, which was considered equivalent to failure. The 10WMT values under slow and fast conditions were similar. As indicated by the RRE% during the SP, the patient had difficulty matching movements with the external cues of the metronome during the stepping task.

Patient 3 had the lowest UPDRS score among the male patients but the third highest FOGQ score (12) among all patients. The RRE% was low under slow and fast conditions during stepping, indicating high reproducibility in rhythm learning. The RRE% was lower and higher while walking under the slow and fast conditions, respectively.

Patient 4 was the oldest (80 years) and had an FAB score of 14 and a TMT-B of >200 s. This patient had difficulty with the stepping task under the slow condition but perceived it as slower than under the fast condition. The RRE% was reduced by almost 6% under both conditions, indicating low rhythm replication while walking.

Patient 5 was the second oldest participant (79 years). This patient had an FAB score of 11 and a TMT-B of >200 s. The patient had difficulty in the stepping task under fast and slow conditions during the SP. However, the RRE% improved during the CP. The RRE% while walking was 9.9% under the slow condition, indicating that the patient was unable to change the rhythm from normal walking speed.

Patient 6 had FAB, TMT-B, and FOGQ scores of 17, 72, and 8, respectively. The RRE% during the stepping task was low; the patient functioned well, particularly under the slow condition. The RRE% while walking was higher than that during the stepping task and was almost the same as that in the healthy group.

DISCUSSION

The RRE% during the CP in Experiment 1 indicated that the participants adequately executed the rhythm-learning task and differentiated the three rhythm conditions. A significant effect of rhythm on gait was evident, suggesting that all three rhythm conditions reflected differences in gait. Furthermore, the ability to replicate the gait rhythm significantly differed among the tasks. Moreover, the stepping task was significantly more reproducible under fast conditions. This might have been due to the lower similarity between the tapping and walking tasks than that between the stepping and walking tasks. Rhythmic movements of the upper extremities are less likely to be transferred to the lower extremities¹⁶⁾. These findings suggest that the lower limbs are more suited to learning internal rhythms that should be transferred to walking. However, we used a two-way ANOVA in Experiment 1 instead of a three-way ANOVA owing to the small sample size. This limitation is acknowledged in the interpretation of the results.

In Experiment 2, all patients with PD safely executed the tasks without falls or other hazardous situations. Patients 1, 3, and 6 immediately adjusted their gait rhythms. Therefore, the internal cueing technique applied herein could be applied to patients with PD.

Patients with FAB scores of <15 and those who exceeded 200 s on the TMT-B (Patients 2, 4, and 5) had less rhythm reproducibility during gait. This implied that the research task paradigm could be related to frontal lobe function. Impaired working memory function might affect the ability to retain rhythmic memory²⁰⁾. The approximately 4-Hz rhythm may have also decreased in the mid-frontal region of the brain, an area associated with time estimation tasks²¹⁾. Theta and beta waves notably decrease and increase, respectively, in patients with PD and Freezing of Gait (FOG+) compared with those without (FOG–) and healthy individuals while executing lower-limb pedaling tasks²²⁾. This indicates that these brain wave activities are associated with motor and cognitive dysfunctions in patients with PD. The decreasing trend in rhythm reproducibility observed in our study was consistent with previous findings.

The reason for low reproducibility while walking was due to decreased reproducibility of the stepping task. A unique feature of the stepping task was the lower RRE% during the CP than during the SP. The RRE% of healthy participants showed fewer errors during the stepping task during the SP than during the CP, which is characteristic of patients with PD. The stepping task during the SP corresponded to an externally guided movement. This approach does not work when patients have impaired frontal lobe function because of increased load on the frontal lobe¹⁴⁾. One study included healthy individuals and patients with frontal lobe damage to execute a simple timing task: pressing a button in response to repeated sounds²³⁾. The results indicated that damage to the right frontal lobe specifically leads to abnormal variability in timing performance. The present findings also suggested that the stepping task during the SP was difficult for patients with significant frontal lobe dysfunction. In contrast, the RRE% was lower for the stepping task during the CP, indicating that internal cueing may confer less stress on the frontal lobe than external cueing, even when the task is loaded.

This study investigated the feasibility of an approach and evaluation using the SCT paradigm in patients with PD. Therefore, we did not compare the effects of internal and external cues on gait stabilization. This is because we primarily focused on an exploratory methodology, anticipating comparisons in future studies. The effectiveness of this method requires further investigation. We were unable to age-match between the healthy and patient groups, which precluded direct comparisons between them. Age-matching between healthy and patient groups in future studies might facilitate the separation of the effects of aging from those of PD. We verified the immediate reproducibility of rhythm during walking and stepping tasks, which should be distinguished from long-term learning results. Particularly, efforts should focus on improving the quality of walking through continuous application and examination of changes in effects as PD progresses.

In conclusion, we attempted to facilitate internal rhythm learning in patients with PD by employing the traditional SCT. Initially, it was unclear whether an internal cueing-based rhythm learning approach would be feasible for older adults and individuals with PD. However, the results of Experiments 1 and 2 showed that this new internal rhythm task was generally feasible for both populations. Furthermore, to determine whether rhythm learning using the upper or lower limbs would be more effective in transferring to gait rhythm, the findings from Experiment 1 suggested that lower-limb-based learning may be more beneficial. Additionally, Experiment 2 indicated that this internal rhythm task might be challenging to perform for patients with PD who present with significant frontal lobe dysfunction. Therefore, when applying this new learning task in clinical settings, it is important to assess patients’ cognitive functions in advance.

Funding

This work was supported by the JSPS KAKENHI under Grant number: 16K16474.

Conflict of interest

The authors have no competing interests to declare.