Quantitative evaluation of human cerebellum-dependent motor learning through prism adaptation of hand-reaching movement

Abstract

The cerebellum plays important roles in motor coordination and learning. However, motor learning has not been quantitatively evaluated clinically. It thus remains unclear how motor learning is influenced by cerebellar diseases or aging, and is related with incoordination. Here, we present a new application for testing human cerebellum-dependent motor learning using prism adaptation. In our paradigm, the participant wearing prism-equipped goggles touches their index finger to the target presented on a touchscreen in every trial. The whole test consisted of three consecutive sessions: (1) 50 trials with normal vision (BASELINE), (2) 100 trials wearing the prism that shifts the visual field 25° rightward (PRISM), and (3) 50 trials without the prism (REMOVAL). In healthy subjects, the prism-induced finger-touch error, i.e., the distance between touch and target positions, was decreased gradually by motor learning through repetition of trials. We found that such motor learning could be quantified using the "adaptability index (AI)", which was calculated by multiplying each probability of [acquisition in the last 10 trials of PRISM], [retention in the initial five trials of REMOVAL], and [extinction in the last 10 trials of REMOVAL]. The AI of cerebellar patients less than 70 years old (mean, 0.227; n = 62) was lower than that of age-matched healthy subjects (0.867, n = 21; p < 0.0001). While AI did not correlate with the magnitude of dysmetria in ataxic patients, it declined in parallel with disease progression, suggesting a close correlation between the impaired cerebellar motor leaning and the dysmetria. Furthermore, AI decreased with aging in the healthy subjects over 70 years old compared with that in the healthy subjects less than 70 years old. We suggest that our paradigm of prism adaptation may allow us to quantitatively assess cerebellar motor learning in both normal and diseased conditions.

Fig 1. Scheme for prism adaptation of hand-reaching.

(A) The experimental apparatus consists of a sensor on the participant’s right earlobe, goggles equipped with an electrically controlled shutter with a plastic or Fresnel prism plate, a touchscreen, and two computers. (B) Time sequence of a single trial shown from left to right. Every trial starts from the touch of a participant’s index finger at the sensor on the right earlobe. As soon as the participant releases their index finger from the sensor, vision is blocked by the shutter (MOVEMENT TIME). Immediately after reaching the touchscreen (TOUCH), the goggles become transparent, and the participant recognizes how their index finger deviated/hit the target for 100 ms (EXPOSURE). Subsequently, the target disappears (TARGET OFF) and the participant returns their index finger to the original position in preparation for the next trial.

Fig 2. Adaptation curves in different subjects and in healthy and patient groups.

(A)–(D) Adaptation curves in HN13 (A), CN4 (B), CN3 (C), and CN15 (D). The ordinate shows the finger-touch error represented by the distance (mm) from the target to the touch point. Positive values indicate rightward shifts and negative values indicate leftward shifts. The abscissa shows the trial numbers. Best-fitted exponential curves (for details, see Materials and Methods) are overlaid on the raw data. Whereas a normal subject (HN13) shows typical adaptation (A), patients with cerebellar diseases show three different patterns of impaired adaptation (B)–(D). (E) and (F) Average adaptation curves for 21 HN subjects (E) and 62 CN patients (F). Thick and thin curves show mean and mean ± 2SD, respectively.

Fig 3. Variability of the finger-touch error in HN and CN groups.

(A) Variability of the finger-touch error in 21 HN subjects. Each dot shows SD of the finger-touch error for every trial throughout BASELINE, PRISM, and REMOVAL. An interpolated curve was drawn by fitting with nonlinear regression. (B) Same analysis as (A) in patients with cerebellar diseases below 70 years old (CN). (C) Comparison of intertrial variability of the finger-touch error shown in A and B between the initial 20 and the last 20 trials. The ordinate shows the ratio (e/d) of the mean SD for the last 20 trials (e) in PRISM to that for the initial 20 trials (d) of PRISM. Note that the intertrial variability markedly decreased following adaptation in the HN group, but not in the CN group. **** p < 0.0001 by Mann-Whitney U-test. Error bar represents SEM.

Fig 4. Quantitative evaluation of prism adaptation.

(A) An example of adaptation in a healthy subject (HN13) shown in Fig. 2A. The finger-touch error of the last 10 trials of PRISM, and that of the initial five and last 10 trials of REMOVAL are extracted from Fig. 2A. Acquisition, retention, and extinction of adaptation were estimated from the probability of success (a) in the last 10 trials of PRISM (10/10), the probability of failure (b) in the initial five trials of REMOVAL (5/5), and the probability of success (c) in the last 10 trials of REMOVAL (10/10), respectively. AI was calculated as a × b × c and 1 in this case. (B) Similar analysis in CN4 shown in Fig. 2B. a = 1/10, b = 1/5, c = 6/10. AI = (1/10) × (1/5) × (6/10) = 0.012. Horizontally shaded areas in (A) and (B) represent the zone of “correct” touch (within ± 25mm). (C)–(F) Frequency distributions of a (C), b (D), c (E), and AI (F). Insets represent cumulative frequency curves. F(x) represents normal cumulative distribution function. (G) Frequency distribution of the time constant τ (for details, see Materials and Methods). Insets represent cumulative frequency curves of τ. Red columns and lines in (C)–(G) show data for 21 HN subjects. Blue columns and lines in (C)–(G) show data for 62 CN patients. (H) Receiver operating characteristic (ROC) curve analysis in the HN and CN groups. A purple line shows ROC curve for AI, a red line for the probability of acquisition, a blue line for the probability of retention, a green line for the probability of extinction, and a black line for τ.

Fig 5. Relationship between AI and incoordination.

(A) AI and the magnitude of dysmetria represented by the SD of the finger-touch error in BASELINE. Data were obtained from 62 CN patients (blue dots) and HN subjects (red dots). Each point represents data obtained from one subject. (B) and (C) Tracking AI (B) and SARA (C) data of each MSA patient (CN56, CN57, CN59, CN60, and CN61; n = 5).

Fig 6. AI of healthy subjects (HN and HE) and cerebellar patients (CN and CE).

(A) Distribution of AIs and ages for all the participants analyzed. AI tended to decrease and showed a widespread distribution in the HE group. Cerebellar patients (CN and CE) showed lower AIs than the age-matched healthy subjects (HN and HE). † indicates four pure parkinsonian MSA patients without clinical cerebellar signs. (B) Comparison of AI among the HN, HE, CN and CE groups. In all panels, red circles and columns represent HN; magenta, HE; blue, CN; and green, CE. **p < 0.01, ****p < 0.0001, Kruskal-Wallis test or Steel-Dwass test. Error bar represents SEM.

Fig 7. AI and other clinical indexes in various cerebellar diseases.

(A)–(C) Scatter plots comparing AI with SARA score (A), 9-Hole Peg Test (B), and disease duration (C) in CN and CE patients. Linear regression lines are overlaid. (D) Comparison of AI between the CBL (n = 24) and CBL+ (n = 32) groups. *p < 0.05 by Mann-Whitney U-test. Error bar represents SEM. (E) AI was significantly higher in pure parkinsonian MSA patients than in SCA6, SCA31, CCA, or MSA (MSA-C and MSA-P) patients. *p < 0.05, post hoc Steel-Dwass test.