2.2.1. Before Exercise

Each participant was instructed to stand in a relaxed position. Surface electrodes were placed on the skin surface of superficial erector spinae muscles, 3-cm from the midline and parallel to the spinous processes of the apical vertebra. The level of the apical vertebra was determined on the radiographs by an experienced clinician. Then, the sEMG signals were recorded for 20 s during the relaxed standing position. The test was repeated three times, with at least 3 min of rest in between to minimize the effect of fatigue in participants.

#### 2.2.2. During Exercise

A physiotherapist specializing in the Schroth exercise instructed participants in performing the exercises. The sEMG data was collected in four different exercise positions, which were performed in a randomized order (Figure 1). Exercise 1 (E1) and exercise 2 (E2) were symmetric exercises, and exercise 3 (E3) and exercise 4 (E4) were asymmetric exercises. It is routine practice that each exercise is repeated three to five times in the clinic as specified in Lehnert-Schroth (2007) [12]. Thus, each test was repeated three times, aiming to minimize the influence of a daily training schedule and reflect the muscle performance in a real clinical situation. Furthermore, rest for at least 3 min in between was allowed to minimize the effects of fatigue in participants. Details of each exercise position are provided below.

E1 (on the fours): The participant kneeled down with the knees apart at shoulder width and kept the thighs in a vertical position. The arms were extended vertically under the shoulders to support the body, with the fingers pointing straight ahead [12]. The participant kept a steady breath and sustained this position for 20 s, during which the sEMG signal was recorded. Each participant repeated this procedure three times to acquire an average sEMG value.

E2 (squatting on the bar): The participant put the feet on the second bar, and the hands apart on bar at shoulder level in a squatting position. The participant would then guide the hip below the rib hump to move laterally, backwards, and downward [12]. Then the participant sustained the downward position for 20 s, during which the sEMG signal was recorded. Each participant repeated this procedure three times to acquire an average sEMG value.

E3 (kneeling on one side): The participant kneeled down with the trunk leaning over to the convex side, then stretched out the leg on the concave side, rotated outwards and placed it laterally to form the leg and the upper body as a line. The participant kept the pelvis upright and hands on the hips. The participant kept a steady breath and sustained this position for 20 s, during which the sEMG signal was recorded. Each participant repeated this procedure three times to acquire an average sEMG value.

E4 (sitting with side bending): The participant sat with the buttock on the heel and kept the pelvis upright, then leaned the trunk over to the convex side and put the hand on the convex side on a block to support the oblique body. Then, the participant sustained this position for 20 s, during which the sEMG signal was recorded. Each participant repeated this procedure three times to acquire an average sEMG value.

**Figure 1.** Four different Schroth exercise positions: (**a**) E1—on the fours, (**b**) E2—squatting on the bar, (**c**) E3—kneeling on one side, (**d**) E4—sitting with side bending.

#### 2.2.3. After Exercise

The participant was instructed to stand in the relaxed standing position. The same procedure of sEMG activity measurement as the pre-exercise was taken again, to record the sEMG signals after exercise.

#### *2.3. Data and Statistical Analysis*

The obtained sEMG signals were amplified and sampled at 1500 Hz using myoMUS-CLE™ software (Noraxon Inc., Scottsdale, AZ, USA). The raw data was band-passed filtered (Butterworth with a cut-off frequency of 20–500 Hz). The sEMG signal of each exercise was divided into three sequences. Each sequence was normalized for time. The root mean square (RMS) quantifying the sEMG amplitude of the averaged sEMG signal was calculated. The paraspinal muscle symmetry index (PMSI) was calculated as RMSconvex/RMSconcave. The PMSI of being close to 1 (e.g., PMSI = 1) referred to the high symmetry of the paraspinal muscle. The PMSI < 1 referred to a greater RMSconcave than RMSconvex, and PMSI > 1 referred to a greater RMSconvex than RMSconcave of the scoliotic curve.

The statistical package SPSS, version 22 (SPSS Inc, Chicago, IL, USA), was used for all statistical analyses. One-way repeated ANOVA was used to compare the PMSI before, during and after exercise, and examine for the existence of significant difference. A post hoc adjusted for multiple comparisons with the Bonferroni method was used if significant differences among overall PMSIs were found. Two-way repeated ANOVA and the Bonferroni correction for multiple comparisons was adopted to analyze the RMS of

sEMG on the concave and convex side before, during and after exercise, and examine for the existence of significant difference. The level of significance was set at 0.05.

#### **3. Results**

*3.1. Participants*

A total of nine patients with AIS participated in this study. Their demographic data are shown in Table 1.

**Table 1.** Demographic data (*n* = 9).

