After jogging on a treadmill for five minutes as a warm-up, surface electrodes were attached to selected muscles of the lower trunk including the left and right rectus abdominis (3 cm lateral to the umbilicus), external obliques (approximately 15 cm lateral to the umbilicus), internal oblique (below the external oblique electrodes and just superior to the inguinal ligament), and lumbar erector spinae (3 cm lateral to L3 spinous process) 22. The skin surfaces where the electrodes were located were cleansed with alcohol and shaved when necessary. Electrodes were placed over the bellies of each muscle parallel to the muscle's line of action with a center-to-center distance of 2.5 cm. Using a MESPEC 4000 telemetry system (Mega Electronics Ltd., Kuopio, Finland), the EMG signals were preamplified with a gain of 500 and band pass filtered at 8-1500 Hz (CMRR > 130 dB) close to the electrodes and telemetrically transmitted to a central receiver (gain = 1, Butterworth filter, 8-500 Hz band pass). The amplified EMG signals were sampled at 900 Hz (12-bit analog-to-digital conversion) using the Peak Motus system.

To obtain maximum EMG levels, two maximal isometric contractions were performed before the experimental trials - the bent-knee sit-up with the trunk inclined at approximately 30° to the horizontal and the trunk hyperextension performed in the prone position on a treatment table. In both maximal contractions, the feet were constrained and the resistance was applied manually at the shoulders. One trial was performed for each maximal contraction and each trial lasted for about 5 s.

After the isometric trials, the EMG transmitter was secured to the left hip of the subject using elastic bands. The subject was then asked to perform three different types of serves - flat, topspin, and slice. In all trials, the subject served with efforts comparable to his first serves during competition. They were asked to target their serves at the corner near the center line (flat and topspin) and sideline (slice). Seven trials were performed for each serve type and the serve type was presented in a random order. At the end of each trial, the subject was asked to rate his own performance based on the pace of the ball and landing location using a 5-point scale (5 = excellent, 0 = poor). For each trial, the EMG signals were collected for 5 s.

After the EMG trials, electrodes were removed and eight reflective markers (each 1 cm in diameter) were placed on the back of the subject. The marker locations were right and left tips of the 11th rib, T9 and T12 spinous processes, right and left posterior superior iliac spines (PSISs), and L3 and L5 spinous processes (Figure 2). The location of the spine level was estimated using the techniques proposed by Tully and Stillman 34 and Schache et al. 35. These markers were used to estimate the 3-dimensional (3D) orientations of the spinal regions directly above and below the lumbar spinal segments. These two regions were considered the middle and lower trunk for the purpose of this study. As a result, the change in relative motion between the middle and lower trunk was treated as motion in the lumbar spinal segments.

To establish the neutral orientation for the markers attached to each subject, the subject was asked to adopt a self-selected comfortable standing posture with arms folded in front of the torso while marker locations were recorded. The subject was then asked to perform seven trials for each type of serve and the order for the type of serve was assigned randomly. Again, the subjects served with efforts comparable to their first serves during competition and rated their own performance at the end of each trial.

The raw EMG signals were filtered using a recursive digital filter (Matlab Elliptic filter, 10-450 Hz band pass) and full-wave rectified. The maximum isometric trial data were smoothed using a moving average of 2 s and the largest average EMG value recorded for each muscle was considered the maximum EMG level. The experimental trial data were smoothed using a moving average of 50 ms before normalizing to the respective maximum EMG levels. An average normalized EMG value was computed for each muscle in each phase for each trial analyzed.

For each standing or serving trial analyzed, coordinate data were extracted from the video pictures (automatic tracking) using a video-based motion analysis system (Peak Motus Motion Measurement System). The three-dimensional (3D) coordinates the eight reflective markers located on the back of the trunk were transformed from the Peak reference frame to local reference frames embedded in the middle and lower trunk (Figure 2). Considering the middle and lower trunk as adjacent segments of a joint, the anatomical joint (AJ) angles between the two segments is the relative orientation of the two segment-embedded local reference frames (see Additional file 1, 36). The three components of the AJ angles represent the rotations about the medio-lateral axis (flexion/extension angle), antero-posterior axis (lateral flexion angle), and longitudinal axis (twisting angle). For each subject, the AJ angles obtained during serving trials were expressed as the angular deviation from the AJ angles recorded at standing posture (i.e., the AJ angles at standing posture are all 0°). The dependent variables for trunk motion were the maximal extension, left lateral flexion, and left and right twisting angles during a serve. For each subject, average normalized EMG values and maximum AJ angles over two trials of the same serve type were used for subsequent analysis.

A significant main effect for the skill was found in the LRA activity (p = .008). The AV group showed significantly higher LRA activation than the AI group. Regardless of the serve type, the activation patterns of the LRA and RRA are quite similar - activations in the descending wind-up and acceleration phases are greater than the ascending windup and follow-through phases (Table 1).

Different patterns of activation were observed in the two EO muscles (Table 1). In general, the subjects in the AV group exhibited greater LEO activity than the subjects in the AI group and the difference was a statistical trend (p = 0.055).

Among the muscles examined, the IO muscles have the largest overall EMG levels (Figure 4). Different patterns of activation were observed in the two IO muscles (Table 1). The LIO was generally more active than the RIO throughout a serve except in the follow-through phase. Very high activation levels (average EMG levels greater than 100% max) were found in the descending windup and acceleration phases in the LIO.

Moderate activity was observed in all phases except the ascending windup for both ES muscles and the descending windup phase for the RES (Table 1). Different patterns of activation were observed in the two ES muscles (Figure 4). Instead of a relatively constant average EMG level after the ascending wind-up phase observed in the LES, the activity of RES increased steadily from ascending wind-up to follow-through phase. Although not statistically significant (p = 0.069), the AV group showed greater RES activation than the AI group (Table 1).

The RA muscles are active during trunk flexion motions like curl-up, sit-up or leg raising exercises 42. Due to its vertical (longitudinal) alignment, the RA has minimal contribution to the production of torque in the transverse plane 43. High activation level of anterior supporting muscles of the lower trunk including RA may lead to unfavorable forces on the spine 44. It has been suggested that an average normalized EMG value to MVIC (maximal voluntary isometric contraction) below 30% is not considered very stressful to the spine structures 45. During the descending windup and acceleration phases of a tennis serve, the EMG level of both RA muscles increased to 40% or higher (Table 1). The AV group showed a higher level of activation during the acceleration phase than the AI group. With the center of gravity of the upper body located behind the lumbar spine, the RA activity (co-contraction) in a hyperextended posture during the descending windup and acceleration phases can drastically increase the loads on the lumbar spine and lead to harmful stress to the lumbar spine structures.

The EO muscle is one of the anterior supporting muscles of the lower trunk that is also active during trunk flexion exercises 42. However, it has been well recognized that the contralateral EO muscle is one of the main movers for axial rotation of the trunk 4647. An appreciable antagonistic activity of the ipsilateral EO during axial rotation was also reported in the literature 48. In the present study, the co-contraction of bilateral EO muscles helps to stabilize the lumbar spine during the tennis serve. This type of co-contraction helps to increase the compressive load and lead to the torsional stiffness of the lumbar spine segments 4950. High activation of LRA and LEO for the right-handers in the present study indicated that these muscles acted as prime movers of twisting to the right occurred during the descending windup and early acceleration phases.

Ipsilateral IO is the agonist to the contralateral EO for the axial rotation of the trunk 46. Similar to the EO muscles, an appreciable co-contraction of contralateral IO is common during trunk twists. Interestingly, bilateral difference in muscle activation is more pronounced in IO than in the EO 51. One explanation is that the function of EO is more complicated than just acting as a prime axial rotator of the lower trunk 4852. Like the EO muscles, bilateral co-contraction of IO observed during a tennis serve helps to provide a stabilizing force to the lumbar spine.

The lumbar ES lies lateral to the multifidus muscle and forms the prominent dorsolateral contour of the back muscles in the lumbar region. The lumbar ES consists of two muscles - the longisimus dorsi and iliocostalis. The lines of action of these two muscles are mostly vertical (or longitudinal) and a bilateral ES contraction can act as a posterior sagittal extensor. However, when contracting unilaterally, these muscles can act as lateral flexors of the lumbar vertebrae 53. During axial rotation, the ipsilateral ES is more active than the contralateral ES 54. It also has been suggested that, during axial rotation, back muscles maintain the spinal posture and stabilize the lumbar spine 2243.

In the present study, we found different patterns of muscle activation during a tennis serve for the two ES muscles. The LES is quite active throughout a tennis serve except during the ascending windup phase while sequentially increased activity from ascending windup to follow-through phases was observed in the RES (Table 1). It is obvious that, for the right-handed subjects in this study, the LES assisted the lateral flexion to the left after the ascending windup phase. To stabilize the trunk during an unbalanced posture in the follow-through phase, the RES becomes highly active during the follow-through phase. The bilateral ES co-contraction is more pronounced in the AV group. This is probably related to the greater left lateral flexion found in the AV subjects (Figure 6). In addition to bilateral co-contractions, front/back co-contractions exist throughout a tennis serve and are especially high toward to end of a serve. This may suggest that the lumbar spine is subject to large compression loads during the follow-through phase.

During the extension motion, the vertebral bodies undergo posterior sagittal rotation and a small posterior translation. A downward movement of the inferior articular processes and the spinous process is also involved which limited by bony impaction between spinous processes 58. This type of impaction is accentuated when the joint is subjected to the action of the back muscles 59. The maximal extension AJ angles obtained in the present study fall within the extension ROM values reported in the literature 5660. It should be emphasized that it may not be adequate to compare the AJ angles in this study with the ROM values reported by other investigators because of the differences in measuring techniques.

The reason why the AI group had significantly greater maximum extension AJ angles than the AV group is difficult to understand. One explanation is that, instead of relying on lumbar hyperextension like the AI subjects did, the subjects in the AV group relied more on the hyperextension of the upper trunk (i.e., thoracic spine) to achieve the overall trunk hyperextension needed for an execution of a tennis serve.

The lateral flexion of the lumbar spine involves a complex and variable combination of lateral bending and rotatory movements of the inter-body joints and diverse movements of the zygapophysial (facet) joints. When compared to lateral bending ROM values for lumbar spinal motion segments measured by X-rays 6061, the maximal left lateral flexion AJ angles are close to the summed value of each motion segment in the lumbar spine. The significantly greater maximal left lateral flexion AJ angle exhibited by the AV group indicates that highly-skilled right-handed players can reach for a greater height during a tennis serve because of the greater left lateral flexion. The significantly greater lateral flexion AJ angle corresponds to the significantly greater LRA activity found in the AV group. This implies that highly-skilled players are subjected to greater asymmetric loads on their lumbar spines due to the greater lateral flexion.

Axial rotation of the lumbar spine involves twisting of the intervertebral discs and impaction of zygapophysial joints. During axial rotation of an intervertebral joint, all the fibers of the annulus fibrosus which are inclined to the direction of rotation will be strained. The other half will be relaxed. Based on the observation that an elongation of collagen beyond about 4% of its resting length can lead to injury of the fiber, it has been estimated that the maximum range of rotation of an intervertebral disc without injury is about 3° 62. The twisting ROM for each lumbar spinal motion segment ranges from 0 - 2° 61. The maximal twisting AJ angles found in the present study are slightly greater than the ROM values reported in the literature. The maximal twisting AJ angles are considered small compared to the amount of shoulder movement during a tennis serve. This clearly indicates that the axial rotation of the trunk during a tennis serve is mostly from the twisting of the upper trunk.

Both bilateral and abdominal/back co-contractions among lower trunk muscles are unavoidable during trunk movements because these muscles function as units to maintain the balance between mobility and stability of the spinal column. As a result, the lumbar spine is subjected to a large amount of compressive and torsional stress during athletic movements due to the co-contraction. Although the importance of torsional stress in the etiology of disc degeneration and prolapse is inconclusive 6364, the link between high compressive load and low back injury and pain is well documented 6566. The activation patterns of the lower trunk muscles clearly demonstrate a high degree of co-contraction during a tennis serve, especially in the descending windup and acceleration phases. In addition to the compressive load, the hyperextension and lateral flexion of the trunk during various phases of a tennis serve may cause shear loads on the lumbar spine. Consequently, stresses upon the various anatomical structures may result in spinal injury and back pain.