Athletes from three Australian-rules football clubs (two in national competition, one in state competition) gave informed consent before the onset of the 1999 season. Approval was also obtained from the clinical research ethics committee before the study. Between February and September 1999, 37 athletes with clinically suspected hamstring injuries (defined as posterior thigh pain that prevented training or playing) underwent clinical assessment and MR imaging on a total of 44 occasions. Seven athletes were examined and imaged twice because of recurrent injury, but the data pertinent to later injuries were excluded from analysis because the preceding injury may have influenced the appearances of the second MR imaging study. Therefore, during the 1999 season, 37 cases had no preceding hamstring injury.

All 37 athletes were male, the median age was 24 years (range, 17-32 years), and the body mass index ranged from 20.9 to 28.2 kg/m². During clinical assessment, athletes were asked to rate their pain on a scale of 0-10 (0 = no pain, 10 = maximal pain intensity). All athletes were treated according to a predefined protocol involving an initial period of immobilization (rest, ice compression, and elevation) immediately after injury followed by gradually increasing mobilization (walking, stretching, and physiotherapy) and activity according to athlete pain levels. Sonography and electric stimulation were not part of the protocol. MR imaging data other than the presence or absence of hamstring abnormality were not made available to managing physicians. The number of days between injury and recommencement of competition was noted during follow-up.

Athletes underwent MR imaging with a 1.5-T imaging unit (Vision; Siemens Medical Systems, Erlangen, Germany) 2-6 days (median, 3 days) after injury. Before imaging, we affixed a vitamin E capsule to the athlete's leg overlying the point of maximal tenderness indicated by the athlete. A flexible circularly polarized body array coil was used with the patient in the supine position. After coronal localization, we acquired axial T1-weighted turbo spin-echo MR images (TR/TE, 802/12; section thickness, 10 mm; field of view, 30-32 × 40.0-42.7 cm; echo-train length, 3; matrix, 213 × 512; intersection gap, 20%; acquisitions, 2) and axial inversion recovery T2-weighted turbo spin-echo images (5032/30; section thickness, 10 mm; field of view, 30.0-31.9 × 40.0-42.5 cm; echo-train length, 7; matrix, 182 × 256; inversion time, 150 msec; intersection gap, 20%; acquisitions, 1). Additionally, we obtained sagittal T1-weighted turbo spin-echo images (676/12; section thickness, 7 mm; field of view, 24 × 32 cm; echo-train length, 3; matrix, 213 × 512; intersection gap, 20%; acquisitions, 2), sagittal inversion recovery T2-weighted turbo spin-echo images (5000/30; section thickness, 7 mm; field of view, 24 × 32 cm; matrix, 189 × 256; echo-train length, 7; inversion time, 150 msec; intersection gap, 20%; acquisitions, 1), and axial gradient-echo images (610/18; flip angle, 20°; section thickness, 10 mm; field of view, 30.0-31.4 × 40.0-41.9 cm; matrix, 192 × 512; intersection gap, 20%; acquisitions, 1). The latter images were acquired with the intent of detecting hemorrhage with greater sensitivity than on turbo spin-echo sequences.

Images were independently interpreted by two radiologists who were unaware of clinical details other than the suspected hamstring injury. Each radiologist recorded the presence or absence of abnormal focal intra- and extramuscular T2 hyperintensity, the muscles involved, and the location of the abnormality in each muscle. The muscle exhibiting the most extensive T2 hyperintensity was noted, and abnormal signal in this muscle was categorized as involving the musculotendinous junction (proximal or distal) or the muscle belly (adjacent to the intramuscular tendon or not). Injuries in the muscle belly were denoted as proximal (above the femoral origin of the short head of the biceps) or distal. Hamstring muscle injury was considered to be present when abnormal T2 signal was seen in one or more hamstring muscles. Discordant cases (regarding either the presence or absence of hamstring muscle injury or the muscle most involved) were reviewed and consensus reached. At that time, transverse images showing the maximal intra- and extramuscular extent of T2 hyperintensity were selected for further analysis.

Measurements were performed at the console of the MR imaging unit and were directed to the selected transverse T2-weighted image of the muscle showing the most extensive hyperintensity. The maximal anteroposterior and transverse dimensions of the region of intramuscular hyperintensity were measured from the designated image (Fig. 1A). Slice position notation on images was used to estimate the craniocaudal extent of intramuscular T2 hyperintensity in that muscle belly. Sagittal T2-weighted images were not used for the latter measurement because in some cases T2 hyperintensity was obliquely oriented with respect to the sagittal plane. The volume of abnormal muscle was approximated using the formula for a prolate ellipsoid ([π/6] × anteroposterior × transverse × craniocaudal extent) [8]. Irregular regions of interest were drawn around the muscle to obtain the cross-sectional area of the whole muscle belly and of intramuscular T2 hyperintensity (Fig. 1B). The ratio of abnormal to total muscle cross-sectional area was thus obtained. The length of extramuscular T2 hyperintensity was also measured (multiple line segments) from the selected transverse T2-weighted image that best displayed this finding (Fig. 2).

Statistical analysis was performed using Statistical Package for the Social Sciences software (version 10.0; SPSS, Chicago, IL). The numbers of patients in this study were relatively small, so statistical analysis was used to indicate patterns of association.

The gamma statistic [9] was used to investigate the association between the number of days lost from competition and the presence or absence of intramuscular T2 hyperintensity. Spearman's rank correlation coefficient was used to assess relationships between the number of days lost from competition and the following MR indexes of muscle injury: percentage of cross-sectional area of abnormal muscle, approximate volume of abnormal muscle, and length of linear extramuscular T2 hyperintensity. Spearman's rank correlation coefficient was also used to examine relationships between an athlete's subjective pain scores and the percentage of cross-sectional area of abnormal muscle, the approximate volume of abnormal muscle, and the length of extramuscular T2 hyperintensity. The Mann-Whitney U test (exact) for nonpaired comparisons was used to assess relationships between the number of days lost from competition and the following factors: primary involvement of the long head of the biceps femoris muscle, primary involvement of the semitendinosus muscle, proximal muscle injury, muscle injury extending to the upper or lower musculotendinous junction, and the presence of greater than 50% cross-sectional area of abnormal muscle. The same statistical test was used to examine relations between MR imaging measurements of the extent of muscle injury and primary involvement of the biceps femoris and semitendinosus muscles.

A p value of 0.05 was considered an indicator of statistical significance; however, because of the number of statistical tests undertaken, the levels of statistical significance should be viewed with caution.

Thirty-two (86%) of 37 MR imaging studies showed focal intramuscular T2 hyperintensity that was considered to represent muscle injury. In all but two of these 32 cases (isolated abnormalities of the adductor magnus and vastus lateralis muscles), the most extensive MR imaging abnormality was in relation to the long head of the biceps or semitendinosus muscle, leaving 30 cases in which hamstring muscle injury was considered to be present. T2 hyperintensity was also noted in the rectus femoris (n = 2) and gracilis (n = 1) muscles, but injuries in these three athletes were considered secondary, there being prominent hamstring abnormality on T2-weighted images.

Altered muscle signal intensity was least evident on T1-weighted images, most obvious on inversion recovery T2-weighted images, and of intermediate conspicuity on gradient-echo images (Fig. 3A,3B,3C,3D). Gradient-echo images revealed focal hypointensity suggestive of blood products in evolution in only one case (primary adductor magnus injury, Fig. 4A,4B,4C). Linear extramuscular T2 hyperintensity (range, 1.8-24.9 cm; median, 8.6 cm) was present in 25 (68%) of 37 cases.

The distribution of hamstring muscle injury is shown in Table 1. No avulsion injuries and no complete tendon ruptures were present. The long head of the biceps muscle was the most commonly affected muscle (26 [87%] of 30 cases), and this occurred in association with semitendinosus muscle strain in 11 (37%) of 30 athletes (long head of biceps muscle the dominant injury in six cases and semitendinosus muscle injury dominant in five cases). The long head of the biceps was the only muscle injured or the major site of abnormality in 21 (70%) of 30 cases, and the semitendinosus muscle was the sole or dominant abnormality in nine cases (30%). The short head of the biceps was affected in five (17%) of 30 cases of hamstring muscle injury, but in each case involvement was minor and adjacent to the long head of the biceps muscle belly. When the long head of the biceps was the sole or primary site of injury, no greater intramuscular extent of injury occurred on the basis of approximate volume (Mann-Whitney U test, p = 0.26) or percentage of abnormal surface area (Mann-Whitney U test, p = 0.48). Similarly, sole or primary involvement of the semitendinosus muscle was not associated with greater intramuscular extent of injury.

The musculotendinous junction was involved in 28 (93%) of the 30 cases with hamstring muscle injury at MR imaging, the two exceptions being small isolated injuries to the long head of the biceps that extended to the epimysium. Of these 28 cases, the intramuscular tendon was involved in 24 (86%) and the musculotendinous junction at the muscle end was solely involved in the other four cases (14%; three distal, one proximal). Five (21%; five proximal, none distal) of 24 injuries affecting the intramuscular tendon extended to the musculotendinous junction. Muscle or tendon injury was situated above the origin of the short head of the biceps in 11 instances (37%).

Although the long head of the biceps was the sole or dominant site of injury in 21 cases, involvement of this muscle did not influence the duration of absence from professional sport (Mann-Whitney U test, p = 0.86). Similarly, sole or dominant semitendinosus muscle injury did not appear to result in greater rehabilitation time (Mann-Whitney U test, p = 0.86). Using the short head of the biceps as a criterion for classifying injury as proximal (above femoral origin) or distal, we noted 11 proximal and 19 distal injuries. No association was shown (Mann-Whitney U test, p = 0.17) between proximal muscle injury and time lost from competition. Hamstring muscle injury involving both muscle belly and the adjacent musculotendinous junction did not influence the convalescence interval (Mann-Whitney U test, p = 0.93).

The number of days lost from competition by those with hamstring muscle injury ranged from 13 to 48 (median, 27 days). An association between the presence of focal intramuscular T2 hyperintensity and the number of days lost from competition (γ = 0.69, p = 0.04) was present. Measurements of the percentage of abnormal cross-sectional area (range, 8-100%; median, 46%) and approximate volume of abnormal muscle (range, 0.04-175.6 cm³; median, 16.8 cm³) were widely and not normally distributed and are provided in Table 2. Athletes with hamstring muscle injury recorded subjective pain scores ranging from 2 to 8 (median, 5). Table 3 provides statistics indicating relationships between MR imaging measurements of injury extent and the number of days missed because of injury as well as subjective pain scores. Abnormal cross-sectional muscle area greater than 50% was associated with a longer rehabilitation time (Mann-Whitney U test, p < 0.001).