This retrospective study was approved by our Institutional Review Board, which waived the requirement for informed consent due to this study's retrospective nature. We searched the medical records of a tertiary hospital for clinically suspected subacromial impingement between November 2012 and May 2013; of those patients, 125 were enrolled in the study. Sixty-three patients (31 men, 32 women) were diagnosed with subacromial impingement on the basis of the following inclusion criteria: arthroscopic RC repair and preoperative plain radiography, tomosynthesis, and CT or MR arthrography of the most symptomatic shoulder within 1 month. The left shoulder was affected in 28 patients. Subacromial impingement was clinically diagnosed by an orthopedic specialist and was defined as pain that persisted for more than 6 months without response to conservative treatment. The physical examination of subacromial impingement was positive in the entire study population.

Of the 62 excluded patients, four had a history of inflammatory joint disease. Twenty-five had a clinical history of shoulder surgery. Two patients had primary adhesive capsulitis which was characterized by the functional restriction of active and passive shoulder motion without evidence of osteopenia or calcific tendinitis on plain radiographs (16). Seventeen patients had a primary diagnosis of shoulder instability. Five had localized pain with direct compression of the acromio-clavicular joint that was suggestive of acromio-clavicular joint degeneration (17181920). Nine were excluded because of the absence of CT or MR arthrography (n = 8) or postoperative re-tear (n = 1). We reviewed each patient's electronic medical record for demographic data, including height and weight.

A digital radiographic (DR) examination (Discovery XR656; GE Healthcare, Milwaukee, WI, USA) that included anteroposterior (AP) and scapular Y views of the shoulder was performed separately for each patient while in a supine position. A DR system was used for indirect conversion with a flat-panel detector (cesium iodide). The maximum detector size was 40.4 × 40.4 cm; the active matrix was 2022 × 2022 pixels and the pixel pitch was 200 microns. A standard AP view was obtained while the patient was standing with the posterior part of the shoulder touching the X-ray film cassette and the patient's arms were along the trunk in an external rotation with the palms facing upwards (21). This system can be used for both conventional radiographic imaging and tomosynthesis imaging. The scout image was a shoulder AP projection image, which was collected using automatic exposure control (AEC). The tube voltage was 70 kV, and a total filtration of 2.7 mm Al was used. All radiographs were collected using AEC and an antiscatter grid (grid ratio 10:1, strip density 70 lines/cm). The tube current-exposure time product was 25 mAs, and the mean total dose at the skin was 0.42 mGy per patient. This skin dose was estimated from the dose calculated by the device using a mathematic method reported by Okkalides and Fotakis (22).

The present study was based on shoulder tomosynthesis examinations performed in an X-ray room that had equipment for tomosynthesis (Discovery XR656 with VolumeRAD option; GE Healthcare). The tomosynthesis examination was performed with the patient standing against the wall stand, and the source-to-image distance was 150 cm for the central projection. Tomosynthesis was obtained in the same position that was used in plain radiography. The examinations were performed using a tube voltage of 70 kV, a total filtration of 2.7 mm Al, and an antiscatter grid (grid ratio 10:1, strip density 70 lines/cm). The dose ratio was 5:1. The low-dose projection images were collected with a minimum of 1.6 mAs per projection. During the AP shoulder tomosynthesis examination, the X-ray tube was moved in a continuous sweeping motion along a straight line in the caudo-cranial direction. With an angular movement between −20° to +20°, a mean of 23.3 low-dose projection radiographs (range: 14–54 images depending on patient size) were acquired with a fixed detector position (23). The reconstructions were 2 mm thick. The effective doses from the shoulder examinations were calculated individually for each patient included in the study. The registered dose-area product, field size, and tube angle were obtained from the Digital Imaging and Communications in Medicine header of each projection radiograph. These data, together with information on the individual patient's height and weight, were used in the Monte Carlo program PCXMC 2.0 (STUK-Radiation and Nuclear Safety Authority, Helsinki, Finland) (2425) to calculate the effective dose for each projection radiography. The mean total effective dose for the entire examination (including scout and tomosynthesis acquisition for shoulder) was 0.32 mSv per patient.

A CT arthrography was performed in 16 patients using a 16-MDCT scanner (Sensation 16; Siemens, Erlangen, Germany). All patients underwent CT arthrography after an injection of approximately 12 mL (range, 10–15 mL) of contrast mixture (13 mL iopromide [Ultravist 370; Bayer AG, Berlin, Germany] and 7 mL normal saline). The patient was positioned supine with the thumb pointing upward. The conditions were a kilovoltage setting of 120 kV, an effective tube current-time product of 150 mAs, CT dose index volume of 11.61 ± 4.58 mGy, and dose-length product of 163.16 ± 86.00 mGy·cm which were automatically produced by the CT vendor as the dose report.

A MR arthrography was performed in 47 patients using a 3T system with a dedicated shoulder coil (Achieva TX; Philips Healthcare, Best, The Netherlands). All patients underwent MR arthrography after an injection of approximately 12 mL (range, 10–15 mL) of gadobutrol (Gadovist; Bayer AG) solution at a concentration of 2 mmol/L. The patients were examined in the supine position with their arm at the side, palm facing up, and hand under their hip to keep the shoulder motionless. The following sequences were acquired: T1-weighted spin-echo images in the coronal oblique plane with fat suppression (repetition time/echo time = 583/10 ms; slice thickness, 4 mm; field of view, 150 × 150 mm; matrix size, 340 × 265) and in the sagittal oblique plane (583/10; slice thickness 4 mm; field of view, 150 × 150 mm; matrix size, 340 × 265); T2-weighted spin-echo images in the coronal oblique plane (2668/80 ms; slice thickness 4 mm; field of view, 150 × 150 mm; matrix size, 340 × 258) and sagittal oblique plane (2668/80 ms; slice thickness, 4 mm; field of view, 150 × 150 mm; matrix size, 340 × 258). The intersection gaps were set at 0.4 mm, respectively.

Two musculoskeletal radiologists independently evaluated the AHD on AP plain radiography, tomosynthesis, and CT or MR arthrography. Each reader measured the AHD three times at 3- to 4-week intervals to avoid recall bias. Both readers were blinded to the radiologic reports, clinical information, and surgical results. The AHD was measured as the shortest distance from the inferior surface of the acromion to the superior cortex of the humerus (26). The first line was drawn through the undersurface of the acromion, and the second line was drawn parallel to the first line and contacted the superior cortex of the humeral head. The distance between the two lines was recorded as the AHD. Three groups were stratified according to the AHD on AP plain radiographs as follows: group 1, 7 mm or less; group 2, 8–10 mm; and group 3, greater than 10 mm (4).

The tomosynthetic AHDs were divided into three compartments: anterior, middle, and posterior (Figs. 1, 2). The anterior AHD was defined as the first point with a clear margin of the inferior surface of the acromion. The middle AHD was measured on the acromio-clavicular joint level which displayed the clear margin of articulation with the shortest distance from the inferior surface of the acromion to the superior cortex of the humerus. The posterior AHD was defined as the last point with a clear margin of the inferior surface of the acromion. Figure 3 provides the X-ray tube position according to the plain radiography and tomosynthesis.

The CT or MR arthrographic AHDs were also divided into three compartments as with the tomosynthetic AHD. The reformatted oblique sagittal CT image or T2-weighted oblique sagittal image was chosen for the measurement. We measured the distances from the undersurface of the acromion at the point directly above the head of the humerus to the center and the top of the humeral head (4). On the MR images, the intermediate signal intensity cartilage of the humeral head was not included in the measurements (26).

To adjust for patient size, one reader measured the humeral head diameter, glenoid height, and glenoid width on the AP plain radiographs three times. The humeral head diameter was defined as the diameter of the humeral head at the level of the anatomical neck, which was used to estimate the anatomic index in the shoulder arthroplasty (27). The glenoid height was defined as the distance from the most superior to the most inferior point on the glenoid (27). The glenoid width was defined as the distance from the most anterior to the most posterior point on the glenoid (27).

For a reference standard, a single orthopedic surgeon with 11 years of experience reviewed each patient's arthroscopic operation record. The RC tears were divided into four groups according to the involved tendon location: isolated supraspinatus, supra-plus-infra-spinatus, supraspinatus-plus-subscapularis, or all RC tendons. The size of the RC tear was estimated during surgery and was given in square centimeters (sagittal × coronal extension). The type of RC tear was distinguished according to the Snyder classification (full-thickness tear; small, medium, large, and massive) (28) and the Ellman classification (partial-thickness tear; articular, bursal, and intratendinous) (29).

An independent t test was used to compare the mean AHDs among the three modalities. Quantitative data were expressed as the mean ± standard deviation (SD). The Pearson correlation coefficient was used to test the association between the tear size and variable anatomic indices and the AHD on plain radiography, tomosynthesis, CT or MR arthrography obtained by two readers. A two-way ANOVA was used to determine whether there were significant differences in AHD according to group and size or type of RC tear. The relationships between the results of the AHD on plain radiography, tomosynthesis, and CT or MR arthrography were assessed by using Spearman correlation coefficients. The intraobserver and interobserver agreement for each imaging method was assessed using the intraclass correlation coefficient (ICC) and the lower limit of its 95% confidence interval. To assess the diagnostic efficacy of the imaging technique, the mean AHD value for the two readers on each modality was used.

Results were considered significant when a p value of less than 0.05 was obtained. All calculations were performed with the Statistical Package for the Social Sciences (SPSS) version 21.0 software (IBM Corp., Armonk, NY, USA).

Table 3 provides the correlation between the tear size and AHD on the plain radiography, tomosynthesis, and CT or MR arthrography. A significant inverse correlation was observed between the tear size and posterior tomosynthetic AHD (Pearson correlation coefficient = -0.312, p = 0.047) and all AHDs on CT or MR arthrography (p < 0.001). No significant association was found between the tear size and plain radiographic AHD (p = 0.169).

There was no significant association between the tear type and plain radiographic AHD (p = 0.111) or tomosynthetic AHD (anterior, p = 0.187; middle, p = 0.161; posterior, p = 0.213) according to a Pearson correlation analysis.

The results of a linear regression analysis demonstrated a significant relationship between the tear type and middle tomosynthetic AHD (r = 0.345, p = 0.025) or posterior tomosynthetic AHD (r = 0.365, p = 0.017).

The humeral head diameter and glenoid height had a significant positive correlation with patient height and weight. No statistically significant correlation was observed between the humeral head diameter and the size or type of RC tear.

A moderate correlation was found between the plain radiographic AHD and anterior tomosynthetic AHD (r = 0.433, p < 0.001). A strong correlation was found between the plain radiographic AHD and middle (r = 0.714, p < 0.001) and posterior tomosynthetic AHDs (r = 0.754, p < 0.001). A moderate correlation was found between the plain radiographic AHD and anterior (r = 0.591, p < 0.001), middle (r = 0.503, p < 0.001), and posterior (r = 0.401, p < 0.001) CT or MR arthrographic AHDs.

The interobserver and intraobserver agreement for the AHD was excellent on the radiography (ICC = 0.966–0.988, 0.935–0.976), tomosynthesis (ICC = 0.947–0.983, 0.975–0.987), and CT or MR arthrography (ICC = 0.896–0.928, 0.974–0.982). The comparisons of the interobserver and intraobserver variabilities for AHD are summarized in Tables 4 and 5.