Reliability of Manual Measurements Versus Semiautomated Software for Glenoid Bone Loss Quantification in Patients With Anterior Shoulder Instability

Abstract

Background: The presence of glenoid bone defects is indicative in the choice of treatment for patients with anterior shoulder instability. In contrast to traditional linear- and area-based measurements, techniques such as the consideration of glenoid concavity have been proposed and validated.

Purpose: To compare the reliability of linear (1-dimensional [1D]), area (2-dimensional [2D]), and concavity (3-dimensional [3D]) measurements to quantify glenoid bone loss performed manually and to analyze how automated measurements affect reliability.

Study design: Cohort study (diagnosis); Level of evidence, 3.

Methods: Computed tomography images of 100 patients treated for anterior shoulder instability with differently sized glenoid defects were evaluated independently by 2 orthopaedic surgeons manually using conventional software (OsiriX; Pixmeo) as well as automatically with a dedicated prototype software program (ImFusion Suite; ImFusion). Parameters obtained included 1D (defect diameter, best-fit circle diameter), 2D (defect area, best-fit circle area), and 3D (bony shoulder stability ratio) measurements. Mean values and reliability as expressed by the intraclass correlation coefficient [ICC]) were compared between the manual and automated measurements.

Results: When manually obtained, the measurements showed almost perfect agreement for 1D parameters (ICC = 0.83), substantial agreement for 2D parameters (ICC = 0.79), and moderate agreement for the 3D parameter (ICC = 0.48). When measurements were aided by automated software, the agreement between raters was almost perfect for all parameters (ICC = 0.90 for 1D, 2D, and 3D). There was a significant difference in mean values between manually versus automatically obtained measurements for 1D, 2D, and 3D parameters (P < .001 for all).

Conclusion: While more advanced measurement techniques that take glenoid concavity into account are more accurate in determining the biomechanical relevance of glenoid bone loss, our study showed that the reliability of manually performed, more complex measurements was moderate.

Keywords: BSSR; glenoid bone loss; glenoid concavity; glenoid defect; shoulder instability; shoulder segmentation software.

Figure 1.

En face view of the glenoid on a 3-dimensional CT. The diameter (A; red) and the area (B; green) of the BFC positioned on the inferior aspect of the glenoid were determined. The extent of the defect was determined 1-dimensionally by assessing the defect diameter (C; yellow) in relation to the BFC diameter (A) and 2-dimensionally by assessing the defect area (D; blue) in relation to the BFC area (B).

Figure 2.

For calculating the BSSR, the concavity diameter obtained by drawing a tangent line from one apex of the concavity to the opposite concavity (A; yellow line) as well as the concavity depth, defined as the distance from the deepest point of the concavity to the tangent line (B; red line), were determined, and the BSSR was calculated according to Moroder et al. Multiplanar reconstruction was used to obtain standardized axial images that were perpendicular to the long axis of the glenoid and passed through the center of the best-fit circle.

Figure 3.

Computed tomography–based measurements conducted automatically using ImFusion software. (A) The result was a label map representing the scapula and the humerus. The segmentation output of the humerus and scapula was reviewed and refined by the user. (B) Consequently, the geometric properties of the glenoid were characterized and the best-fit circle estimated for further refinement.

Figure 4.

Evaluation of a shoulder CT image using ImFusion software. (A) Linear- and area-based glenoid bone loss were calculated by projecting the glenoid mesh contour (red shaded area) onto the BFC plane (green circle). Linear-based glenoid bone loss was defined as the maximum perpendicular distance between the BFC and the glenoid contour projection (red line), and area-based glenoid bone loss was defined as the BFC area minus the area of the glenoid contour projection inside the BFC (yellow line). (B) To compute the BSSR, the concavity depth was determined. For this purpose, the glenoid mesh was cut along its short axis by a plane passing through the BFC center, and the concavity depth was estimated as the farthest point from a line parallel to the glenoid short axis that touched the glenoid mesh. As the measurement is very sensitive to glenoid mesh irregularities, the calculation was repeated along 10 different mesh cuts, performed with planes parallel to the first plane (yellow lines), to produce the final concavity depth value as an average.