Usefulness of computed tomography based three-dimensional reconstructions to assess the critical shoulder angle

Predictability and functional impact of lateral acromioplasty using preoperative 2D planning on the correction of the critical shoulder angle (CSA) in patients with rotator cuff repair

Background

Rotator cuff tears are highly prevalent, and their association with critical shoulder angle (CSA) has been reported. There is controversy regarding whether the morphology of the acromion influences its incidence, as well as whether acromioplasty would impact the results of a rotator cuff repair. Lateral acromioplasty does not play a decompressive role; rather, it aims to correct the deltoid vector. According to some publications, this would achieve less loading on the repaired rotator cuff, a lower retear rate, and better function. CSA correction with lateral acromioplasty can be planned with radiography (2-dimensional [2D]), but its predictability has not been fully studied. The primary objective of this study is to evaluate the predictability of 2D planning with radiography in CSA correction in patients with rotator cuff repair. The secondary objective is to analyze the association between the correction of the CSA and the functional outcomes.

Methods

This single-center, prospective, observational, analytical study included candidates for arthroscopic repair of a rotator cuff tear with a CSA > 35°. Lateral acromioplasty was performed as planned with preoperative radiography (2D) to achieve a CSA of 35°. The degrees to be corrected were calculated. CSA was recalculated with a postoperative radiography; and the error in the planned grades to be corrected was calculated. At the end of follow-up, Visual Analog Scale, Subjective Shoulder Value (SSV), and Quick-Disabilities of the Arm, Shoulder, and Hand (Quick-DASH) scores were evaluated.

Results

Forty one cases were included, 43.9% were men, and the mean age was 55.5 ± 8.6 years. The mean preoperative and postoperative CSA were 39.6° ± 1.9° and 35.7° ± 2.3°, respectively; 41.5% achieved a postoperative CSA ≤ 35°. The mean CSA planned correction error was 45.7 ± 28.8%. At the end of the follow-up, 33 (78.6%) had a functional evaluation, with an average follow-up of 41 ± 6.8 months. The mean Visual Analog Scale, Quick-DASH, and SSV were 0.9 ± 1.6, 5.3 ± 7.5, and 92.7 ± 10.6, respectively. There was a significant difference in Quick-DASH (P = .01) and SSV (P = .02) according to whether a postoperative CSA ≤ 35° was achieved.

Conclusion

In lateral acromioplasty, planning of CSA correction with radiography (2D) is imprecise. Reaching a CSA ≤ 35° positively influences functional results.

The Critical Shoulder Angle Can be Accurately and Reliably Determined from Three-Dimensional Computed Tomography Images

OBJECTIVE

Anteroposterior (AP) radiographs do not necessarily offer the optimal approach to measuring the critical shoulder angle (CSA) due to the malposition of the scapula. Three-dimensional computed tomography (3D-CT) may offer some advantages, including the ability to rotate the scapula for position alignment and pre-operative planning for reducing CSA. This study aimed to investigate the accuracy and reliability of CSA measurement in 3D-CT and to determine whether there is an association between CSA and rotator cuff tears (RCTs).

METHODS

In this retrospective study we identified 200 patients who received shoulder arthroscopy from 2019 to 2021, including 142 patients (81 females, 61 males) with RCTs and 58 patients (14 females, 44 males) with non-RCTs. For each participant, CSA was measured from standard shoulder AP radiographs and anterior views of 3D-CT of the scapula by two independent assessors. Inter- and intra-observer agreements were assessed by the intraclass correlation coefficient (ICC). The relationship between the two measurement methodologies was determined by Spearman's correlation coefficient and Bland-Altman plots. Discriminative capacity was calculated by using receiver operating curve (ROC) analyses in the whole cohort and age sub-groups above and below 45 years.

RESULTS

We found perfect inter-observer (ICC >0.96) and intra-observer (ICC >0.97) reliabilities for CSA measurements obtained from the standard AP radiographs and the 3D-CT. There was a strong correlation between the two methods (r = 0.960, P < 0.001). The mean CSA was 31.7° ± 4.2° in the standard AP radiographs and 31.8° ± 4.4° in the 3D-CT (mean difference 0.02°, P = 0.940; bias 0.02°, limits of agreement -2.29° to +2.33°). ROC analysis of the whole cohort showed that the CSA measured in the standard AP radiographs (area under the ROC curve [AUC] = 0.812, P < 0.001) and the 3D-CT (AUC = 0.815, P < 0.001) predicted RCT with high confidence. ROC analysis of patients aged ≥45 years showed that the CSA measured from the standard AP radiographs (AUC = 0.869, P < 0.001) and the 3D-CT (AUC = 0.870, P < 0.001) were very good at predicting RCTs.

CONCLUSION

CSA measured from standard AP radiographs and 3D-CT showed high consistency, and the CSA could be accurately and reliably measured using 3D-CT. CSAs measured from standard AP radiographs and 3D-CT could predict RCTs, especially in patients aged ≥45 years.

Deep-Learning-Based Segmentation of the Shoulder from MRI with Inference Accuracy Prediction

Three-dimensional (3D)-image-based anatomical analysis of rotator cuff tear patients has been proposed as a way to improve repair prognosis analysis to reduce the incidence of postoperative retear. However, for application in clinics, an efficient and robust method for the segmentation of anatomy from MRI is required. We present the use of a deep learning network for automatic segmentation of the humerus, scapula, and rotator cuff muscles with integrated automatic result verification. Trained on N = 111 and tested on N = 60 diagnostic T1-weighted MRI of 76 rotator cuff tear patients acquired from 19 centers, a nnU-Net segmented the anatomy with an average Dice coefficient of 0.91 ± 0.06. For the automatic identification of inaccurate segmentations during the inference procedure, the nnU-Net framework was adapted to allow for the estimation of label-specific network uncertainty directly from its subnetworks. The average Dice coefficient of segmentation results from the subnetworks identified labels requiring segmentation correction with an average sensitivity of 1.0 and a specificity of 0.94. The presented automatic methods facilitate the use of 3D diagnosis in clinical routine by eliminating the need for time-consuming manual segmentation and slice-by-slice segmentation verification.