Comparison Between 3-Dimensional Multiple-Echo Recombined Gradient Echo Magnetic Resonance Imaging and Arthroscopic Findings for the Evaluation of Acetabular Labrum Tear

A Deep Learning Approach for MRI in the Diagnosis of Labral Injuries of the Hip Joint

BACKGROUND

The diagnosis of labral injury on MRI is time-consuming and potential for incorrect diagnoses.

PURPOSE

To explore the feasibility of applying deep learning to diagnose and classify labral injuries with MRI.

STUDY TYPE

Retrospective.

POPULATION

A total of 1016 patients were divided into normal (n = 168, class 0) and abnormal labrum (n = 848) groups. The abnormal group consisted of n = 111 with class 1 (degeneration), n = 437 with class 2 (partial or complete tear), and n = 300 with unclassified injury. Patients were randomly divided into training, validation, and test cohort according to the ratio of 55%:15%:30%.

FIELD STRENGTH/SEQUENCE

Fat-saturation proton density-weighted fast spin-echo sequence at 3.0 T.

ASSESSMENT

Convolutional neural network-6 (CNN-6) was used to extract, discriminate, and detect oblique coronal (OCOR) and oblique sagittal (OSAG) images. Mask R-CNN was used for segmentation. LeNet-5 was used to diagnose and classify labral injuries. The weighting method combined the models of OCOR and OSAG. The output-input connection was used to correlate the whole diagnosis/classification system. Four radiologists performed subjective diagnoses to obtain the diagnosis results.

STATISTICAL TESTS

CNN-6 and LeNet-5 were evaluated by area under the receiver operating characteristic (ROC) curve and related parameters. The mean average precision (MAP) evaluated the Mask R-CNN. McNemar's test was used to compare the radiologists and models. A P value < 0.05 was considered statistically significant.

RESULTS

The area under the curve (AUC) of CNN-6 was 0.99 for extraction, discrimination, and detection. MAP values of Mask R-CNN for OCOR and OSAG image segmentation were 0.96 and 0.99. The accuracies of LeNet-5 in the diagnosis and classification were 0.94/0.94 (OCOR) and 0.92/0.91 (OSAG), respectively. The accuracy of the weighted models in the diagnosis and classification were 0.94 and 0.97, respectively. The accuracies of radiologists in the diagnosis and classification of labrum injuries ranged from 0.85 to 0.92 and 0.78 to 0.94, respectively.

DATA CONCLUSION

Deep learning can assist radiologists in diagnosing and classifying labrum injuries.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

3D MRI of the Hip Joint: Technical Considerations, Advantages, Applications, and Current Perspectives

Magnetic resonance imaging (MRI) is a common choice among various imaging modalities for the evaluation of hip conditions. Conventional MRI with two-dimensional acquisitions requires a significant amount of time and is limited by partial-volume artifacts and suboptimal fluid-to-cartilage contrast. Recent hardware and software advances have resulted in development of novel isotropic three-dimensional (3D) single-acquisition protocols that cover the volume of the entire hip and can be reconstructed in arbitrary planes for submillimeter assessment of bony and labro-cartilaginous structures in their planes of orientation. This technique facilitates superior identification of small labral tears and other hip lesions with better correlations with arthroscopy. In this review, we discuss technical details related to 3D MRI of the hip, its advantages, and its role in commonly encountered painful conditions that can be evaluated with great precision using this technology. The entities described are femoroacetabular impingement with acetabular labral tears, acetabular dysplasia, avascular necrosis, regional tendinopathies and tendon tears, bursitis, and other conditions.

MRI-based 3D models of the hip joint enables radiation-free computer-assisted planning of periacetabular osteotomy for treatment of hip dysplasia using deep learning for automatic segmentation

Introduction

Both Hip Dysplasia(DDH) and Femoro-acetabular-Impingement(FAI) are complex three-dimensional hip pathologies causing hip pain and osteoarthritis in young patients. 3D-MRI-based models were used for radiation-free computer-assisted surgical planning. Automatic segmentation of MRI-based 3D-models are preferred because manual segmentation is time-consuming.To investigate(1) the difference and(2) the correlation for femoral head coverage(FHC) between automatic MR-based and manual CT-based 3D-models and (3) feasibility of preoperative planning in symptomatic patients with hip diseases.

Methods

We performed an IRB-approved comparative, retrospective study of 31 hips(26 symptomatic patients with hip dysplasia or FAI). 3D MRI sequences and CT scans of the hip were acquired. Preoperative MRI included axial-oblique T1 VIBE sequence(0.8 mm³ isovoxel) of the hip joint. Manual segmentation of MRI and CT scans were performed. Automatic segmentation of MRI-based 3D-models was performed using deep learning.

Results

(1)The difference between automatic and manual segmentation of MRI-based 3D hip joint models was below 1 mm(proximal femur 0.2 ± 0.1 mm and acetabulum 0.3 ± 0.5 mm). Dice coefficients of the proximal femur and the acetabulum were 98 % and 97 %, respectively. (2)The correlation for total FHC was excellent and significant(r = 0.975, p < 0.001) between automatic MRI-based and manual CT-based 3D-models. Correlation for total FHC (r = 0.979, p < 0.001) between automatic and manual MR-based 3D models was excellent.(3)Preoperative planning and simulation of periacetabular osteotomy was feasible in all patients(100 %) with hip dysplasia or acetabular retroversion.

Conclusions

Automatic segmentation of MRI-based 3D-models using deep learning is as accurate as CT-based 3D-models for patients with hip diseases of childbearing age. This allows radiation-free and patient-specific preoperative simulation and surgical planning of periacetabular osteotomy for patients with DDH.

Use of a 3D Virtually Reconstructed Patient-Specific Model to Examine the Effect of Acetabular Labral Interference on Hip Range of Motion

Background:

The labrum is likely to influence impingement, which may also depend on acetabular coverage. Simulating impingement using 3-dimensional (3D) computed tomography (CT) is a potential solution to evaluating range of motion (ROM); however, it is based on bony structures rather than on soft tissue.

Purpose:

To examine ROM when the labrum is considered in a 3D dynamic simulation. A particular focus was evaluation of maximum flexion and internal rotation angles before occurrence of impingement, comparing them in cases of cam-type femoroacetabular impingement (FAI) and borderline developmental dysplasia of the hip (BDDH).

Study Design:

Descriptive laboratory study.

Methods:

Magnetic resonance imaging (MRI) and CT scans of 40 hips (20 with cam-type FAI and 20 with BDDH) were reviewed retrospectively. The thickness and width of the labrum were measured on MRI scans. A virtual labrum was reconstructed based on patient-specific sizes measured on MRI scans. The impingement point was identified using 3D dynamic simulation and was compared with the internal rotation angle before and after labral reconstruction.

Results:

The thickness and width of the labrum were significantly larger in BDDH than in FAI (P < .001). In FAI, the maximum internal rotation angles without the labrum were 30.3° at 90° of flexion and 56.9° at 45° of flexion, with these values decreasing to 18.7° and 41.4°, respectively, after labral reconstruction (P < .001). In BDDH, the maximum internal rotation angles were 48.0° at 90° of flexion and 76.7° at 45° of flexion without the labrum, decreasing to 31.1° and 55.3°, respectively, after labral reconstruction (P < .001). The differences in the angles before and after labral reconstruction were larger in BDDH than in FAI (90° of flexion, P = .03; 45° of flexion, P = .01).

Conclusion:

As the labrum was significantly more hypertrophic in BDDH than in FAI, the virtual labral model revealed that the labrum’s interference with the maximum internal rotation angle was also significantly larger in BDDH.

Clinical Relevance:

The labrum has a significant effect on impingement; this is more significant for BDDH than for FAI.

Computed Tomography-Based Three-Dimensional Analyses Show Similarities in Anterosuperior Acetabular Coverage Between Acetabular Dysplasia and Borderline Dysplasia

PURPOSE

(1) To compare the acetabular coverage between dysplasia, borderline dysplasia, and control acetabulum in a quantitative 3-dimensional manner; and (2) to evaluate correlations between the radiologic parameters and the 3-dimensional zonal-acetabular coverage.

METHODS

We reviewed contralateral hip computed tomography images of patients 16 to 60 years of age who underwent 1 of 3 types of surgeries: eccentric rotational acetabular osteotomy, curved intertrochanteric varus osteotomy, and total hip replacement with minimum 1-year follow-up from January 2013 to April 2018. A point-cloud model of the acetabulum created from computed tomography was divided into 6 zones. Three-dimensional acetabular coverage was measured radially at intervals of 1°. Mean radial acetabular coverage for each zone was named ZAC (zonal acetabular coverage) and was compared among the 3 subgroups (control: 25° ≤lateral center-edge angle [LCEA] <40°; borderline: 20° ≤LCEA <25°; and dysplasia: LCEA ≤20°) statistically. Further, the correlations between the ZAC in each zone and the LCEA were analyzed using Pearson's correlation coefficient.

RESULTS

One-hundred fifteen hips were categorized as control (36 hips), borderline (32 hips), and dysplasia (47 hips). The mean anterocranial ZAC in the borderline (87.5 ± 5.7°) was smaller than that in the control (92.6 ± 5.9°, P = .005) but did not differ compared with the dysplasia (84.5 ± 7.6°, P = .131). In contrast, the anterocaudal (71.2 ± 5.0°), posterocranial (85.0 ± 6.4°), and posterocaudal (82.4 ± 4.5°) mean ZACs in the borderline were not different from those in the control (anterocaudal, 74.3 ± 4.6°, P = .090; posterocranial, 87.9 ± 4.3°, P = .082; posterocaudal, 85.1 ± 5.0°, P = .069) respectively. Although there was a very strong positive correlation with supra-anterior ZAC and LCEA (r = 0.750, P < .001), the correlation between the anterocranial ZAC and LCEA was relatively weak (r = 0.574, P < .001).

CONCLUSIONS

The anterosuperior acetabular coverage in the borderline dysplastic acetabulum is more similar to the dysplastic acetabulum than to the normal acetabulum.

CLINICAL RELEVANCE

This study emphasizes the importance of evaluating not only the lateral but also the anterior coverage in borderline dysplasia.

Correlations and Reproducibility Between Radiographic and Radial Alpha Angles in the Evaluation of Cam Morphology

Background:

The alpha angle used to evaluate cam morphology can be determined on different imaging views; however, 2-dimensional (2D) imaging can present limitations in terms of the reproducibility of the radial alpha angle. Recent developments in 3-dimensional (3D) high-resolution magnetic resonance imaging (MRI) have allowed detailed evaluations of the radial alpha angle.

Purpose:

To determine whether there are any correlations or discrepancies between the 2D alpha angle on plain radiography and the maximum radial alpha angle on 3D MRI.

Study Design:

Cohort study (diagnosis); Level of evidence, 2.

Methods:

A total of 42 hips from 39 patients (19 males, 20 females) were analyzed, including 22 hips with femoroacetabular impingement (FAI; mean age, 41 years) and 20 hips with borderline developmental dysplasia of the hip (BDDH; mean age, 43 years). Radial images were reconstructed from 3D multiple echo recombined gradient echo (MERGE) MRI. Differences in the maximum radial alpha angle on MRI between hips with FAI and BDDH were evaluated. Correlations and discrepancies between the maximum radial alpha angle on MRI and alpha angles on the anteroposterior, cross-table lateral, and 45° Dunn views of radiography were also evaluated.

Results:

The maximum radial alpha angle was significantly higher for hips with FAI than for hips with BDDH. On average, the greatest alpha angle on radial MRI was higher than the alpha angle on each of the 3 radiographic views for both FAI and BDDH. The 45° Dunn view revealed the smallest discrepancy for both FAI (P = .005) and BDDH (P = .002). The cross-table lateral view had the highest correlation with the maximum radial alpha angle for BDDH (P < .001).

Conclusion:

We reconfirmed the utility of the 45° Dunn view, with it presenting the best reproducibility for the maximum radial alpha angle in the evaluation of cam morphology, while the cross-table lateral view revealed the best correlation with the maximum radial alpha angle, particularly for hips with BDDH.

High spatial resolution whole-neck MR angiography using thin-slab stack-of-stars quiescent interval slice-selective acquisition

PURPOSE

To report a 3D multi-echo thin-slab stack-of-stars (tsSOS) quiescent-interval slice-selective (QISS) strategy for high-resolution magnetic resonance angiography (MRA) of the entire neck in under seven minutes.

METHODS

The neck arteries of eight subjects were imaged at 3 Tesla. Multi-echo 3D tsSOS QISS using a FLASH readout was compared with 3D tsSOS FLASH, 2D QISS, 2D TOF, and 3D TOF. A root-mean-square (RMS) combination of echo time images was tested. Evaluation metrics included arterial signal-to-noise ratio (SNR), arterial-to-muscle contrast-to-noise ratio (CNR), and image quality.

RESULTS

3D multi-echo tsSOS QISS using a RMS combination of echo time images increased SNR and CNR by 60% and 63% with respect to the reconstruction obtained with the shortest echo time. 3D tsSOS QISS showed superior CNR with respect to 3D tsSOS FLASH imaging, and more than 3-fold higher SNR and CNR with respect to 2D radial QISS when normalized for voxel size. 3D tsSOS QISS provided good to excellent image quality that exceeded the image quality of 2D QISS, 2D TOF, and 3D TOF (P < .05).

CONCLUSION

Whole-neck high-resolution nonenhanced MRA is feasible using 3D tsSOS QISS, and produced image quality that exceeded those of competing nonenhanced MRA protocols at 3 Tesla.

Editorial Commentary: Advances in 3-Dimensional Imaging are the Key to Improving our Surgical Precision in Hip Arthroscopy and Beyond

Advances in high-resolution magnetic resonance imaging have driven a wealth of knowledge in orthopaedic basic science. The application of these novel techniques to clinical practice is the next logical step for enhancing our understanding of intra-articular pathology and morphology. The specific diagnostic challenge presented by hip labral and chondral pathology is a particular point of interest, given the increasing popularity of hip arthroscopy. As our field continues to progress in complexity, the integration of new, higher-resolution imaging sequences such as multiple-echo recombined gradient echo and double-echo steady state provide the potential to enhance preoperative planning and ultimately the effectiveness of our arthroscopic techniques.