Ruling out rotator cuff tear in shoulder radiograph series using deep learning: redefining the role of conventional radiograph

Objective To develop a deep learning algorithm that can rule out significant rotator cuff tear based on conventional shoulder radiographs in patients suspected of rotator cuff tear. Methods The algorithm was developed using 6793 shoulder radiograph series performed between January 2015 and June 2018, which were labeled based on ultrasound or MRI conducted within 90 days, and clinical information (age, sex, dominant side, history of trauma, degree of pain). The output was the probability of significant rotator cuff tear (supraspinatus/infraspinatus complex tear with > 50% of tendon thickness). An operating point corresponding to sensitivity of 98% was set to achieve high negative predictive value (NPV) and low negative likelihood ratio (LR−). The performance of the algorithm was tested with 1095 radiograph series performed between July and December 2018. Subgroup analysis using Fisher’s exact test was performed to identify factors (clinical information, radiography vendor, advanced imaging modality) associated with negative test results and NPV. Results Sensitivity, NPV, and LR− were 97.3%, 96.6%, and 0.06, respectively. The deep learning algorithm could rule out significant rotator cuff tear in about 30% of patients suspected of rotator cuff tear. The subgroup analysis showed that age < 60 years ( p  < 0.001), non-dominant side ( p  < 0.001), absence of trauma history ( p  = 0.001), and ultrasound examination ( p  < 0.001) were associated with negative test results. NPVs were higher in patients with age < 60 years ( p  = 0.024) and examined with ultrasound ( p  < 0.001). Conclusion The deep learning algorithm could accurately rule out significant rotator cuff tear based on shoulder radiographs. Key Points • The deep learning algorithm can rule out significant rotator cuff tear with a negative likelihood ratio of 0.06 and a negative predictive value of 96.6%. • The deep learning algorithm can guide patients with significant rotator cuff tear to additional shoulder ultrasound or MRI with a sensitivity of 97.3%. • The deep learning algorithm could rule out significant rotator cuff tear in about 30% of patients with clinically suspected rotator cuff tear.

[1]  E. S. Pearson,et al.  THE USE OF CONFIDENCE OR FIDUCIAL LIMITS ILLUSTRATED IN THE CASE OF THE BINOMIAL , 1934 .

[2]  D. Resnick,et al.  Shoulder impingement syndrome: radiographic evaluation. , 1984, Radiology.

[3]  D. C. Hardy,et al.  The shoulder impingement syndrome: prevalence of radiographic findings and correlation with response to therapy. , 1986, AJR. American journal of roentgenology.

[4]  E. DeLong,et al.  Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. , 1988, Biometrics.

[5]  V. Mow,et al.  The relationship of acromial architecture to rotator cuff disease. , 1991, Clinics in sports medicine.

[6]  G. Samsa,et al.  Likelihood ratios with confidence: sample size estimation for diagnostic test studies. , 1991, Journal of clinical epidemiology.

[7]  F A Matsen,et al.  Repairs of the rotator cuff. Correlation of functional results with integrity of the cuff. , 1991, The Journal of bone and joint surgery. American volume.

[8]  T. Yamamuro,et al.  Use of a thirty-degree caudal tilt radiograph in the shoulder impingement syndrome. , 1992, Journal of shoulder and elbow surgery.

[9]  G. Guyatt,et al.  Users' Guides to the Medical Literature: III. How to Use an Article About a Diagnostic Test: B. What Are the Results and Will They Help Me In Caring for My Patients? , 1994 .

[10]  G. Guyatt,et al.  Users' guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? The Evidence-Based Medicine Working Group. , 1994, JAMA.

[11]  W. Totty,et al.  Acromial arch shape: assessment with MR imaging. , 1995, Radiology.

[12]  G. Walch,et al.  Critical analysis of the supraspinatus outlet view: rationale for a standard scapular Y-view. , 1998, Journal of shoulder and elbow surgery.

[13]  D. Rubin,et al.  Greater tuberosity changes as revealed by radiography: lack of clinical usefulness in patients with rotator cuff disease. , 1999, AJR. American journal of roentgenology.

[14]  M. Päivänsalo,et al.  Supraspinatus outlet view in the diagnosis of stages II and III impingement syndrome. , 2001 .

[15]  M. Päivänsalo,et al.  Supraspinatus outlet view in the diagnosis of stages II and III impingement syndrome , 2001, Acta Radiologica.

[16]  J. Steurer,et al.  Accuracy of Ottawa ankle rules to exclude fractures of the ankle and mid-foot: systematic review , 2003, BMJ : British Medical Journal.

[17]  C. Helms,et al.  Radiographic findings associated with symptomatic rotator cuff tears. , 2003, Journal of shoulder and elbow surgery.

[18]  P. Bossuyt,et al.  Sources of Variation and Bias in Studies of Diagnostic Accuracy , 2004, Annals of Internal Medicine.

[19]  M. Mayerhoefer,et al.  Comparison of MRI and conventional radiography for assessment of acromial shape. , 2005, AJR. American journal of roentgenology.

[20]  J. Ide,et al.  Arthroscopic Transtendon Repair of Partial-Thickness Articular-Side Tears of the Rotator Cuff , 2005, The American journal of sports medicine.

[21]  Frederick A Matsen.  Open rotator cuff repair without acromioplasty. , 2005, The Journal of bone and joint surgery. American volume.

[22]  Andrea J. Frangos,et al.  Accuracy of MRI, MR arthrography, and ultrasound in the diagnosis of rotator cuff tears: a meta-analysis. , 2009, AJR. American journal of roentgenology.

[23]  Yoshua Bengio,et al.  Understanding the difficulty of training deep feedforward neural networks , 2010, AISTATS.

[24]  K. Takagishi,et al.  Prevalence and risk factors of a rotator cuff tear in the general population. , 2010, Journal of shoulder and elbow surgery.

[25]  J. Jacobson.  Shoulder US: anatomy, technique, and scanning pitfalls. , 2011, Radiology.

[26]  G. Garavaglia,et al.  The frequency of subscapularis tears in arthroscopic rotator cuff repairs: A retrospective study comparing magnetic resonance imaging and arthroscopic findings , 2011, International journal of shoulder surgery.

[27]  C. Wijdicks,et al.  The accuracy of magnetic resonance imaging and magnetic resonance arthrogram versus arthroscopy in the diagnosis of subscapularis tendon injury. , 2012, Arthroscopy : the journal of arthroscopic & related surgery : official publication of the Arthroscopy Association of North America and the International Arthroscopy Association.

[28]  R. Buchbinder,et al.  Magnetic resonance imaging, magnetic resonance arthrography and ultrasonography for assessing rotator cuff tears in people with shoulder pain for whom surgery is being considered. , 2013, The Cochrane database of systematic reviews.

[29]  J. Yoo,et al.  True anteroposterior (Grashey) view as a screening radiograph for further imaging study in rotator cuff tear. , 2013, Journal of shoulder and elbow surgery.

[30]  A. Boonstra,et al.  Cut-off points for mild, moderate, and severe pain on the visual analogue scale for pain in patients with chronic musculoskeletal pain , 2014, PAIN®.

[31]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[32]  J. Roy,et al.  Diagnostic accuracy of ultrasonography, MRI and MR arthrography in the characterisation of rotator cuff disorders: a systematic review and meta-analysis , 2015, British Journal of Sports Medicine.

[33]  R. Narasimhan,et al.  Prevalence of subscapularis tears and accuracy of shoulder ultrasound in pre-operative diagnosis , 2016, International Orthopaedics.

[34]  Tan Hwee Chye Andrew,et al.  Systematic review on risk factors of rotator cuff tears , 2017, Journal of orthopaedic surgery.

[35]  Michael J. Carroll,et al.  Tears of the Subscapularis Tendon: A Critical Analysis Review , 2017, JBJS reviews.

[36]  M. Bey,et al.  Characterization of Rotator Cuff Tears: Ultrasound Versus Magnetic Resonance Imaging. , 2017, Orthopedics.

[37]  Joseph S. Yu,et al.  ACR Appropriateness Criteria® Shoulder Pain-Atraumatic. , 2018, Journal of the American College of Radiology : JACR.

[38]  Li Yao,et al.  Weakly Supervised Medical Diagnosis and Localization from Multiple Resolutions , 2018, ArXiv.

[39]  Alexander D. Weston,et al.  What Does Deep Learning See? Insights From a Classifier Trained to Predict Contrast Enhancement Phase From CT Images. , 2018, AJR. American journal of roentgenology.

[40]  A. Hussain,et al.  Effectiveness Of Plain Shoulder Radiograph In Detecting Degenerate Rotator Cuff Tears. , 2018, Journal of Ayub Medical College, Abbottabad : JAMC.

[41]  Kenneth S. Lee,et al.  ACR Appropriateness Criteria® Shoulder Pain-Traumatic. , 2018, Journal of the American College of Radiology : JACR.

[42]  Expert Panel on Neurologic Imaging.  ACR Appropriateness Criteria® Shoulder Pain–Traumatic , 2018 .

[43]  A. Pripp,et al.  At a 10-Year Follow-up, Tendon Repair Is Superior to Physiotherapy in the Treatment of Small and Medium-Sized Rotator Cuff Tears. , 2019, The Journal of bone and joint surgery. American volume.

[44]  K. Chin,et al.  The accuracy of plain radiographs in diagnosing degenerate rotator cuff disease , 2019, Shoulder & elbow.

[45]  Enhua Wu,et al.  Squeeze-and-Excitation Networks , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.