Unboxing AI | Friday Webinars @ 11 AM ET | Dr Leonardo K Bittencourt, University Hospitals (July 12) | Kevin Harris, CureMetrix (July 19) | Dr Ryan Lee, Jefferson Health (July 26) | Register Now
  • 2020-01-13

Are orthopaedic surgeons as good as radiologists at detecting glenoid labral lesions? A comparative study


In the current era of information age, medical education and training has undergone a rapid shift in the style and structure. Radiology training, due to its nature of complete digitization of information is in the forefront of this shift. This change offers unique opportunity to educate radiologists as well as clinicians in a practical way to enhance their practice. In this work, we propose to study if it possible to impart capsule training to general radiologists and clinicians to improve their interpretation skills in specific pathologies like glenoid labral tears. We also propose to compare the performance of a general radiologist with a shoulder surgeon in detecting glenoid labral tears.


Seventy-two cases of MRI of the shoulder (Axial, Coronal & Sagittal FSPD, Sagittal T1) with diagnosis of glenoid labral tear on clinical report were extracted from PACS, anonymised and randomised. A sub-specialist musculoskeletal radiologist independently read the scans to establish ground truth for the comparative study. A shoulder surgeon and a general radiologist, each with 12 years’ experience in their respective fields were given a capsule course of training on imaging findings of glenoid labral tears with cases including all common variants. The training cases didn’t include any of the cases from the test set. After the training, they independtly read the test scans and reported presence or absence of labral tears in 4 quadrant (anterosuperior, posterosuperior, anteroinferior, posteroinferior). The results were compared with the ground truth and percentage observed agreement with ground truth was calculated.


Out of the 72 cases, 5 cases had 270º tears, 24 cases had anteroinferior glenoid labral tears,14 had posteroinferior tears, 25 had posterosuperior tears and two had anterosuperior tears. The percentage observed agreement for SLAP (posterosuperior) and Bankart’s lesion (anteroinferior) was 58% and 72% for the radiologist and 46% and 58% for the orthopaedic surgeon. The percentage agreement for anterosuperior and posteroinferior labrum was similar for the radiologist (61% and 69%) and the orthopaedic surgeon (60% and 69%).


Surprisingly, the general radiologist performed better than the sub-specialist orthopaedic surgeon paving the way for training programs in MRI reading for orthopaedic surgeons. Note that our sample size of readers was very small, and the study should be replicated with a larger number of readers. Many orthopaedic surgeons, especially in the developing world, where there is a paucity of radiologists, choose to read their scans themselves. We demonstrate that as far as possible an opinion from a radiologist should be obtained.

The EPOS can be viewed here: http://dx.doi.org/10.26044/ecr2020/C-13084

Unlock the potential of CARPL platform for optimizing radiology workflows

Talk to a Clinical Solutions Architect