Revolutionizing Shoulder Instability Research with 3D MRI Imaging and AI

Advanced MRI imaging and AI learning models are transforming the study of shoulder instability, potentially improving treatment for a debilitating cause of injury in military personnel. 

A service member lies face up on a gurney, head first, entering an MRI machine. Red laser lines are visible across their head and chest.
MIRROR Project 6, titled “Early Identification of Glenohumeral Pathomorphology to Prevent Recurrent
Shoulder Instability,” studies shoulder instability to inform optimal treatment measures through integration of
high-resolution 3D models from MRI data and AI machine learning tools. (U.S. Air Force photo by Airman
1st Class Zachary Foster)

June 24, 2025 by Hadiyah Brendel

Shoulder instability—when the shoulder joint slips out of place or dislocates—is a common and serious injury among service members, affecting their ability to stay mission-ready. It's one of the top three musculoskeletal injuries in the military. The Uniformed Services University’s (USU) Musculoskeletal Injury Rehabilitation Research for Operational Readiness (MIRROR) program is working to improve care and prevention through research, education, and training focused on more than 40 types of these common joint and muscle injuries. 

MIRROR Project 6, titled “Early Identification of Glenohumeral Pathomorphology to Prevent Recurrent Shoulder Instability,” one of 84 current MIRROR research projects, studies shoulder instability to inform optimal treatment measures through integration of high-resolution 3D models from MRI data and AI machine learning tools. 

This project focuses on detecting early signs of abnormal shoulder structure to help prevent repeated shoulder dislocations in service members. Specifically, it looks at the main shoulder joint–the glenohumeral joint—where the upper arm bone (the "ball") fits into a shallow socket in the shoulder blade. Identifying issues in this area early on can help reduce the risk of ongoing shoulder problems. 

“Think of the shoulder joint like a golf ball and tee,” says Dr. Matthew Bradley, professor and chair of the USU School of Medicine’s Department of Surgery and project manager for the research. The ball is cupped by the tee, providing stability, and together this ball-and-socket structure allows for a wide range of shoulder motion. But, he notes, if you chip the tee, the ball can fall off. Bradley further explains, “Whenever the shoulder bone pops out, there’s injury to the soft tissue structures, and also some of the bone chips off.”

The research involves examining the structural changes and abnormalities within the glenohumeral joint, including issues with the bone, cartilage, ligaments, and the surrounding soft tissues, which result from shoulder instability events. 

A three-panel image showing different 3D anatomical models of a human shoulder joint from various perspectives.
Fig 8. Views of a 3D model reconstruction developed from a 3D magnetic resonance imaging scan, completed
using the 3D Slicer application. Present are the scapula (green), humerus (blue), and clavicle (yellow) on a
left shoulder, in the anterior (A), posterior (B), and en face glenoid (C) perspectives. (3D, 3-dimensional; A,
anterior; I, inferior; L, left; P, posterior; R, right; S, superior.) (Figure courtesy of MIRROR)

This loss of muscle integrity and bone leads to a high-rate of re-injury. Up to 80% of civilian individuals experience re-injury to the shoulder. For military trainees, the risk is significantly higher, being 40 times more likely for a recurrence. Understanding the glenohumeral pathomorphology in service members with shoulder instability is crucial for addressing this pervasive problem.

Traditional methods of diagnosis and treatment have limitations. For a first-time shoulder instability event, both surgical and non-surgical management are common, with varied success rates. A lack of definitive evidence exists for optimal rehabilitation methods. And recurring instability heightens risks, leading to more bone loss, surgical failures, and post-traumatic arthritis. MIRROR’s research seeks to develop military-specific clinical practice guidelines (CPGs) to optimize treatment, improve patient outcomes, and maximize return to duty for service members. 

CT scans are commonly used for 3D visualization and bone loss assessment. However, they involve radiation exposure, additional time, and higher costs. Alternatively, MRIs excel in generating a 2D soft-tissue assessment, but are generally less effective than a CT scan for detecting bone loss.

The sequence technique developed with this research utilizes a 3D-slicer, software for medical image viewing and reconstruction, to segment the scapula and humerus from MRI datasets, creating realistic 3D models of the shoulder joint. The model allows for more in-depth analysis to assess bone loss and helps determine between surgical and non-surgical pathways, all while avoiding radiation exposure associated with CT scans. 

A collaboration with Duke University stemming from Dr. Jonathan Dickens, a shoulder, knee and sports medicine surgeon at Duke and former principal investigator for the project who still helps oversee it, has led to the creation of an AI learning model that automates the 3D models from the MRI scans. This automation significantly reduces the time required for model generation, a process that currently takes Jacob Dowe, the research coordinator, approximately three hours to complete manually. This intricate work is further supported by a dedicated research team, including Audrey Hartin and Dr. Judd Robins both at the U.S. Air Force Academy (USAFA), and Dr. Ken Cameron at the U.S. Military Academy (USMA) at West Point.

The AI will also enable large-scale comparative analyses, identifying patterns and answering investigative questions such as whether the shape and symmetry of the glenohumeral affect risk. Bradley says they are curious if factors such as an oval shape for their shoulder joint over a circular shape could lead to more instability events. Additionally, the 3D models will aid in evaluating whether pre-existing methods for assessing bone loss—such as comparing the injured shoulder to the uninjured shoulder—remain an appropriate measure for quantifying the extent of bone loss. 

A U.S. Air Force physical therapist, wearing a mask, helps a male patient in military fatigues extend his arm during a rehabilitation session.
If a 3D model reveals less bone loss than initially perceived on 2D scan, it might indicate physical therapy as a
more viable option and potentially avoid unnecessary surgery. (U.S. Air Force photo by Airman 1st Class
Makensie Cooper)

Project 6 is a longitudinal study, aiming to study service members over a period of 10 to 20 years. The goal is to enroll participants recruited within their first year of entering the USAFA, USMA, and the U.S. Naval Academy (USNA). This long-term approach allows researchers to track shoulder health and injury development over time, hoping to gain understanding into the long-term effects of shoulder instability on service members’ careers. Since August 2020, approximately 90 participants have been enrolled in the study and researchers have taken over 1,800 scans.

Bradley expresses optimism about the project’s current juncture, stating, “We’re at a really exciting point.” He believes based on the work already completed and the discoveries made, the research has the potential to fundamentally “change the course of care” for anyone who experiences a shoulder instability event, regardless if surgery is ultimately required. 

While the specific aim of differentiating surgical from non-surgical candidates based on 3D models is still in the long-term surveillance phase, Bradley highlights an immediate impact: the implementation of this advanced MRI sequence itself. He explains that simply incorporating this 3D modeling into standard MRI scans can yield significantly more information than traditional 2D images. This enhanced visualization could directly inform surgical decision-making, allowing surgeons to more precisely identify the extent of bone loss and select the most appropriate treatment pathway from various surgical options. Conversely, if a 3D model reveals less bone loss than initially perceived on 2D scan, it might indicate physical therapy as a more viable option and potentially avoid unnecessary surgery. 

“The advancements driven by MIRROR Project 6 are anticipated to enhance force medical readiness,” said Dr. Brad Isaacson, chief of Research and Operations for MIRROR, associate professor, Department of Physical Medicine and Rehabilitation at USU, and principal investigator, Geneva Foundation. “By optimizing the management of shoulder instability, this research aims to improve patient outcomes, maximize the rate at which warfighters can return to duty, and reduce the risk of re-injury among service members.”