Skip to content

Advances in Orthopedic Diagnosis Using Artificial Intelligence

In This Video

  • Foot & Ankle Research and Innovation Laboratory (FARIL) researchers have found that machine learning algorithms and artificial intelligence techniques on X-ray and CT scan increase the diagnosis of occult foot and ankle fractures with up to 95% accuracy
  • Soheil Ashkani-Esfahani, MD, and his FARIL colleagues are developing comprehensive software that can not only interpret the images of patients' injuries, but also consider other patients' data in its judgment for that specific condition or illness
  • FARIL takes a multidisciplinary approach, freely collaborating with specialists and researchers across the Mass General Brigham system and at other leading institutions in order to spread diagnostic knowledge and improve foot and ankle outcomes globally

In this video, Soheil Ashkani-Esfahani, MD, a physician investigator and fellow with the Department of Orthopaedic Surgery at Massachusetts General Hospital, discusses the latest diagnostic advances made by the hospital's Foot & Ankle Research and Innovation Laboratory (FARIL).


To give you background, as you know in the field of orthopedic surgery just as many other fields of medicine the early and accurate diagnosis of musculoskeletal or MSK conditions play a more important role compared to the treatment protocols. So faster and more precise diagnosis will lead to not only significant decrease in morbidities and mortalities in the patient but also a significant decrease in the costs and the burden on the health care system.

In the field of orthopedic surgery other than patient's data, notes, physical and lab examinations, imaging modalities play a significant role in detecting injuries and MSK conditions, sometimes not only in settings with limited resources, experience and expertise, but also in highly facilitated settings. Subtle and occult injuries, such as occult fractures, can easily be missed by the healthcare provider, the clinician or the physician who is examining the patient. We have thought about how we can enhance the accuracy of these diagnostic protocols by applying artificial intelligence, or AI, to these diagnostic methods, more specifically, applying image analysis methods and machine learning.

So we have started by working on X-rays both weight bearing and non-weight bearing X-rays and conventional CT scans as the two most abundant modalities in clinical settings all over the world. Surprisingly after applying machine learning algorithms and AI techniques to these imaging modalities we saw that, for example, regarding occult fractures of the foot and ankle, these factors are sometimes being missed by 50% in some clinical settings. And after we applied our AI protocols to these diagnostic modalities, we saw an increase up to 95%. As another example, we applied machine learning algorithms on CT scans, conventional CT scans, and X-rays, weight-bearing X-rays, for detection of unstable Lisfranc joint, mostly subtle cases. And we saw that our clinicians at Mass General had the accuracy of about 80% in detecting these injuries but after applying machine learning algorithms we had an accuracy of 96% in detecting subtle Lisfranc instability.

We tried to also include portable ultrasound to our imaging modalities included in our project. What is so special about portable ultrasound is that it is easily accessible and portable and it can be used at the site of injury like the battlefield or the sports field, but a gap here is not all clinicians are familiar with the use, and how to use, portable ultrasound. So our first goal is to educate them. Our ultimate goal is to create software that includes all necessary imaging modalities that can be used for diagnosis of musculoskeletal or MSK conditions. For that goal, we are also going to include patients' data including notes, physical and lab examinations, and also operative notes. We have also started a collaboration with Harris Orthopaedic Lab at Mass General, Dr. Orhun Muratoglu and his colleagues, working on weight-bearing CT scan which is a newly emerging modality, especially in the field of orthopedic surgery, working on different subtle injuries, mostly in the field of foot and ankle, and even in those projects we saw a significant increase in the accuracy of diagnosis.

We have a started a collaboration with Harvard Global Orthopaedic Collaborative Group, Dr. George Dyer, Dr. Agarwal-Harding and Dr. Kwon, that helped us in creating educational content that are already online and free to use, particularly for healthcare providers in poor and low income countries to educate them and to update them about the new and currently used diagnostic and treatment methods. Thanks to my engineer colleague Reza Yazdi and also the director of the lab, Dr. Lubberts and also the chief of the division Dr. DiGiovanni, and my other colleagues and teammates, we have created a prototype of the software with the aim of creating the comprehensive software that can not only interpret the images of the patients but also consider other patients' data in its judgment for that specific condition or illness. The ultimate goal here is not only to enhance the diagnostic methods for orthopedic conditions but also to provide educational content and slides to update our colleagues, healthcare providers, those who are near or far from us, about the currently used diagnostic and treatment methods which leads to democratizing knowledge, expertise and experience throughout the world.

Something which is so special about doing research at Mass General is not only you are provided with a robust database, with a robust facility in terms of research, but also you are in a think-tank with a bunch of professors, mentors and experts who are otherwise humble and so friendly and can guide you through your journey in the field of research.

Learn more about the Foot and Ankle Research and Innovation Laboratory at Mass General

Refer a patient to the Department of Orthopaedic Surgery


Massachusetts General Hospital researchers have created an automated musculoskeletal image interpretation system that uses artificial intelligence to significantly improve orthopedic diagnosis.


Based on a systematic review, researchers in Massachusetts General Hospital's Foot and Ankle Research and Innovation Lab believe the commonly used threshold of 2.0 mm for distal tibiofibular diastasis may lead to overdiagnosis of syndesmotic instability during arthroscopy, and they suggest more evidence-based cutoffs.