April 24, 2025
The AI Future of Orthopaedics Imaging
Pitt researchers are collaborating to harness the power of advanced artificial intelligence algorithms that are trustworthy, ethical by design and explainable.
Designs on Aging-Ready
By Strategic Communications
In orthopaedic care, decisions often happen at the intersection of imaging, clinical judgment and time. A plain radiograph or MRI can guide the assessment of cartilage loss or meniscus damage and might determine whether a patient should undergo surgery. A postoperative clinical trajectory may vary because early signs of complications may be subtle and easily missed when health care systems are overloaded.
But researchers from the Pitt HexAI Research Laboratory at the University of Pittsburgh’s School of Health and Rehabilitation Sciences (SHRS) are collaborating with colleagues from the School of Medicine and the Intelligent Systems Program to harness the power of advanced artificial intelligence (AI) algorithms in orthopaedic imaging for preoperative planning and postoperative assessments that are trustworthy, ethical by design and explainable.
Ahmad P. Tafti, assistant professor of health informatics and director of the HexAI Research Laboratory, is leading the charge. “If AI is going to support clinical decisions, it must be reliable, transparent, responsible and equitable across all patients, institutions and real-world medical imaging conditions,” says Tafti.
In Tafti’s HexAI lab, researchers design and validate AI tools that can explain their decisions, quantify uncertainty when confidence is low and identify disparities.
Through AI, Tafti and his team can mitigate inherent biases and disparities, which can profoundly impact patient care and treatment plans.
Eliminating biases
In a recent study, for example, a hybrid AI model was trained with four distinct bias mitigation strategies to ensure fairness and accuracy in knee and hip image segmentation. According to Tafti, these strategies address and minimize potential biases, enhance the accuracy and reliability of the segmentation results and help the health care team identify complications that may occur after total hip or knee replacement surgery.
In the past, the algorithms used in medical image segmentation were inconsistent. They did not consider variabilities in anatomy, joint structures and bone densities across various demographic groups. For example, they did not account for the fact that female pelvises are wider than male, African Americans exhibit higher bone density than their white counterparts, and older patients have less bone mineral density than younger ones.
Through AI, Tafti and his team can mitigate inherent biases and disparities, which can profoundly impact patient care and treatment plans.
“To the best of our knowledge, this study represents the first initiative to build, train and validate deep learning-enabled knee and hip bony anatomy segmentation methods that generate accurate and unbiased results across diverse genders, races and ages,” says Tafti.
In addition to fair AI-powered orthopaedic image segmentation, Tafti’s work spans a wide research agenda, ranging from AI-powered knee osteoarthritis analysis to hip and knee arthroplasty surgical risk prediction, and mobile technology enabling remote follow-up and rehabilitation.
Across these efforts, one theme stays constant: the Pitt HexAI lab builds AI systems physicians, surgeons—and patients—can trust.
“Pitt HexAI has created a rare and powerful bridge between cutting-edge AI research and real surgical and imaging workflows in orthopaedics. That translational focus is what the field needs,” says MaCalus V. Hogan, David Silver Professor and chair, Pitt’s Department of Orthopaedic Surgery.
“These are not models developed in isolation; they are designed around how orthopaedics is practiced and to improve the care for our patients,” continues F. Johannes Plate, associate professor of orthopaedic surgery, Pitt School of Medicine.
Brian J. McGrory, Tufts University School of Medicine clinical instructor and external collaborator at the Pitt HexAI lab, agrees. "As a surgeon, I value innovations that translate into better patient care. Pitt HexAI is developing AI tools that are not only technically impressive, but also clinically meaningful, and designed to integrate into orthopaedic workflows.”
“At Pitt HexAI, we bring advanced AI research into orthopaedics, by developing tools that are transparent, fair and practical for clinical use. By prioritizing trust and responsible innovation, we aim to ensure these technologies meaningfully benefit both clinicians and patients,” says Pitt HexAI Lab Manager Nicole Myers.
Team science
SHRS Professor of Physical Therapy James Irrgang has devoted his career to studying musculoskeletal injury prevention and recovery. For the past three years, he has been collaborating with Tafti to find ways to help patients avoid re-injury after anterior cruciate ligament knee surgery.
They’ve learned that the shape and slope of the shin bone determine how the knee moves, so they are currently using hundreds of data sets and sophisticated machine learning tools to understand how patients with differing tibial slopes can avoid re-injury. The benefits are significant. Patients may be able to return to their sports and other activities sooner and eliminate the need for additional surgeries and related complications down the road, such as arthritis and disability. Furthermore, health care costs are reduced.
“As a physical therapist, I have always collaborated with orthopaedic surgeons and specialists in bioengineering and biomechanics because we each bring a unique perspective to solving orthopaedic problems,” says Irrgang. “But as clinicians, we do not have expertise in AI. Ahmad Tafti and the HexAI lab are helping us arrive at solutions by automating processes and improving accuracy and reliability. Without AI, the process would have been tedious, time-consuming and inaccurate. With AI, it’s a win-win for both patients and health care providers.”