Osteoarthritis (OA) affects over 32.5 million American adults and remains a leading cause of pain, disability, and joint dysfunction worldwide. While interventions such as total knee arthroplasty and offloading braces can mitigate symptoms, they do not restore articular cartilage once degeneration has occurred. We developed a Light Soft Actuated Bionic Regenerative Engineering (Light SABRE) brace system, a pneumatic actuator–driven orthotic that modulates the forces around the knee through a controlled distraction mechanism. By precisely regulating joint mechanics, the device is intended to create a biomechanical environment that encourages cartilage regeneration. We hypothesized that the Light SABRE brace system would promote cartilage regeneration and ultimately reverse OA pathology in a rabbit model, generating data to train machine learning models to predict outcomes and guide personalized brace optimization.
Methods
1. Brace Design and Custom Pneumatic Actuator Prototyping: Light SABRE brace system dimensions were captured via 3D laser scanning of male New Zealand White Rabbits, and brace components were modeled in SolidWorks. The sleeve, made of polyester & neoprene, was paired with the 3D-printed load-bearing exoskeleton and secured to the limb using hook-and-loop fasteners. Custom soft Pneumatic actuators were fabricated with Preform® Elastic 40A® V2 resin. Finite Element Analysis in ANSYS Workbench, load-bearing, and tensile testing were performed before actuator integration.
2. Evaluation of Light SABRE brace system in an OA-rabbit Model: Osteoarthritis was induced in male New Zealand White rabbits with collagenase II injections, confirmed clinically and radiographically and rabbits were randomized to control (OA), Light SABRE (brace), or fixator groups. Rabbit knee joint unloading was done either by brace system or surgically by external fixator. Animals were sacrificed at 6 or 10 weeks. Analyses included histology (H&E, Safranin O, Masson’s Trichrome, Alcian Blue), immunohistochemistry (Collagen I/II), International Cartilage Repair Society (ICRS) cartilage scoring, nanoindentation, and cytokine profiling of serum and synovial fluid.
3. Machine Learning Model Development: Models were developed to predict osteoarthritis cartilage regeneration outcomes using mechanobiological data collected from the Light SABRE brace system. Biomechanical, histological, and cytokine features were used to train regression models (XGBoost & Random Forest). Model performance was evaluated using R² and Root Mean Squared Error metrics, and feature importance analyses were conducted to identify dominant predictors of ICRS Mean scores.
Results
1. Light SABRE brace system actuators provided effective mechanical offloading. Finite Element Analysis simulations showed non-uniform strain and stress, peaking at 15.8% strain and 25.2 MPa at chamber ends and centers, confirming structural integrity under physiological loads. Radiographic measurements demonstrated that the assembled Light SABRE brace system significantly increased joint space width (0.72mm → 1.60mm).
2a. Light SABRE brace system promoted cartilage regeneration and restored joint structure. H&E confirmed smoother surfaces, intact tidemarks, and less remodeling; Safranin O and Alcian Blue demonstrated greater proteoglycan/glycosaminoglycan retention, and Masson’s Trichrome revealed more organized collagen and intact subchondral structure. ICRS scores increased from 6 to 10 weeks. At 10 weeks the Light SABRE group scored highest across all six domains, reflecting progressive cartilage repair and supporting OA reversal. Immunohistochemistry showed Collagen I confined to subchondral cortical layers and uniform Collagen II, also indicating improved cartilage composition.
2b. Light SABRE brace system supported structural recovery and promoted cartilage regeneration. At 6 weeks, both the brace and fixator groups showed higher reduced modulus than controls. By 10 weeks, the brace group exhibited the greatest modulus and lowest indentation depths, with biomechanical properties approaching those of healthy cartilage.
2c. Light SABRE brace system effectively modulated inflammatory and matrix remodeling markers. Pro-inflammatory cytokines (IL-1β, TNF-α, MIP-1β) in synovial fluid were reduced, while anti-inflammatory markers (IL-1RA, IL-13) were elevated. Serum MMP-9 remained low, and TSP-2 was elevated in brace-treated rabbits. RANTES and TWEAK-R were also suppressed, supporting reduced inflammation and a pro-regenerative joint environment.
3. The machine learning model corresponded to histologic outcomes, with XGBoost achieving the highest predictive performance (R² ~1.0, RMSE ~0.000114). Feature importance identified matrix and histological features—cell viability, matrix structure, subchondral bone, and surface integrity—as dominant predictors, while cytokine variables showed indirect effects not captured by direct correlations. Despite cytokines' biological relevance, this suggests that the Light SABRE brace system promotes regeneration primarily through mechanotransductive and matrix remodeling pathways.
Conclusion & Future Directions
The Light SABRE brace system promoted cartilage regeneration and restored joint structure in the OA rabbit model with improved ICRS scores, organized matrix, and biomechanical properties approaching healthy tissue. Inflammatory and catabolic markers were suppressed, creating a pro-regenerative joint environment. Mechanobiological data from the brace system was effectively used by ML models to predict regenerative outcomes, highlighting the potential for personalized optimization. Future integration of Bluetooth sensors and closed-loop feedback will enable real-time adjustment of joint loading to further enhance cartilage recovery. By combining dynamic offloading, biological stimulation, and machine learning–driven customization, the Light SABRE brace system represents a transformative approach to regenerating cartilage and reversing osteoarthritis progression.