Kiapour, Ata M.

AbstractSupervised learning is the most common form of machine learning utilized in medical research. It is used to predict outcomes of interest or classify positive and/or negative cases with a known ground truth. Supervised learning describes a spectrum of techniques, ranging from traditional regression modeling to more complex tree boosting, which are becoming increasingly prevalent as the focus on “big data” develops. While these tools are becoming increasingly popular and powerful, there is a paucity of literature available that describe the strengths and limitations of these different modeling techniques. Typically, there is no formal training for health care professionals in the use of machine learning models. As machine learning applications throughout medicine increase, it is important that physicians and other health care professionals better understand the processes underlying application of these techniques. The purpose of this study is to provide an overview of commonly used supervised learning techniques with recent case examples within the orthopedic literature. An additional goal is to address disparities in the understanding of these methods to improve communication within and between research teams.

Background: Despite basic characterization of the loading factors that strain the anterior cruciate ligament (ACL), the interrelationship(s) and additive nature of these loads that occur during noncontact ACL injuries remain incompletely characterized. Hypothesis: In the presence of an impulsive axial compression, simulating vertical ground-reaction force during landing (1) both knee abduction and internal tibial rotation moments would result in increased peak ACL strain, and (2) a combined multiplanar loading condition, including both knee abduction and internal tibial rotation moments, would increase the peak ACL strain to levels greater than those under uniplanar loading modes alone. Study Design: Controlled laboratory study. Methods: A cadaveric model of landing was used to simulate dynamic landings during a jump in 17 cadaveric lower extremities (age, 45 ± 7 years; 9 female and 8 male). Peak ACL strain was measured in situ and characterized under impulsive axial compression and simulated muscle forces (baseline) followed by addition of anterior tibial shear, knee abduction, and internal tibial rotation loads in both uni- and multiplanar modes, simulating a broad range of landing conditions. The associations between knee rotational kinematics and peak ACL strain levels were further investigated to determine the potential noncontact injury mechanism. Results: Externally applied loads, under both uni- and multiplanar conditions, resulted in consistent increases in peak ACL strain compared with the baseline during simulated landings (by up to 3.5-fold; P ≤ .032). Combined multiplanar loading resulted in the greatest increases in peak ACL strain ( P < .001). Degrees of knee abduction rotation ( R 2 = 0.45; β = 0.42) and internal tibial rotation ( R 2 = 0.32; β = 0.23) were both significantly correlated with peak ACL strain ( P < .001). However, changes in knee abduction rotation had a significantly greater effect size on peak ACL strain levels than did internal tibial rotation (by ~2-fold; P < .001). Conclusion: In the presence of impulsive axial compression, the combination of anterior tibial shear force, knee abduction, and internal tibial rotation moments significantly increases ACL strain, which could result in ACL failure. These findings support multiplanar knee valgus collapse as one the primary mechanisms of noncontact ACL injuries during landing. Clinical Relevance: Intervention programs that address multiple planes of loading may decrease the risk of ACL injury and the devastating consequences of posttraumatic knee osteoarthritis.