Long-Term Bilateral Neuromuscular Function and Knee Osteoarthritis after Anterior Cruciate Ligament Reconstruction

Abstract

Neuromuscular function is thought to contribute to posttraumatic osteoarthritis (PTOA) risk in anterior cruciate ligament (ACL)-reconstructed (ACLR) patients, but sensitive and easy-to-use tools are needed to discern whether complex muscle activation strategies are beneficial or maladaptive. Using an electromyography (EMG) signal analysis technique coupled with a machine learning approach, we sought to: (1) identify whether ACLR muscle activity patterns differed from those of healthy controls, and (2) explore which combination of patient outcome measures (thigh muscle girth, knee laxity, hop distance, and activity level) predicted the extent of osteoarthritic changes via magnetic resonance imaging (MRI) in ACLR patients. Eleven ACLR patients 10-15 years post-surgery and 12 healthy controls performed a hop activity while lower limb muscle EMG was recorded bilaterally. Osteoarthritis was evaluated based on MRI. ACLR muscle activity patterns were bilaterally symmetrical and differed from those of healthy controls, suggesting the presence of a global adaptation strategy. Smaller ipsilateral thigh muscle girth was the strongest predictor of inferior MRI scores. The ability of our EMG analysis approach to detect meaningful neuromuscular differences that could ultimately be related to thigh muscle girth provides the foundation to further investigate a direct link between muscle activation patterns and PTOA risk.

Keywords: anterior cruciate ligament; artificial intelligence; electromyography; neuromuscular function; osteoarthritis; reconstruction; wavelet analysis.

Figure 1

Hop phase determination. The hop activity was divided into three phases based on a combination of EMG, MoCap, and force data. t₀ = start of trial based on increased EMG signal amplitude; t₁ = take-off based on vertical ankle velocity; t₂ = contact based on ground rection force; and t₃ = peak vertical ground reaction force. Figure adapted from [28].

Figure 2

Example of EMG wavelet visualization with 5 frequency bands f1–5. (A). EMG signals are band-pass filtered according to their central frequency (vertical dashed lines). (B). Signal frequencies and amplitudes determine the signal intensity. (C). Examples of wavelet objects that represent frequency content (e.g., height of shape according to y-axis), intensity (e.g., color), and time (e.g., shape width according to x-axis). The dashed blue lines that span (A–C) illustrate how the combination of wavelet components is represented in a wavelet object. Figure adapted from [28].

Figure 3

Average wavelet muscle activity patterns. The horizontal axis represents each activity phase’s normalized time. The y axis in each figure represents the activation frequency (0 to 1000 Hz), and the contour intensity is the normalized activity intensity scaled from 0 to 1 mV/mV with the brighter colors representing a higher relative activation intensity. The cyan vertical lines delineate take-off, airborne, and landing hop phases. Significantly classified patterns between the ACLR_Sx and Control_Idx limbs [28] and ACLR_Sx and ACLR_Contra (e.g., Table 4b)) are highlighted by orange and green boxes, respectively. GM: Gastrocnemius medialis, GL: gastrocnemius lateralis, TA: tibialis anterior, VM: vastus medialis, RF: rectus femoris, BF: biceps femoris, and ST: semitendinosus.