Differences in Pedaling Technique in Cycling: A Cluster Analysis

Abstract

The purpose of this study was to employ cluster analysis to assess if cyclists would opt for different strategies in terms of neuromuscular patterns when pedalling at the power output of their second ventilatory threshold (POVT2) compared to cycling at their maximal power output (POMAX). Twenty athletes performed an incremental cycling test to determine their power output (POMAX and POVT2; first session) and pedal forces, muscle activation, muscle-tendon unit length, and vastus lateralis architecture (fascicle length, pennation angle, and muscle thickness) were recorded (second session) in POMAX and POVT2. Athletes were assigned to two clusters based on the behaviour of outcome variables at POVT2 and POMAX using cluster analysis. Clusters #1 (n=14) and #2 (n=6) showed similar power output and oxygen uptake. Cluster #1 presented larger increases in pedal force and knee power than Cluster #2, without differences for the index of effectiveness. Cluster #1 presented less variation in knee angle, muscle-tendon unit length, pennation angle and tendon length compared to Cluster #2. However, Clusters #1 and #2 showed similar muscle thickness, fascicle length and muscle activation. When cycling at POVT2 versus POMAX, cyclists could opt for keeping a constant knee power and pedal force production, associated with an increase in tendon excursion and a constant fascicle length. In conclusion, increases in power output lead to greater variations in knee angle, muscle-tendon unit length, tendon length, and pennation angle of vastus lateralis for a similar knee extensor activation and smaller pedal force changes in cyclists from Cluster #2 compared to Cluster #1.

Untitled

“…To collect kinematic athletes' clearance data, twenty markers were attached to the body of each participant for digitization. Body markers, using the Hanavan model modified by De Leva (1996), were digitized using the video-based data analysis system SkillSpector ® 1.3.2 [Odense SØ -Denmark], (Bini, Jacques, Lanferdini, & Vaz, 2015;Lanferdini et al, 2016;Mkaouer, 2018;Mkaouer, Chaabene, Amara, Negra, & Jemni, 2018;Mkaouer, Jemni, Amara, Chaabène, & Tabka, 2013). Similarly, the body segments' COM was computed using the Hanavan model modified by De Leva (1996).…”

Application of Kohonen Neural Network in Sports Cluster

“…Commonly used measures are similarity and dissimilarity measures, which quantitatively describe the degree of similarity or dissimilarity between two data objects or clusters. Or the greater the similarity between clusters, the smaller the dissimilarity; conversely, the smaller the similarity, the greater the dissimilarity [17,18]. However, most existing clustering algorithms often use dissimilarity to represent the similarity measure and use it as a measure of computing data objects.…”