Article Text
PDF
Abstract
Introduction Self-Organising Maps (SOM) are a type of Artificial Neural Network (ANN) model used to visualise multidimensional data [Kohonen-00]. SOM provides a correspondence between the original N-dimensional space and a twodimensional space [Haykin-08]. In the case presented in this work, this means that two subjects with similar values of the different variables should appear near in the two-dimensional space [Kohonen-00], [Haykin-08].
Method In order to test this method, 15 athletes were selected. Subjects’ measurements were performed in three different periods: before the exercise test, immediately after performing the exercise and twenty-four hours later. The control measures were the left leg, trained without occluding and the comparisons were made with the right leg, trained with occlusion. Changes in the sural triceps perimeters were measured. Furthermore, medial calf cross section and cross section of the Achilles tendon were measured using Doppler Ultrasound. The method consists in the performing of exercise series with very light loads (30% of 1RM) with occlusion in the zone to be measured.
Results Variables Body Mass Index (BMI), age, differences between 24h later and pre-training exercise test for right perimeter (ΔPdT24-T0), right calf in pretraining (ΔGdT24-T0) and right tendon in pretraining (ΔTdT24-T0) were chosen to be visualised on the map. Figure 1 shows the map structure, where each neuron is represented by a hexagon. Coloured area inside each hexagon is proportional to the number of subjects assigned in each zone. The more coloured area has more number of subjects. Figure 2 shows the “components plane” where the analysed variables are visualised with a colour code.
Areas of the map where the different subjects are located
Analysed variables in the self-organising map
Discussion The Self-Organising Map shows two different areas. Upper zone: In the upper area of the map, patterns with a higher body mass index are shown. These patterns match with youngest athletes. In both perimeter and calf, an increase in length is observed, whereas in tendon, a decreased in length is produced. Lower zone: On the other hand, in the lower area of the map, opposing trends in the physical magnitudes are observed. This zone of the maps represents subjects with lower body mass index and older age, who show a smaller increase in leg perimeter and calf length as well as a larger increase of the tendon length.
ReferenceS [Kohonen-00]; Self-Organizing Maps, T. Kohonen, Springer, 2000.
[Haykin-08]; Neural Networks and Learning Machines, S. Haykin, Prentice Hall, 2008