Classification of movement patterns of subviral particles using a support vector machine

This work deals with the classification of subviral particle movement. Preparatory, tracks were humanly assigned into classes. Coordinates, obtained by a subviral particle tracker and metadata were used to determine parameters that quantify movement patterns in space and time. The information content of the parameters was condensed by principal component analysis. A scatter plot of the components including the manual mapping shows a clear separation of the tracks and confirms the hypothesis. Using this, an algorithm was developed to classify new tracks automatically. Data classification was performed by a support vector machine trained on detected and human-assigned subviral particle tracks.