Abstract
The goal of this work is to investigate the performance of classical methods for feature description and classification, and to identify the difficulties of the ImageCLEF 2010 modality classification subtask. In this paper, we describe different approaches based on visual information for classifying medical images into 8 different modality classes. Since within the same class, images depict very different objects, we focus on global descriptors such as histograms extracted from scale-space, log-Gabor and phase congruency feature images. We also investigated different classification approaches based on support vector machines and random forests. A grid-search associated to a 10 folds cross-validation has been performed on a balanced set of 2390 images to find the best hyperparameters for the different models we propose. All experiments have been conducted with MATLAB on a Workstation with Intel Duo Core 3.16 Ghz and 4Gb of RAM. Our approach based on simple SVM and random forests give best performance and achieve respectively an overall f-measure of 74:13% and 73:59%.
Originalsprache | Englisch |
---|---|
Fachzeitschrift | CEUR Workshop Proceedings |
Jahrgang | 1176 |
Publikationsstatus | Veröffentlicht - 2010 |
Veranstaltung | 2010 Cross Language Evaluation Forum Conference, CLEF 2010 - Padua, Italien Dauer: 22 Sept. 2010 → 23 Sept. 2010 |