A comparative study of morphological and other texture features for the characterization of atherosclerotic carotid plaques
| Autor: | Christodoulou, Christodoulos I., Kyriacou, Efthyvoulos C., Pattichis, Marios S., Pattichis, Constantinos S., Nicolaïdes, Andrew N. |
|---|---|
| Přispěvatelé: | Pattichis, Constantinos S. [0000-0003-1271-8151], Pattichis, Marios S. [0000-0002-1574-1827], Kyriacou, Efthyvoulos C. [0000-0002-4589-519X] |
| Rok vydání: | 2003 |
| Předmět: | |
| Zdroj: | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Lect. Notes Comput. Sci. |
| Popis: | The extraction of features characterizing the structure of atherosclerotic carotid plaques, obtained by high-resolution ultrasound imaging is important for the correct plaque classification and the estimation of the risk of stroke. In this study morphological features were extracted and compared with the well-known texture features spatial gray level dependence matrices (SGLDM), gray level difference statistics (GLDS) and the first order statistics (FOS) for the classification of 330 carotid plaques. For the classification the neural self-organizing map (SOM) classifier and the statistical k-nearest neighbor (KNN) classifier were used. The results showed that morphological and other texture features are comparable, with the morphological and the GLDS feature sets to perform slightly better than the SGLDM and the FOS features. The highest diagnostic yield was achieved with the GLDS feature set and it was about 70%. © Springer-Verlag Berlin Heidelberg 2003. 2756 503 511 Cited By :8 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|