LABERINTO at ImageCLEF 2012 Medical Image Retrieval Task
Mariano Crespo, Jacinto Mata Vázquez, Manuel Jesús Maña López · 2012
Abstract. This paper shows the experimentation and the results obtained for LABERINTO research group at the ImageCLEF 2012 medical task. We focus our work on image retrieval based on textual information related to the image. Last year we demonstrated that query expansion exploiting the hierarchical structure of the MeSH descriptors achieved a significant improvement in image retrieval systems. This year our goal is to improve the results obtained last year adding a relevance factor to the query terms. In addition, we have developed a new strategy combining the expansion strategy based on the hierarchical MeSH structure with another expansion strategy very popular among researchers in this field, where the query terms are expanded using MMTx program. The experiments carried out have shown that a relevance factor for the query terms achieves a significant improvement for the results of the different expansion strategies.