Use of medical ontologies to improve segmentation and medical visualization processes

Computer -assisted medical applications offer the expert radiologist or medical specialist a tool for decision support when emitting a diagnosis. The process of segmentation or identification of anomalous tissues based on the use of medical images allows the specialist to have a better visualization of tissues and organs of interest. We believe that based on the structured and defined knowledge of an ontology it is possible to perform semantic segmentations from medical images in Dicom format. DICOM allows to manage metadata regarding the study described in an image, which coupled with existing medical ontologies can automatically contribute to the segmentation process. Semantic annotations using medical ontologies defined as Radx and the Foundational Model of Anatomy (FMA), Ontology of Biomedical Reality (OBR), Open Biomedical Ontologies (OBO) among many others, conceptually allow the annotation of the areas that surround particular tissues. The application of these annotations can improve the segmentation process, the quality of representation of the volume and the applied visualization technique. Many researchers have defined their own ontologies depending on the medical modality (TC1, RM2, US3, etc.) used and of the particular pathology. However, it is a scientific challenge to identify an ontology that covers the totality of medical knowledge. Our intention is not to create an ontology that encompasses medical knowledge but to make use of existing ontologies, exploit the knowledge encoded in them, and together with the data and metadata contained in the DICOM4 standard, to improve the techniques of segmentation and visualization defining new methods of semantic segmentation guided by ontologies. Once the images have been written down, intelligent searches on a medical image database with respect to some fabric, some pathology, among other capabilities.

PROMETEO PROGRAM - UNIVERSITY OF CUENCA