Topological Concepts applied to Digital Image Processing

Abstract

This article describes an automatic method applicable to the segmentation of mediastinum Computerized Axial Tomography (CAT) images with tumors, by means of Alternating Sequential Filters (ASFs) of Mathematical Morphology, and connected components extraction based on continuous topology concepts. Digital images can be related to topological space structures, and then general topology principles can be straightforwardly implemented. This method allows not only to accurately determine the area and external boundary of the segmented structures but also to obtain their precise location. Throughout these last years, technological development has significantly improved diagnostic imaging, enabling renal tumor and incidental hepatic tumor detection -usually small in size- in younger people and with an eventually lower malignant potential. This has led to a remarkable advance in interventionist techniques such as cryosurgery and radiofrequency ablation, preventing, in some cases, major surgeries, decreasing morbid-mortality rate, hospital stay and total treatment costs. Notwithstanding this, both cryosurgery and radiofrequency ablation, through extremely low and high temperatures, respectively, kill tumor as well as healthy cells, rendering crucial the identification of tumors with an extraordinary spatial accuracy.

Keywords

• segmentation
• topological spaces
• connected component
• cat