The main projects I have been working on are listed as follows. See the papers for a full list of coauthors. For a whole list of my publications, please visit my Google Scholar site.
Image-Derived Arterial Radiotracer Concentration from dynamic PET
Using a low-absorption radiotracer, 18F-choline, we designed and implemented a method to automatically extract the venous and arterial Image-Derived Input Functions (IDIF), that is, the radiotracer concentration with respect to time. The venous IDIF is captured by segmenting the biggest venous pools, whereas the arterial IDIF is derived by contrasting the amount of radioactivity in the intracranial region and the amount of radioactive material entering it (aIDIF) and leaving it (vIDIF).
[Extended abstract, AIME 2019] [Slides, AIME 2019]
Soft Color Morphology
The Soft Color Morphology is an extension of fuzzy mathematical morphology to multivariate images. We process CIELab-encoded natural images with operators that provide good visual results. The soft color erosion and dilation maintain, respectively, the notions of shrinking and enlarging objects.
Kinetic Model of 18F-Choline Uptake in Gliomas
In this project, we study a model of how high-grade gliomas absorb choline through PET imaging. Blood samples are obtained to measure the blood concentration of 18F-choline and a dynamic PET reflects the tissue absorption. Different uptake parameters may indicate a different degree of malignancy or a different type of tumor.
This project has been developed at the Hospital Universitari Son Espases.
Curvilinear Object Detection
Locating and segmenting tubular-shaped regions is a problem common to several image processing tasks: road detection from aerial images, vessel segmentation in diverse medical imaging, bronchi segmentation, crack detection in roads… To unify the literature, we reviewed the state of the art.
Also, some morphological detectors were developed to segment curvilinear objects.
Real-Time Retinal Vessel Segmentation
Vessels appear as thin, elongated objects in eye-fundus images of the back of the eye. The automatic detection of the vascular network may provide useful information to the practitioner to correctly diagnose several medical conditions. We employ fuzzy mathematical morphology techniques to segment vessels in eye-fundus photographs at a rate of 30-60 miliseconds per image.