Máster en Visión por Computador (Computer Vision)

Permanent URI for this collectionhttps://hdl.handle.net/10347/37779

Browse

Recent Submissions

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item type: Item ,
    Impact of the healthy control database on brain pet imaging quantification
    (2025-02-03) Fernández Zorrilla, Elena; Universidade de Santiago de Compostela. Escola Técnica Superior de Enxeñaría; Aguiar Fernández, Pablo; Brea Sánchez, Víctor Manuel
    Positron Emission Tomography (PET) is crucial in neurological research and clinical practice, providing information about brain metabolism, especially through the use of [18F]FDG PET, which aids in the diagnosis of conditions such as epilepsy. While quantitative PET analysis can improve diagnostic precision, its results may be influenced by the selection of the healthy control (HC) dataset. This study uses SimPET-generated synthetic data to explore how healthy control database acquired in six different scanners impact quantitative PET analysis, aiming to improve the robustness of clinical quantification methods. To assess the impact of scanner variability, both VOI-based and voxel-based analyses were performed, comparing a synthetic database of epilepsy patients with the different synthetic HC databases. The results show that scanner variability significantly affected the detection of both focal and extended epileptic lesions. These findings highlight the importance of the data harmonization to ensure reliable and accurate PET quantification. In conclusion, while the choice of a locally acquired healthy control database is important, its impact on PET quantification could be limited when comparing scanners of the same generation.
  • Thumbnail Image
    Item type: Item ,
    Addressing Multiple Object Tracking with Segmentation Masks
    (2024) Bendaña Gómez, Manuel; Universidade de Santiago de Compostela. Escola Técnica Superior de Enxeñaría; Mucientes Molina, Manuel; Brea Sánchez, Víctor Manuel
    Multiple Object Tracking (MOT) aims to locate all the objects from a video, assigning them the same identities across all frames. Traditionally, this problem was addressed following the Tracking by Detection (TbD) paradigm, using detections represented by bounding boxes. However, bounding boxes can contain information from several objects, something that does not happen with segmentation masks. This work takes the ByteTrack MOT system as a starting point. Our proposal, ByteTrackMask, integrates a class-agnostic segmentation method and a segmentation-based tracker in ByteTrack in order to rescue tracks that would have been lost. Results over validation sets of MOT challenge datasets provide improvements in MOT metrics of interest like MOTA, IDF1 and false negatives.