Radiological characterization of lung remodeling in chronic obstructive pulmonary disease

• Autores: Rina Dewi Rudyanto
• Directores de la Tesis: Arrate Muñoz Barrutia (dir. tes.), Carlos Ortiz de Solórzano Aurusa (codir. tes.)
• Lectura: En la Universidad de Navarra ( España ) en 2014
• Idioma: español
• Tribunal Calificador de la Tesis: Juan Pablo de Torres Tajes (presid.), Igone Velez Isasmendi (secret.), Juan José Vaquero López (voc.), Bram Van Ginneken (voc.), Petia Radeva (voc.)
• Materias:
  • Matemáticas
    • Ciencia de los ordenadores
      • Inteligencia artificial
  • Ciencias médicas
    • Ciencias clínicas
      • Oncología
• Texto completo no disponible (Saber más ...)
• Resumen

  • Chronic Obstructive Pulmonary Disease (COPD) is the 4th global leading cause of death, and is defined as the finding of nonreversible pulmonary function impairment. While the definitive diagnosis of COPD is often functional spirometry test, the presence of COPD in cancer screening population has become a source of interest. The aim of the thesis is to characterize COPD lung remodeling from anatomical findings on the CT images. Methods are described for the automatic segmentation and quantitative analysis of anatomical structures from thoracic CT scans. First, segmentation of vessels is obtained, as vessels are required for segmenting other structures.

    To do this, vessel segmentation methods are compared in a standardized framework. After the segmenting vessel trees, the lung lobes are segmented, which then enable local characterization of lung remodeling. To label the lung lobes, the lung fissures are first extracted. When the fissures are incomplete or absent, the vessel trees are used to guide lung lobe segmentation.

    Blood vessel dimensions are then quantified to study vessel tree alteration and loss of vessels that accompany COPD. Airways are another type of anatomical structures directly affected by COPD. Using knowledge of tubularity and intensity, as well as location information of fissures and vessels, the airways are segmented using a fast marching algorithm and refined using pattern recognition techniques. Airways are then quantitatively measured by airway count and airway wall dimensions. We also present a method devoted to the matching and volumetric follow-up of lung tumors in longitudinal images. Finally, we perform an analysis of the lung anatomical structures across longitudinal study to characterize lung remodeling as a function of progress of COPD.