Reconhecimento da gravidade da radiografia de tórax de indivíduos com fibrose pulmonar idiopática

• Autor(a) principal: Costa, Luana dos Santos
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/7716
• topic: Fuzzy Connectedness; Idiopathic Pulmonary Fibrosis; Bland-Altman Agreement; Conectividade Fuzzy; Fibrose Pulmonar Idiopática; Concordância Bland-Altman; Processamento digital de imagens; Segmentação Fuzzy; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

Pulmonary disease affects a significant portion of the world population. The motivation for the project was to achieve cost reduction in pulmonary disease diagnosis by applying recent advances in automatic pattern recognition based on image segmentation methodology. The CImages system for image segmentation, based on Fuzzy Connectedness, may be used as a tool in the follow up and gradation of Idiopathic Pulmonary Fibrosis (IPF), a chronic interstitial scaring disease, by analyzing X-Ray images and categorizing the damage to different sections of the lung as Light, Moderate or Serious, thereby avoiding the subjective evaluation of a human Radiologist. Fuzzy Connectedness is an emerging approach in image processing suited to the imprecise nature of the task. Semiautomatic segmentation demarcates approximate boundaries of interest areas defining an object by using the strength of pixel interconnectivity and mapping connection relationships. An index for the quantification of the degree of impairment of the tissue image being analyzed is created. The Bland-Altman methodology was used in the validation of the fuzzy connectivity approach generating a graphic representation of confidence intervals and limits of agreement permitting the CImages evaluations to be compared to those of expert Radiologists. The methodology was applied to 15 Chest X-Ray from posterior-to-anterior (PA) of different lung regions: top right, top left, medium rigt, medium left, bottom right and bottom left. The agreement of the CImages results and that of the expert was acceptable for the evaluation of the top right region of four patients aged 60 to 69 and another four aged 80 to 89. In analogous fashion there was also acceptable agreement in the evaluation of medium-left regions of four other patients aged 70 to 79. For the top left and medium left regions there was good agreement for three patients aged 50 to 59. The use of new pharmacological treatments of IPF require efficacy monitoring by measuring Forced Vital Capacity and radiological analysis, low cost procedures and of lower radiation exposure compared to Tomography. The CImages software reduces subjectivity in X-Ray image interpretation by an expert radiologist, contributing to the validation of new treatments to halt IPF progression at lower cost. It is understood that the option of analyzing X-Ray images using two tonalities, namely black and white, motivates us to upgrade the technique to include different shades of gray, even though this option could be hampered by hardware limitations such as memory space and processing capacity. The gravity of IPF disease underlines the importance of the effort