Avaliação de radiografias abdominais em recém nascidos prematuros com enterocolite necrosante: uma ferramenta baseada em processamento de imagens digitais
Carregando...
Data
2017-07-31
Autores
Sampaio, Ana Julia Martins [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A enterocolite necrosante (NEC – do inglês necrotizing enterocolitis) é caracterizada pela necrose isquêmica da mucosa intestinal de recém-nascidos prematuros. Uma vez que existe a suspeita de NEC, é instituída uma rotina de radiografias abdominais de acordo com a gravidade da doença. Os principais achados radiológicos de pacientes com NEC são: distensão abdominal generalizada, pneumatose intestinal, pneumoperitônio. Entretanto, a interpretação dessas radiografias é um processo difícil devido à falta de especificidade da maioria dos achados radiológicos. O objetivo desse estudo foi desenvolver uma ferramenta computacional que auxilie o corpo clínico na análise de radiografias abdominais para a diferenciação de alças normais e alças inflamadas em recém-nascidos prematuros. Para o desenvolvimento desta pesquisa foi utilizado um banco de dados composto por 45 radiografias abdominais e algoritmos computacionais desenvolvidos em ambiente MatLab. As espessuras das alças intestinais foram quantificadas através da ferramenta computacional Largura a Meia Altura (FWHM – do inglês Full Width at Half Maximum), e classificadas como alças edemaciadas ou alças normais. Para a análise de textura e extração de características, a fim de diferenciar regiões de pneumatose, aplicamos a técnica de Transformada Wavelet. Com a utilização do algoritmo, as alças intestinais normais apresentaram mediana igual a 10,30 pixels, enquanto as alças edemaciadas, foram estatisticamente maiores (Mann Whitney, p < 0,05), com mediana de 15,13 pixels. Para as alças com pneumatose, as quantificações foram realizadas em três direções (horizontal, vertical e diagonal) para oito níveis de decomposição. Os níveis 7 e 8 da decomposição horizontal apresentaram diferenças significantes (Mann Whitney, p < 0,05) quando comparadas as alças normais com as regiões de pneumatose. Para o nível 7 foram obtidos o valor 0,034 para a mediana das alças normais e 0,088 para alças com pneumatose; já para o nível 8, as medianas obtidas foram 0,19 e 0,34, respectivamente. Podemos concluir, portanto, que a combinação de técnicas distintas de processamento de imagens aplicadas em imagens médicas apresentou diferenças consideráveis pouco visíveis em avaliações subjetivas. Os resultados obtidos nesta pesquisa contribuem como uma ferramenta a mais, de modo a auxiliar neonatologistas e radiologistas a tomada de decisões com maior seguridade na detecção de alças acometidas pela NEC, dando maiores subsídios ao diagnóstico e acompanhamento clínico.
Necrotizing enterocolitis (NEC) is characterized by ischemic necrosis of the intestinal mucosa of preterm newborns. Since there is a suspicion of NEC, a routine of abdominal radiographs is instituted according to the severity of the disease. The main radiological signs of patients with NEC are generalized abdominal distension, intestinal pneumatosis, pneumoperitoneum. However, the interpretation of these radiographs is a difficult process due to the lack of specificity of most radiological findings. The aim of this study was to develop a computational tool to assist the clinical staff in the differentiation of normal loops of inflamed loops in radiographs of premature newborns. For the development of this research we started with a database consisting of 45 abdominal radiographs and computational algorithms developed in MatLab environment. The thicknesses of the intestinal loops were quantified using the Full Width at Half Maximum (FWHM) computational tool, and classified as distended loops or normal loops. For the analysis of texture and extraction of characteristics in order to differentiate regions of pneumatosis, we apply the Wavelet Transform technique. With the use of the algorithm, the normal intestinal loops presented a median of 10,30 pixels, whereas distended loops were statistically larger (Mann Whitney, p <0.05), with a median of 15.13 pixels. For the pneumatosis loops, the quantifications were performed in three directions (horizontal, vertical and diagonal) for eight levels of decomposition. Levels 7 and 8 of horizontal decomposition showed significant differences (Mann Whitney, p < 0,05) when compared to normal loops with pneumatosis regions. For level 7, a value of 0.034 was obtained for the median of the normal loops and 0.088 for the loops with pneumatosis; already for level 8, the medians obtained were 0.19 and 0.34, respectively. We can conclude, therefore, that the combination of different image processing techniques applied in medical images presented considerable differences that are not very visible in subjective evaluations. The results obtained in this research contribute as an additional tool, in order to help neonatologists and radiologists to make decisions with greater safety in the detection of NEC affected handles, giving greater subsidies to diagnosis and clinical follow up.
Necrotizing enterocolitis (NEC) is characterized by ischemic necrosis of the intestinal mucosa of preterm newborns. Since there is a suspicion of NEC, a routine of abdominal radiographs is instituted according to the severity of the disease. The main radiological signs of patients with NEC are generalized abdominal distension, intestinal pneumatosis, pneumoperitoneum. However, the interpretation of these radiographs is a difficult process due to the lack of specificity of most radiological findings. The aim of this study was to develop a computational tool to assist the clinical staff in the differentiation of normal loops of inflamed loops in radiographs of premature newborns. For the development of this research we started with a database consisting of 45 abdominal radiographs and computational algorithms developed in MatLab environment. The thicknesses of the intestinal loops were quantified using the Full Width at Half Maximum (FWHM) computational tool, and classified as distended loops or normal loops. For the analysis of texture and extraction of characteristics in order to differentiate regions of pneumatosis, we apply the Wavelet Transform technique. With the use of the algorithm, the normal intestinal loops presented a median of 10,30 pixels, whereas distended loops were statistically larger (Mann Whitney, p <0.05), with a median of 15.13 pixels. For the pneumatosis loops, the quantifications were performed in three directions (horizontal, vertical and diagonal) for eight levels of decomposition. Levels 7 and 8 of horizontal decomposition showed significant differences (Mann Whitney, p < 0,05) when compared to normal loops with pneumatosis regions. For level 7, a value of 0.034 was obtained for the median of the normal loops and 0.088 for the loops with pneumatosis; already for level 8, the medians obtained were 0.19 and 0.34, respectively. We can conclude, therefore, that the combination of different image processing techniques applied in medical images presented considerable differences that are not very visible in subjective evaluations. The results obtained in this research contribute as an additional tool, in order to help neonatologists and radiologists to make decisions with greater safety in the detection of NEC affected handles, giving greater subsidies to diagnosis and clinical follow up.
Descrição
Palavras-chave
Enterocolite necrosante, Largura à meia altura, Processamento de sinais, Raios X, Transformada Wavelet, Necrotizing enterocolitis, Full width at half maximum, Signal processing, X ray, Wavelet Transform