Mi SciELO
Servicios Personalizados
Articulo
Indicadores
- Citado por SciELO
Links relacionados
- Similares en SciELO
Compartir
Ingeniería Electrónica, Automática y Comunicaciones
versión On-line ISSN 1815-5928
Resumen
PLA MARTINEZ, Gilbert y IRIZAR MESA, Mirtha. Automatic Classifier of blood sample images based on deep neural networks. EAC [online]. 2019, vol.40, n.1, pp. 18-30. ISSN 1815-5928.
Massive neonatal screening is a test performed on all newborns to detect and prevent congenital, hereditary and metabolic diseases that affect the child's normal growth and development. This test consists of making a small puncture in the heel of the baby to take a few drops of blood and place them on a filter paper to be analyzed in the laboratory. One of the stages of this analysis is the quality evaluation of the samples based on elements such as the coloring of the samples, the drying, the presence of impurities and the presence of clots. Because this stage requires high experience, agility and visual memory of the specialists, sometimes samples are taken to the laboratory and they do not provide the best results due to the low quality they present. This paper presents a study to develop an automatic classifier based on deep artificial neural networks that evaluates the quality of the blood samples analyzed. This classifier is based on pattern recognition associated to the color moments and the HSV components, extracted from the images of blood samples. For this, a deep neural network formed by two deep autoencoders and a softmax classifier is trained with examples of images of blood samples, obtaining satisfactory results in the validation of the method with the correct classification of the samples submitted for analysis
Palabras clave : neonatal screening; color moments; HSV components; deep autoencoders.