Research on Biomedical Engineering
https://rbejournal.org/article/5889fb6e5d01231a018b4619
Research on Biomedical Engineering
Original Article

Segmentación de biopsias de médula ósea mediante filtros morfológicos y rotulación de regiones homogéneas

Segmentation of bone marrow biopsies by morphological filters and labeling of flat zones

Pastore, Juan Ignacio; Moler, Emilce; Meschino, Gustavo

Downloads: 0
Views: 388

Resumen

Este trabajo presenta un método semi-automático de detección de tejidos en biopsias de médula ósea utilizando técnicas de Procesamiento Digital de Imágenes. Las técnicas utilizadas combinan filtrados morfológicos y detección de regiones homogéneas, con el fin de realizar un cálculo preciso de la celularidad medular. Los informes anátomo-patológicos de estos cortes histológicos entregan resultados porcentuales de la celularidad medular, indicando la presencia de trabéculas, células adiposas y hematopoyéticas. Dichos porcentajes per miten evaluar la presencia y/o el grado de algún desorden metabólico, estableciendo comparaciones entre los valores normales y los patológicos. Generalmente, estas mediciones se realizan por simple inspección visual. El método propuesto permite calcular el porcentaje de trabéculas, células adiposas y hematopoyéticas. La identificación de trabéculas se basa en la aplicación de Filtros Morfológicos Alternativos Secuenciales por Reconstrucción y rotulación de regiones homogéneas. Todo el proceso es implementado en lenguaje MatLab® 6.5 con una interfase visual que permite operar el sistema sin ser experto en Procesamiento de Imágenes. El método fue probado con 26 imágenes de biopsias de médula ósea que presentan distribuciones variadas de trabéculas. Se presenta un análisis comparativo entre los resultados obtenidos con el método propuesto y los calculados por dos patólogos es pecialistas donde la diferencia entre ambas estimaciones es menor al 12%, mientras que las diferencias entre ellos llegan a ser hasta de 20%. Tanto los resultados visuales como los numéricos indican que el método es altamente eficiente para esta aplicación.

Palabras clave

Filtros morfológicos secuenciales, Médula ósea, Segmentación

Abstract

This work presents a method for semi-automatic detection of dif ferent types of tissues in bone marrow biopsies using techniques of Digital Image Processing. The techniques are based on combina tions of morphological filtering and flat zones detection, to perform an exact calculation of the medullar celullarity. The anatomic and pathologic rof trabeculae, hematopoietic and fatty cells. The identification of trabecular structures is achieved by the application of Alternating Sequential morphological Filters by reconstruction (ASFs) and labeling of flat zones. The whole process is implemented on MatLab® 6.5 with a guide user interface that allows operation without Image Processing acknowledgement. Methods were tested with 26 images of bone marrow biopsies presenting varied distributions of cellular ity and trabeculae topography, presenting varied distributions of cellularity and trabeculae topography. The results were compared to others obtained by visual inspection by two expert pathologists, and differences of less than 12 % were observed, while differences up to 20% were observed between results given by experts. The visual and numerical results indicate that this method is highly efficient for this application.eports of the histological cuts give results expressed in percentages, indicating the presence of trabeculae, hematopoietic and fatty cells. This allows evaluating the existence or grade of some metabolic disorder, comparing normal values with pathologic ones. Generally, quantifications are carried out by simple visual inspection. The proposed method allows calculating the percentage of trabeculae, hematopoietic and fatty cells. The identification of trabecular structures is achieved by the application of Alternating Sequential morphological Filters by reconstruction (ASFs) and labeling of flat zones. The whole process is implemented on MatLab® 6.5 with a guide user interface that allows operation without Image Processing acknowledgement. Methods were tested with 26 images of bone marrow biopsies presenting varied distributions of cellular ity and trabeculae topography, presenting varied distributions of cellularity and trabeculae topography. The results were compared to others obtained by visual inspection by two expert pathologists, and differences of less than 12 % were observed, while differences up to 20% were observed between results given by experts. The visual and numerical results indicate that this method is highly efficient for this application.

Keywords

Alternating sequential morphological filters, Bone marrow, Segmentation.
5889fb6e5d01231a018b4619 rbeb Articles
Links & Downloads

Res. Biomed. Eng.

Share this page
Page Sections