V Congreso Latinoamericano de Ingeniería Biomédica (CLAIB2011)

Detección de Bacilos de Tuberculosis en Muestras de Esputo por medio de Técnicas de Procesamiento de Imágenes
Aguilar Guerrero Norman Gustavo, Castañeda Aphan Benjamín, Lavarello Roberto

##manager.scheduler.building##: Palacio de Convenciones de La Habana
##manager.scheduler.room##: Sala 8
Fecha: 2011-05-20 12:00  – 12:15
Última modificación: 2011-04-15 06:14


Tuberculosis is one of the most deadly diseases according to the World Health Organization. In 2008, 1.1-1.7 million people died and 8.9-9.9 million new cases were regis-tered. Currently, the most useful tool of diagnosis is the direct examination of sputum smears of possibly affected people. Since early diagnosis is the main strategy to control this dis-ease, the development of faster methods of diagnosis is re-quired. In this paper, an algorithm to automatically screen sputum samples is described. First, a large database of 8,200 microscopic images of sputum smears was created. Then, specialized staff labeled all objects in the images. Therefore, the algorithm was developed. The segmentation stage was based on color and shape. The following shape descriptors were obtained: Fourier descriptors, Hu moments and Zernike moments. The classification system was based on Principal Component Analysis and Support Vector Machines. The algo-rithm reached 41.24% sensitivity. The algorithm can be used to rapidly detect risky cases.