Resumen

VALERO GOMEZ, Juan Carlos; ZUNIGA INCALLA, Alex Peter  y  CLARES PERCA, Juan Carlos. Tuberculosis detection with Deep Learning algorithms in chest X-ray images. Vive Rev. Salud [online]. 2021, vol.4, n.12, pp.190-199. ISSN 2664-3243.  https://doi.org/10.33996/revistavive.v4i12.119.

It is estimated that around 1.4 million people infected with tuberculosis died in 2019, most of them in developing countries. If tuberculosis had been diagnosed in time, the death of infected people would have been prevented. One of the most relevant tuberculosis detection methods is the analysis of chest radiographs; However, having highly trained professionals for the diagnosis of tuberculosis in all health centers is impossible in emerging countries, this is one of the main reasons why this method is not widely used. In recent decades, neural networks have played a very relevant role in solving problems in society and especially in the health sector. Three recognized Deep Learning algorithms have been used in the development of computational vision that are VGG19, MobileNet and InceptionV3, it has been possible to obtain very auspicious results for the detection of tuberculosis. MobileNet has been a special case, which has stood out among the others, giving important results in the different evaluation metrics used. In addition, MobileNet has a less complex architecture and the weights obtained after training are very less compared to the other two algorithms. It is concluded that MobileNet is the most efficient Deep Learning algorithm compared to VGG19 and InceptionV3, it has better precision for the detection of tuberculosis and the computational cost and processing time is significantly lower.

Palabras clave : neural networks; Deep Learning; tuberculosis; x-ray.

        · resumen en Español | Portugués     · texto en Español     · Español ( pdf )