Automatic Tuberculosis Detection using Binary Pattern of Phase Congruency

Computational solutions to assist the automatic analysis of radiographic images are already a reality. Commercial solutions for the diagnosis of tuberculosis have shown exceptional results similar to those of radiologists. However, these solutions are still far from regions of greater vulnerability, given their high costs. Researchers have shown a growing interest in developing increasingly efficient and effective solutions that require the least resources. Thus, in this work, we explore the use of chest X-ray and Phase Congruence Binary Pattern features to build a minimalist, low computational cost, and high-efficiency Feed-Forward Neural Network model to aid the diagnosis of tuberculosis. Our results showed high performance compared to related research, placing our solution as a viable alternative.