Object Detection and Instance Segmentation in Chest X-rays for Tuberculosis Screening

Abstract

Tuberculosis (TB) is a highly contagious disease leading to the deaths of approximately 2 million people annually. TB primarily affects the lungs and is spread through the air when people cough, sneeze, or spit. Providing healthcare professionals with better information, at a faster pace, is essential for combating this disease, especially in Low and Middle Income Countries (LMICs) with resource-constrained health systems. In this paper we describe how using convolution neural networks (CNNs) with an object level annotated dataset of chest X-rays (CXRs) allows us to identify the location of pulmonary issues indicative of TB. We compare the performance of Faster R-nobreakdash-CNN, Mask R-nobreakdash-CNN, Cascade versions of each, and SOLOv2, demonstrating reasonable results with a small dataset. We present a method to reduce the false positive rate by comparing the location of a detected object with the known location of areas where the detected class is likely to occur in the lung. Our results show that object detection and instance segmentation of CXRs can be achieved with a dataset of high-quality, object level annotations, and could be used as part of an automated TB screening process. This work has the potential to improve the speed of TB diagnosis in LMICs, if properly integrated into the healthcare system and adapted to existing clinical workflows and local regulations.