Search

We demonstrate that Domain Invariant Feature Learning (DIFL) can improve the out-of-domain generalizability of a deep learning Tuberculosis (TB) screening algorithm. It is well known that state of the art deep learning algorithms often have difficulty generalizing to unseen data distributions due to “domain shift.” In the context of medical imaging, this could lead to unintended biases such as the inability to generalize from one patient population to another. We analyze the performance of a ResNet-50 classifier for the purposes of TB screening using the four most popular public datasets with geographically diverse sources of imagery. We show that without domain adaptation, ResNet-50 has difficulty in generalizing between imaging distributions from a number of public TB screening datasets with imagery from geographically distributed regions. However, with the incorporation of DIFL, the out-of-domain performance is greatly enhanced. Analysis criteria includes a comparison of accuracy, sensitivity, specificity and AUC over both the baseline, as well as the DIFL enhanced algorithms. We conclude that DIFL improves generalizability of TB screening while maintaining acceptable accuracy over the source domain imagery when applied across a variety of public datasets.

Nipah virus (NiV) is a paramyxovirus (genus Henipavirus) that emerged in the late 1990s in Malaysia and has since been identified as the cause of sporadic outbreaks of severe febrile disease in Bangladesh and India. NiV infection is frequently associated with severe respiratory or neurological disease in infected humans with transmission to humans through inhalation, contact or consumption of NiV contaminated foods. In the work presented here, the development of disease was investigated in the African Green Monkey (AGM) model following intratracheal (IT) and, for the first time, small-particle aerosol administration of NiV. This study utilized computed tomography (CT) and magnetic resonance imaging (MRI) to temporally assess disease progression. The host immune response and changes in immune cell populations over the course of disease were also evaluated. This study found that IT and small-particle administration of NiV caused similar disease progression, but that IT inoculation induced significant congestion in the lungs while disease following small-particle aerosol inoculation was largely confined to the lower respiratory tract. Quantitative assessment of changes in lung volume found up to a 45% loss in IT inoculated animals. None of the subjects in this study developed overt neurological disease, a finding that was supported by MRI analysis. The development of neutralizing antibodies was not apparent over the 8-10 day course of disease, but changes in cytokine response in all animals and activated CD8+ T cell numbers suggest the onset of cell-mediated immunity. These studies demonstrate that IT and small-particle aerosol infection with NiV in the AGM model leads to a severe respiratory disease devoid of neurological indications. This work also suggests that extending the disease course or minimizing the impact of the respiratory component is critical to developing a model that has a neurological component and more accurately reflects the human condition.