Identification of Tuberculosis with AI techniques

by | Nov 6, 2017 | Healthcare, Research, Technology

Tuberculosis is mainly found in poorer regions of Africa and Asia. Resources and trained medical personal are scare in theses regions, leading to more deaths than necessary. With upcoming artificial intelligence and big data techniques identification of tuberculosis in an early stage becomes possible. Fast and early detection may support the radiologist to more effectively fight the infectious tuberculosis disease and prevent it from spreading across the population.


At Genematics Scientific we are developing a trained model that supports the medical professional in the detection and identification of tuberculosis. As a starting point we took public tuberculosis datasets for its training and developed a classifier around the Google’s deep-learning framework: Tensorflow. The system was trained multiple days on the Microsoft Azure platform in combination with Nvidia Tesla GPUs (Tensorflow supports CUDA out-of-the-box). The first version of our trained model is currently implemented in the free to acess public version of our Genematics Cloud Platform which be accessed by registering at our sign-up page or sign-in at the login page of the Genematics Cloud Platform. We’d love to hear feedback on the model, and if you or your organization is interested in a collaboration we’d love to hear aswell.


Thumbnail image credit: Yale Rosen. Licensed via Creative Commons 2.0. Article image credit: own work. Licensed via Creative Commons 2.0.