Predicting Bovine Tuberculosis Status from Milk Mid-Infrared Spectral Data Using Deep Learning

  • Scott Denholm (Speaker)

Activity: Talk, evidence or presentation typesOral presentation


Mid-infrared (MIR) analysis of milk samples is used routinely to predict the fat and protein content of dairy cow milk as part of the production process, but it's also an effective predictor of other economically important (and often hard-to-record) phenotypes. Learn how we can use individual cow MIR profiles, matched to corresponding reference phenotypes, to train deep convolutional neural networks and develop a phenotype prediction pipeline. By running these models on NVIDIA DGX Station with RAPIDS, the team was able to speed up the time it took to develop models at least tenfold. We'll highlight the effectiveness of deep learning applied to agriculture with a case study predicting bovine tuberculosis (bTB), a chronic zoonotic disease of cattle that's of great economic, welfare, and societal importance. We'll show this method of prediction provides a rapid, low-cost, and (more importantly) non-invasive method for the routine collection of phenotypes that can subsequently be used for herd management, health and fertility monitoring, and genetic improvement.
Period14 Apr 2021
Held atNVIDIA, United States
Degree of RecognitionInternational