Classifying medical images is a manually intensive process, requiring expertise from pathologists, radiologists and other trained experts. Deep learning can help automate and accelerate image analysis as well as improve insights by enabling researchers to contextualize images with other data sources such as genetic, electronic health record data and more. However, most organizations attempting to build deep learning pipelines face challenges such as scaling legacy infrastructure, experimenting with rapidly proliferating DL frameworks, developing models efficiently within the team and deploying models into production.

In this webinar, Human Longevity Inc (HLI), a leader in medical imaging and genomics, will share how they overcame these challenges to deliver novel insights with deep learning at scale. They’ll walkthrough how they use Databricks and open-source technologies like Apache Spark, Tensorflow, and MLFlow to build a comprehensive imaging database of 14,000+ de-identified individuals and power an agile environment for model development, training and deployment.

Join this session to learn:

• HLI’s implementation to prepare and de-identify millions of images and lessons learned
• How to build reliable deep learning pipelines using Apache Spark, Tensorflow, and MLFlow on Databricks and AWS
• How HLI uses deep learning to extract biomarkers for chronic age-related diseases for integrated imaging and genetic risk prediction and phenotype-genotype discovery
• Live demo of a deep learning model for metastasis identification on the Databricks Unified Analytics Platform

• Christine Swisher, PhD, Director of Machine Learning, Human Longevity Inc
• Michael Wibbeke, Data Engineer, Human Longevity Inc
• Frank Austin Nothaft, PhD, Technical Director of Healthcare and Life Sciences, Databricks