Aurotech supports four different Data
Analytics and Data Warehousing initiatives at the FDA: a Unified Data
Warehouse, a Data Lake as service in the cloud (AWS), SAP HANA in the Cloud
(AWS), and use of natural language processing (NLP) in medical trials over that
data lake.
For the HHS Accelerate program we also
used AI/ML for contracting writing.
Aurotech is part of the IAAI (Intelligence Automation and Artificial
Intelligence) government-approved list. To learn more, click HERE
With response to COVID-19 as a top
priority, the FDA needed a solution to rapidly find valuable correlations and
uncover insights across scientific publications on the types of treatments
being prognosticated, administered, and assessed. We leveraged World Health
Organization COVID-19 Open Research Dataset (CORD) repository of almost 200,000
scientific publications relevant to the pandemic. The CORD data sets are
integrated with FDA’s SAP HANA backend that analyzes the text in memory using
Natural Language Processing (NLP) and correlates between a custom biological
dictionary and the NLM Medical Subject Headings (MeSH) dictionary. The result set is refined
by other metadata, such as the location of the researcher, their affiliated
organizations, and the date of publication. Aurotech created knowledge graphs
that capture facts and relationships related to people, treatments, chemical
compounds, drugs, and biologics.
FDA’s Office of Prescription Drug
Promotion currently receives 125,000 unique promotional materials annually with
receipts projected to increase 5% each year while reviewer staffing levels are
projected to remain unchanged. Aurotech developed a solution to augment the
reviewers with Docxonomy, an
AI-based platform, to help automate the analysis of promotional materials and
compare documents to ensure they remained unaltered to be consistent with FDA
polices. We converted source materials contained in PDF documents to structured
text and employed NLP toolkits to define risk categories such as boxed warning
and contraindication information. We used AI modeling services to build and
train ML algorithms that compare documents to ensure continuous process
improvement.
created with
Website Builder Software .