NCI-DOE Collaboration, MOSSAIC project (Modeling Outcomes using Surveillance Data and Scalable AI for Cancer)

Overview

Development stage

DeployedSource: c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

2021-01 (month precision)Source: Jan-21

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Science

Operational functions

Document processing and extraction, Clinical and medical Derived by keyword rules; see methodology

Problem the AI is intended to solve

MOSSAIC applies deep learning natural language processing (NLP) and foundation models to population-based cancer data collected by NCI's Surveillance, Epidemiology, and End Results (SEER) program. DOE's Oak Ridge National Lab (ORNL) has data use agreements (DUAs) with multiple SEER registries to access and train models using SEER data. Two APIs are in production use in the data management system used by the SEER registries -- OncoID which predicts whether a pathology report is related to cancer and OncoIE which extracts key tumor characteristics from unstructured pathology report text.  Together these APIs are an important part of moving the US towards near real-time cancer incidence reporting.  In addition, a third API OncoMetsID, which predicts whether a pathology report is indicative of metastatic disease, is in a pilot phase to use in conjunction with other sources of information in the registries to identify recurrent disease.

Expected benefits

MOSSAIC enhances the infrastructure of the SEER cancer registries by providing tools that can increase the efficiency and accuracy of manual data abstraction by automatically extracting cancer surveillance data elements.  SEER registries receive millions of unstructured clinical text documents that must be manually reviewed, leading to a lag in reporting of US cancer incidence trends.  Automated tools such as those developed by MOSSAIC will help us achieve near real-time incidence trends and ultimately a more meaningful report card on the status of cancer in the US.

System outputs

Input: unstructured (free text) cancer pathology reports.                         Output: varies depending on the algorithm but generally a predicted class (eg, tumor site) and associated relative confidence score that can be used to tune accuracy

Technology

AI classification

Natural language processingSource: Natural Language Processing: AI that processes, interprets, and shares information in human language.

System name(s)

SEER*Data Management System (SEER*DMS)

Custom-developed code

No

Public source code

computational.cancer.gov ↗

Sourcing

How it was built

Contract and in-houseSource: c) Developed with both contracting and in-house resources

Vendor (as reported)

development -- Oak Ridge National Lab; maintenance -- Information Management Services (IMS)

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

No

Privacy Impact Assessment

Not reported

Authorization to Operate

Yes

Demographic features

k) None of the above

Training and evaluation data

Data is owned by the NCI SEER registries, which are funded by the NCI

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported