Key Cancer AI Glossary Terms

Federated Learning (FL)

An AI training approach that preserves the anonymity of sensitive clinical data. Instead of bringing sensitive patient data to a centralized location, federated learning brings the AI model to the data.

Learn more about how CAIA harnesses federated learning to accelerate cancer research.

Data Federation

A framework aimed at aggregating clinical insights across a distributed network of cancer centers without compromising patient data, privacy, or security. CAIA’s data federation framework is underpinned by an AI training approach called federated learning.

Read our blog post on the role of data federation in cancer research.

Central Orchestration Layer

A coordinating element in CAIA’s federated learning platform that facilitates collaboration between participating cancer centers without ever accessing sensitive patient data.

Edge Node

A secure compute environment (like a computer or a server) that serves as a gateway between the cancer center and the rest of the Alliance.

Learn more about how edge notes facilitate decentralized federated learning.

Gradients

Mathematical representations that AI models use to improve their predictions. A gradient tells an AI model how to adjust its calculations based on the underlying data.

Read our detailed guide into how gradient updates create smarter AI models.

Baseline AI model

The initial model code and weights sent by researchers to participating cancer centers’ edge during the FL process.

Global AI Model

The “smarter” model that contains aggregated updates as the end results of federated learning. 

NVIDIA FLARE

NVIDIA’s Federated Learning Application Runtime Environment (FLARE) provides the framework and code for secure communication between the edge nodes and the orchestration layer.

Multi-Cloud Federated Learning

An interoperable approach that enables cancer centers to operate on the cloud environments that work best for them.

Data Standardization

Data Standardization

A process by which a set of parties agree to a common format for the data collected about patients. A common data format can be interpreted, and reliably analyzed across different institutions.

Common Data Model (CDM)

A Common Data Model (CDM) provides a framework for how different types of patient data can be organized and structured.

Observational Medical Outcomes Partnership (OMOP)

CAIA has adopted OMOP as its CDM. OMOP is an open data standard maintained and supported by OHDSI, an international open science community. It is an implementation of data standardization that can translate local hospital data into a universal standardized system.

Electronic Health Records (EHR)

Records that capture both unstructured and structured data about how a patient experiences care. 

Structured Data

Clean, organized, and standardized data elements, such as entries in an Electronic Health Record about the dose of a drug or the results of a lab test. 

Unstructured DataData

Elements of a patient record that include free-text – like clinical notes, pathology reports or radiology reports.

Gen-1 dataset

A federated dataset comprising standardized, structured clinical data from over one million patients across 4 participating cancer centers.

CAIA’s Research and Impact

Clinical innovation projects

These projects use AI to answer critical questions related to patient care today by leveraging the federated Gen-1 dataset

AI innovation projects

These projects are building critical AI infrastructure such as massive foundation models and applications trained on the federated Gen-1 dataset.

Real-world data

Routine data that is gathered during care visits at cancer centers outside of structured clinical trials. EHR data from cancer centers is one source of real-world data. 

Clinical trials

A clinical trial is a research study involving a select group of people — either healthy individuals or patients dealing with a disease — who have volunteered to try a new intervention, an existing treatment, or a comparison of those under a defined protocol and close medical monitoring. 

NCI-designated cancer center

The National Cancer Institute (NCI) provides this designation to cancer centers that “meet rigorous standards for transdisciplinary, state-of-the-art research focused on developing new and better approaches to preventing, diagnosing, and treating cancer.” All cancer centers that are currently part of CAIA are NCI-designated Comprehensive Cancer Centers that demonstrate “an added depth and breadth of research, as well as substantial transdisciplinary research that bridges these scientific areas.”