Using AI to streamline precision oncology: Northeastern collaborates with JAX’s Maine Cancer Genomics Initiative

A system designed to swiftly parse and summarize vital information could help MCGI’s Genomic Tumor Board experts dedicate their full attention to their expertise in genomics and cancer care.

Here's how a team of AI experts from Northeastern University is working with The Jackson Laboratory’s Maine Cancer Genomics Initiative (MCGI) to achieve this, and help bring precision oncology to more patients in rural areas:

About MCGI

The Maine Cancer Genomics Initiative (MCGI) is a collaboration between Maine’s oncology providers and The Jackson Laboratory. MCGI’s goal is to improve access to advanced cancer care and precision medicine for Maine and New England. The work has proven particularly impactful in low-resource, rural areas. MCGI has impacted more than 3,500 patients since its founding in 2016 - providing access to advanced genomic testing technologies, precision medicine interpretation support or local clinical trial availability.

The Problem

Cancer care is becoming more complicated as physicians utilize increasingly detailed biomarker testing data from individual patients and their tumors to identify the best treatments, such as advanced immunotherapies. For many oncologists (especially in rural areas where per-capita doctor/patient coverage is lowest), the pace and volume of growing information limits their ability to keep pace with progress in precision oncology. These oncologists are often not specialized in treating one type of cancer, so they are considered generalists and must be able to treat many different diagnoses with little additional expert support.

To ensure cancer patients in the sprawling, largely rural state of Maine get access to cutting-edge care, MCGI facilitates virtual meetings in which a panel of external advisors, along with experts in clinical trials and genomic testing, are paired with oncologists and pathologists from MCGI’s network to discuss patient cases. Ahead of the meetings, which are known as Genomic Tumor Boards (GTBs), experts parse through complex, unstructured documents ranging from test results to handwritten notes and fragments of medical records. The meetings also generate extensive minutes which summarize the discussion for the treating oncologist. As a result, more than half of the experts’ time is devoted to case preparation and summarization, limiting the number of patients they can help.

Solution

In an effort to streamline meeting preparation and summaries, Northeastern’s team of experts from the Institute for Experiential AI (EAI) and the Roux Institute are collaborating with MCGI and leveraging AI tools to summarize and extract key information from patient health records. The processes perform optical character recognition to read handwritten notes and other hard-to-

discern records using specialized prompt algorithms for large language models (LLMs) and neural networks. The systems aim to perform two primary functions:

1. Automate document processing to reduce manual entry

2. Automatically structure meeting transcripts into queryable summaries, making it easier for experts to retrieve insights and streamlining decision-making

"Working with Northeastern's Institute for Experiential AI and the Roux Institute to automate the time-consuming process of cleaning up data for the treating oncologist and our team of experts ultimately will give our GTB team more time for the crucial work of identifying treatment options for patients," said Leah Graham, Ph.D., Program Director, MCGI.

‍

Results

These solutions could bring substantial efficiency improvements to the GTB workflow, saving time for genetic experts and physicians so that they are free to focus on what they do best – helping patients.

MCGI compared the AI outputs with work from MCGI’s human experts across more than 3,000 pages of unstructured intake documents for 148 patients. In this run, they determined the LLM tools successfully extracted complicated genomic testing data and processed clinical intake documents, and that the neural networks were 90 percent accurate in reproducing materials from intake documents.

"The team at EAI and the Roux Institute go beyond the standard project management methodologies one might expect from a typical development partner. Their expertise and collegiality make this project a true collaborative effort," said Jennifer Bourne, M.S., Operations Manager, MCGI.

Next Steps

Recognizing the impact of these AI-driven tools, MCGI, MaineHealth, and Northeastern University aim to build on this work and vet AI tools in additional real-world settings.

The work will facilitate processes that empower human subject matter experts to deliver advanced care to more patients. The next phase will focus on:

• AI assistants that automate data assembly and presentation
• AI capable of parsing GTB meeting records to provide transcripts and treatment recommendation summaries

"Cancer remains the second leading cause of death in the US, and precision oncology has the potential to change that. By using AI to simplify complex processes, we're helping clinicians across the country access the insights they need to provide more personalized, effective treatments — and ultimately, better care for more patients." - Maria Giovanna Trovato, Global Strategy & Business Development in Healthcare and Life Science, The Institute for Experiential AI

Find Your Own Solutions With AI

The work exemplifies the power and potential of AI to streamline workflows in healthcare and improve decision-making for patients. Experts at the Institute for Experiential AI and the Roux Institute harness a team of leading data scientists, AI researchers, practitioners, and Northeastern University faculty members to build impactful solutions for organizations like The Jackson Laboratory.

Schedule a strategy session to see how the Institute for Experiential AI can help your organization solve its most pressing problems with AI, or learn more about our work.

Learn more about The Jackson Laboratory here.

‍