Northwell Health

Pancreatic cancer is notoriously difficult to detect early, with more than 80% of patients diagnosed at an advanced stage. Traditional diagnostic workflows resulted in lengthy delays between imaging, biopsy, specialist consultation, and treatment initiation, severely limiting outcomes.

Northwell developed iNav, an in-house machine learning-based NLP model that analyzes more than 10,000 MRI and CT scan reports weekly across its 28-hospital network. The system flags high-risk individuals in real time, enabling care coordinators to accelerate specialist referrals and treatment connections.

A study of 71 patients published in The Oncologist found that iNav cut the time from imaging to biopsy from 12 days to 6 days. Oncologist appointment wait times dropped from 27 to 17 days, and time to treatment start shortened from 56 to 35 days. Benefits were observed consistently across racial and ethnic minority groups, suggesting potential for reducing health disparities.

• In-house AI development (via an internal innovation challenge) can produce clinically validated tools without external vendor dependency.
• NLP applied to radiology reports at scale can surface high-risk patients faster than traditional workflows, compressing multi-week diagnostic timelines.
• Equity impact should be a measured outcome: iNav showed consistent improvements across racial and ethnic minority groups, not just aggregate populations.