Introduction/Background: Digital pathology programs generate massive volumes of unstructured imaging data - at NewYork-Presbyterian (NYP), the pathology team scans tissue slides continuously, producing more than 2PB of new data annually, with individual slides reaching 50GB or more. Medical images must be retained on-premises for decades to meet regulatory requirements, yet supporting cloud-based AI pipelines demands fast, selective data movement. Traditional copy-and-sync tools copy everything indiscriminately, making cloud AI workflows prohibitively expensive and creating unnecessary PHI exposure.

Methods/Design: NYP implemented Komprise Intelligent AI Ingest to automate the discovery, classification, tagging, and selective ingestion of pathology slide data into AWS S3 for analysis by PathAI. The workflow curates only the newest, most relevant files - approximately 28GB every five minutes - transfers them to the cloud at high speed, enables pathologist review and verification, then deletes cloud copies after 30 days per policy. Original data remains on-premises for long-term compliance. A Komprise API integration allows PathAI to reingest older slides on demand.

Results: The approach reduced AWS cloud storage from 1PB to a rolling 33TB - a 96% reduction in cloud storage costs. Data ingestion speed improved 10x over prior transfer methods. Security exposure was dramatically reduced alongside costs. Researchers gained self-service access to clinical files via policy-based metadata tagging through the Komprise Global Metadatabase, eliminating redundant AI processing.

Conclusion/Discussion: NYP's initiative proves that clinical AI at scale demands purpose-built data management. By moving only the right data at the right time, health systems can self-fund AI programs through infrastructure savings while protecting PHI, accelerating pathologist workflows, and improving diagnostic precision in oncology and immuno-oncology.

Learning Objectives:

1. Evaluate key data infrastructure considerations for operationalizing digital pathology and AI in the clinical environment.
2. Recognize approaches to managing large-scale pathology image datasets without disrupting access for diagnostic or research.
3. Explain how tiered data management can support cost control, governance, and AI readiness in pathology programs.