By Nathan Opitz, Medicom VP, Data Services

With the ability to pull up a scan in one click at the point of care, most academic medical centers (AMCs) have solved the clinical side of imaging access. However, that same data often remains difficult for researchers to access and operationalize. Despite being centers of innovation, many institutions still struggle with fragmented architectures where imaging is siloed across disconnected PACS. And, because the EHR was designed for clinical operations rather than discovery, using data for longitudinal studies or AI training is onerous for investigators and data scientists. 

Medicom developed Research Assistant, an on-premise PHI-removal toolkit that preserves longitudinal accuracy while maintaining institutional control, to address the EHR bottleneck through enterprise-grade automation. Let’s examine how Research Assistant tackles three hurdles of imaging interoperability in an AMC world: maintaining clinical history, securing narrative text, and protecting image-level data.

The Manual Bottleneck: A Barrier to Institutional Innovation

Many research departments rely on homegrown scripts or manual redaction to extract and de-identify imaging data. As the demand for real-world evidence grows and participation in large-scale data consortiums becomes a strategic priority, these high-touch methods cannot keep pace with the speed and volume of modern data requirements. When a research project requires thousands of longitudinal studies to validate a new biomarker or train a machine-learning model, manual de-identification doesn't just slow down the process. It makes it cost-prohibitive and technically unscalable.

Then there’s the potential for human error. A single oversight in a DICOM tag or a missed patient name burned into an ultrasound frame can result in a PHI leak, jeopardizing the institution's compliance, its reputation, and patient trust. Medical investigation requires a scalable, automated bridge between the clinical silo and the research environment.

Research Assistant replaces fragmented, labor-intensive workflows with a precision extraction tool designed for speed and scale. By allowing investigators to define cohorts internally and trigger an automated data pull via a CSV upload, the system removes the operational bottlenecks of manual de-identification. Research Assistant also works to eliminate the liability of human error. Its automated redaction capabilities secure both metadata and pixel-level data, ensuring that large-scale medical investigation remains compliant, protected, and fully scalable.

‍Preserving the Patient Story: The Longitudinal Challenge

One of the greatest hurdles in clinical research is preserving the chronological integrity of a patient's journey. In a clinical setting, dates are essential for care coordination, scheduling, and historical context. In a research setting, dates are considered PHI and must be removed or transformed to comply with HIPAA Safe Harbor requirements.

However, if you simply delete the dates, you destroy the research value of the data. A researcher needs to know the interval between a baseline scan and a follow-up after a specific intervention. Without that longitudinal context, critical clinical insights and analytical value can be lost.

To maintain both privacy and scientific rigor, Research Assistant uses a smart randomization process to mask visit dates. Rather than stripping date information, the system shifts the entire timeline by a random number of days. This shift remains consistent for every study associated with that specific patient, ensuring the longitudinal relationship between scans stays intact while keeping the data fully de-identified and HIPAA-compliant.

The system automates complex HIPAA Safe Harbor requirements that teams may overlook, such as automatically aggregating all patient ages over 89 into a single category and masking ZIP codes for any geographic unit containing 20,000 or fewer residents. 

Researchers receive a longitudinal timeline where they can see that a follow-up scan occurred exactly 90 days after an initial procedure, but they never see the actual dates. This maintains both patient privacy and scientific rigor, allowing AMCs to conduct accurate, meaningful research without compromising compliance.

Beyond the Metadata: Unlocking Narratives and Pixels

True research interoperability requires more than just moving a DICOM file; it requires the intelligent transformation of the entire diagnostic package. To be useful for clinical trials, imaging data must be accompanied by its context, which is often trapped in narrative radiology reports.

• The Narrative Gap: Radiology reports are unstructured and packed with identifiers. Research Assistant connects directly to the EHR to retrieve relevant reports, then processes them using NLP within a secure, dedicated environment to identify and mask PHI while preserving the medical substance. This allows the radiologist's observations and diagnostic conclusions to follow the image out of the EHR silo and into the research lab.
• Pixel Scrubbing: Research Assistant allows users to define blackout zones for images based on modality (e.g., ultrasound) and remove burned-in PHI. This process occurs in a secure cloud environment where studies are scanned, scrubbed, and immediately returned to the institution's specified output folder.

The Re-Identification Safety Valve

One of the most significant concerns for AMCs moving data out of clinical silos for research is the black-box effect. Once data is fully de-identified, what happens if a researcher discovers an incidental finding?

Research Assistant uses a re-identification safety valve that leverages an underlying SQL database to maintain a verifiable audit trail and ensure that every data movement and transformation is documented for HIPAA compliance and oversight. Because the de-identification occurs on-premise within the AMC’s own network, the system maintains a secure, internal audit database. This allows the institution, and only the institution, to link a research study back to a specific patient if an incidental finding requires immediate care. 

Enabling the Modern AMC: Consortiums and Data Strategy

When AMCs stop treating radiology as a clinical endpoint and start treating it as a shared resource, the possibilities for institutional growth expand exponentially:

• Consortium Collaboration: AMCs can seamlessly participate in multi-site research networks without the overhead of manual data preparation.
• Internal AI/ML Development: By providing data scientists with high-quality, sanitized datasets, AMCs can drastically reduce the data cleaning phase of AI development, moving from pilot to publication faster. Research Assistant offers a turnkey solution for standardized compliance or a customized solution that allows investigators to determine specific values and parameters required for their unique protocols.
• Sustainable Research Funding: Secure, high-quality imaging data strategies empower IT and research informatics departments to support data monetization and industry partnerships, creating new revenue streams to fund future discovery.

Unlocking the Future of Discovery

The EHR was the vital first step in the digitization of the academic medical center, but for clinical research, it cannot be the final destination. To solve the most complex medical issues, we must liberate imaging data from its clinical silos.

By implementing automated, on-premise infrastructure that prioritizes both privacy and longitudinal accuracy, academic medical centers can turn their vast repositories of imaging data into a liquid asset—powering the research that will define the future of medicine. 

If manual de-identification is slowing down your research, we'd love to help. Schedule a discovery call with the Medicom team today to learn how Research Assistant can automate PHI removal and accelerate your institution's path to clinical discovery. Grab the Research Assistant product overview here.

About the author: With 17 years in imaging and experience at every level from PACS administrator to enterprise director, Nathan Opitz oversaw Sanford Health's full imaging ecosystem, including Epic Radiant and Epic Cupid.