A clinical archive becomes a research-ready cohort in five steps: digitize, link, connect, match, and deliver. Each step removes one reason a researcher cannot find or use material that already exists. The specimens and records stay with the institution that holds them. What changes is whether anyone outside the archive can tell what is there.
Most of the data research needs already exists. FFPE blocks, glass slides, pathology reports, and clinical records sit in lab and health system archives, in separate systems, indexed for patient care rather than for research questions. An FFPE block is a tissue sample preserved in formalin and embedded in paraffin wax, the standard way pathology departments store tissue for years. The steps below describe how that material becomes discoverable, and what each step asks of the institution.
Step 1: How are glass slides digitized for research?
Digitizing turns each glass slide into a whole-slide image. A whole-slide image is a high-resolution digital scan of an entire microscope slide, often several gigapixels in size, that can be viewed and analyzed like the glass itself.
In the Biobase digitize step, each slide passes through six stages:
- Scan: the glass slide becomes a gigapixel image.
- Redact: the slide label, which can carry identifiers, is hidden from the image.
- Watermark: an invisible mark records the image’s provenance.
- Quality-check: masks identify tissue, blur, and folds, so unusable regions are flagged before anyone relies on them.
- Embed: a foundation model converts image tiles into vectors, numerical summaries that make the image searchable by its visual content.
- Catalog: the slide becomes searchable and deliverable.
The result of the digitize step is a slide image that has been scanned, quality-checked, and is ready to use. The glass slide stays in the archive. The slide digitization guide covers each decision in more detail.
Step 2: How are specimens and records linked to the same patient?
Linking matches each FFPE block, whole-slide image, pathology report, and clinical record to the same patient. Without that link, a researcher can find a slide but cannot tell the diagnosis, stage, or treatment of the patient it came from.
The linking step does three things:
- It ties the physical specimen (the FFPE block) to its scanned, quality-checked whole-slide image.
- It turns the pathology report, usually free text, into structured fields such as diagnosis, anatomical site, and stage.
- It connects clinical data, including diagnosis, stage, treatment, and pathology findings, to the same patient.
Linked records are what make a cohort usable. A request for “stage III colorectal cancer with matched slide images and clinical data” can only be answered when all three are attached to the same patient. The linkage guide explains how the records are matched.
Step 3: How does an institution connect its systems without moving its data?
Connecting means integrating the institution’s existing systems with the Biobase network using a method that fits the institution. There is no single integration method. Biobase matches the method to three things:
- Organizational needs: how the institution’s systems and teams work today.
- Compliance posture: the de-identification and review steps the institution requires.
- IT requirements: security reviews and how the IT team prefers to connect.
De-identification is the removal of information that could identify a patient, following standards such as the HIPAA de-identification guidance from HHS. In the Biobase model, de-identification, lineage, and audit trails are enforced by the pipeline itself.
Discovery is federated. Federated means patient-level data stays at the institution while researchers search: the network carries queries and governed results, not copies of the archive. The source archive never moves. A dataset leaves only when a specific request is approved and delivered under governance, as described in Step 5. The institution decides what is discoverable, to whom, and under what terms. Biobase connects to the systems the institution already runs, with no migration and no new data warehouse. More detail is on the Healthcare Partners page, and the integration guide compares federated and extract approaches.
Step 4: How are researcher requests matched to what an archive holds?
Matching answers the question a researcher asks before committing to a study: does the material exist? A researcher describes a cohort in plain English, for example “Stage III colorectal cancer: FFPE blocks with matched whole-slide images and clinical data.” Biobase agents check every partner site and return feasibility counts.
A feasibility count is the number of patients or specimens at a site that appear to meet the stated criteria. It is an estimate of availability, not a reservation. One site might return 118 matching cases, another 42, another none (illustrative counts). The feasibility count guide explains what those numbers do and do not tell a researcher, and the cohort description guide shows how to frame the request.
For the institution, matching means its collections are seen by researchers whose requirements they actually fit. Counts and attributes are shared; the data itself does not change hands at this stage.
Step 5: How does a cohort get from request to delivery?
Delivering runs governance and fulfillment in one flow. Governance covers four steps:
- Compliance review: confirming the request fits the institution’s policies and the applicable regulations.
- IRB review, where it applies: an institutional review board (IRB) is the committee that oversees research involving human subjects under the Common Rule. The institution determines whether a study needs IRB review, an exemption determination, or another approval.
- Chain of custody: a documented record of who held a specimen or dataset, when, and under what conditions, from release to receipt (chain of custody, explained).
- Pricing: agreed commercial terms for the material and the work to prepare it.
When governance is complete, the cohort is delivered: physical specimens, digital datasets, or both. Suppliers are paid on delivery. Partner institutions recover their costs and share in the value their data creates. Agents move the process; humans approve the consequential steps. The request-to-delivery guide walks through each of them.
What does an institution need to start?
An institution needs a collection-level picture of what it holds, not a finished catalog. Specimen types, preparation formats, indications, and approximate scale are enough for a first conversation. The collection preparation guide gives a starting structure, and the biospecimen request guide shows what researchers will ask for.
The National Cancer Institute’s best practices for biospecimen resources describe the quality and documentation standards that make a collection useful for research. Do not include patient identifiers in an initial enquiry.