The InPractice flow
Connect the records and images you already create. InPractice protects and links them into an evidence graph, while your EMR stays the source of truth.
Your EMR. Your workflow. Your control.
Two paths. One protected flow. Clinical records and images are processed in parallel, with identity protection and structured context.
01Your existing system
02Two parallel pipelines
03One evidence graph
Start to continuous evidence
Scroll through the deployment path from the clinic's current EMR to a governed cohort that updates every week.
The source stays put
The evidence layer connects to the clinic's existing system of record. The clinic keeps its existing workflows, software and operating rhythm. There is no ERP-style migration and no clinic-wide change-management program.
Secure connection
A lightweight, on-premise agent reads approved data from the clinic's existing EMR. The connection is read-only and designed to work alongside established workflows without interrupting patient care.
Clinical pipeline
The initial cohort is de-identified, harmonized across formats and units, then mapped into the event-based InPractice ontology. Encounters, treatments, observations, images and outcomes become linked research signals.
Vision pipeline
Computer vision performs feature extraction, quality assessment, de-identification when required and structured analysis. This work runs alongside the clinical pipeline rather than waiting for it to finish.
Evidence layer
Clinical events and image-derived features resolve into one governed graph. Researchers can form cohorts around treatment, anatomy, device, product, observation and outcome without exposing direct patient identifiers.
Ongoing rhythm
After the initial load, the pipeline handles new or modified records and newly added images. The work becomes incremental, faster and aligned with the clinic's normal operating cadence.
Parallel by design
Structured records and clinical images move at the same time, then meet inside the ontology.
Clinical records
24–48 hours
Approximately 30,000 patients
Clinical images
24–72 hours
Approximately 100,000 images
Timing note: these ranges describe a best-case scenario. Actual duration depends on EMR access, data quality, governance review, data volume and available CPU or GPU resources.
After the initial load
Weekly updates only process new or modified records. Newly added images follow the same incremental pattern.
LLMs, computer use and computer vision are also making migrations between systems of record easier. That progress is useful context, but it is not a prerequisite for InPractice because the clinic does not need to migrate.
1–2 hours
New and modified records only
Minutes to hours
Only newly added images
The cohort grows while the clinic keeps caring for patients.
Inside the evidence graph
A treatment is one part of the story. Explore how records, anatomy, observations and images connect to give it context.
Three illustrative pathways. One connected model.
Three invented pathways link records, encounters, treatments, anatomy, observations and images through one shared graph. Each connection links two records; it does not establish a clinical effect. Select any node to read its facts and its incoming and outgoing relationships.
Illustrative examples · No real patient data
All identifiers, treatments and observations are invented. Connections show record relationships, not clinical results or evidence of treatment effectiveness.
07 / The result
Without replacing the EMR. Without changing how the clinic delivers care.