Sylwester Cyrwus
CEO and co-founder
Technology executive in healthcare and life sciences. At QuantumMed, he leads business strategy, partnerships and commercial development.
SafeMedAI
Data for AI.
Control on your side.
We are developing a gateway for controlled sharing of health data. You define the recipient, purpose and scope.
SafeMed AI by QuantumMed
Control begins
before sharing.
Data preparation starts with a purpose. We determine what the recipient needs, what to remove or transform and how to preserve data utility.
01 / Types of data
Data comes in
many forms.
Names, addresses, dates and identifiers. Inside documents, tables and measurement series. See what changes in the copy, what stays and how duplicates can be removed.
01
Personal data inside text
A pseudonym replaces the name, contact details, address and test national ID are removed, and dates become less precise. The same rules apply inside the note. The visit description remains.
BeforeAfter
Synthetic examples of operations. Integration scope, risk assessment and data utility are defined for each workflow. A retained reversible link means pseudonymisation.
02 / Example: DICOM imaging
From study
to result.
Imaging is one use case. Here, a copy from PACS reaches the selected AI and the result returns to the right study.
A copy is retrieved from PACS for preparation. The original stays in place.
The policy defines the data scope, purpose and recipient. The identifier mapping stays local.
Transfer follows the agreed checks. The recipient returns a result.
The local mapping associates the result with the right study and supports a transfer record.
03 / Transfer rulesSafeMed AI
Transfer conditions.
A clear decision.
The policy defines the permitted purpose, recipient and data scope. Changing any of these conditions affects the transfer decision.
Transfer being assessed 01 / 03
MatchConditions match the policy
- 01PurposeMatched
- Summarising records
- 02Recipient
- Partner A
- 03Scope
- Content and transfer identifier
Transfer statusChecking
Verifying conditions
Recorded outcome
TRANSFER_ALLOWEDA record of every decision
The policy and its version. Completed checks. The decision and time. This record is intended to explain why data was transferred or held back.
Worth knowing
Before the first
transfer.
Where do we start?
With a specific workflow, data type and sharing purpose. We agree on the recipient, necessary scope and evaluation criteria, then select demonstration material and an integration approach.
Do the data automatically become anonymous?
Removing a name or changing an identifier does not resolve every identification risk. When a reversible link is retained, we explicitly describe it as pseudonymisation. Assessment depends on the data, purpose and recipient.
Is every format already supported?
The examples show different data types and intended preparation rules. Support for a particular format, source and recipient must be confirmed within an agreed integration and testing scope.
What can we examine together?
The flow, required data scope, responsibilities and evaluation criteria. Demonstration material and work scope are agreed before any data are shared.
Leadership
Krzysztof Bokiej
CTO and co-founder
IT architect with experience in enterprise systems integration. At QuantumMed, he leads SafeMed AI architecture, technology development and the technical evaluation of integrations.
The next step
Let’s talk about secure health data sharing and how SafeMed AI could support a workflow in your organisation.
info@quantummed.eu