Security
Patient data doesn't leave your network without your explicit permission.
Neurmorph processes studies in your PACS environment. Here's how the data flows, what we store, and how to configure it.
TLS 1.3
encryption in transit
AES-256
at-rest encryption
24h
max PHI retention (hosted mode)
Data flow
What moves, where, and when.
On-premise deployment
Neurmorph processing node runs inside your datacenter or cloud VPC. DICOM studies never leave your network. Model weights downloaded once at install; inference is local.
Hosted processing (opt-in)
Studies sent to Neurmorph's processing environment, designed with HIPAA administrative, physical, and technical safeguards. Annotated DICOM SR written back to your PACS. Studies deleted within 24 hours of SR delivery. No long-term storage of PHI.
Technical safeguards
Technical controls in every deployment.
TLS 1.3 in transit
All DICOM transfers over encrypted channel. DIMSE-TLS or DICOMweb over HTTPS.
AES-256 at rest
Temporary study storage during processing encrypted at rest.
Role-based access
Admin and radiologist roles. Audit log of every annotation action with timestamp and user ID.
BAA available
We sign Business Associate Agreements same day. Template available on request.
De-identification
PoC mode uses de-identified studies only. DICOM tags stripped per PS 3.15 Annex E.
FAQ
Questions we hear from hospital IT.
In hosted mode: no. Studies are deleted within 24 hours of SR delivery. We retain only aggregate model performance metrics, fully de-identified.
Yes. On-premise deployment runs inference inside your datacenter. No outbound traffic required.
CT Image Storage (1.2.840.10008.5.1.4.1.1.2), MR Image Storage, Enhanced CT, Chest CAD SR, and Comprehensive SR for output.
No. Neurmorph operates entirely within the DICOM layer. No HL7, no FHIR, no Epic access required.
Request the security documentation package.
Includes: penetration test summary, BAA template, data flow diagram, and DICOM conformance statement. Available to qualified radiology departments and hospital IT teams evaluating Neurmorph.