LAMINATE / How it works
From the scanner to the report,
and back.
LAMINATE connects to your PACS through a small gateway at your site. The gateway removes patient identifiers before anything leaves, an automation rule can start the analysis, and the result can return to where the study came from.
The journey of a study
- 01Your site
The scanner or PACS sends the study
Over DICOM to the on-site gateway. Only MR, CT and PT images are accepted; anything else is refused and logged at the gateway.
- 02Your site
Identifiers are replaced
Patient name and ID become a pseudonym, dates are shifted, and private tags are removed, all before the images are written to disk.
- 03Your site
The study is packed and sent
A minute after the last image arrives, the gateway zips the study and uploads it over HTTPS with its own key. It starts every connection itself.
- 04The cloud
The study is checked again
The cloud de-identifies every upload a second time, then files it as a subject, session and scans in the workspace.
- 05The cloud
Automation decides
If an administrator has switched on a rule for this gateway and the study matches, the analysis is queued. Otherwise the study simply waits in the workspace.
- 06The cloud
The analysis runs
Queued, preparing, running, finalizing, completed: the status shows in the Analysis Studio. The result opens in the viewer for review.
- 07Your site
The result returns
For single timepoint analysis, the FLAIR* series goes back to the gateway, which restores the original identity and sends it to the PACS nodes you chose.
Automation: when does an analysis start by itself?
Nothing runs automatically until an administrator saves and switches on a rule. Each gateway has one rule for single timepoint analysis and one for longitudinal analysis. A rule has three parts.
When to run
A filter on the study, for example: the modality is MR and the study description contains BRAIN.
Which scans to use
For each role, a rule that picks the series. Single timepoint needs a FLAIR and an SWI, and can take a phase series. Longitudinal needs a FLAIR and a T1-weighted series from each of two sessions.
After completion
Fixed, not a setting: the result is registered in the workspace, and a single timepoint result can be sent back to PACS.
- One click to start. “Use standard setup” fills in sensible defaults: MR studies only, with the FLAIR and SWI series picked by name.
- Single timepoint runs per study. Longitudinal analysis looks at the last two studies in one upload, and needs a FLAIR and a T1-weighted series in each.
- Repeats are skipped. A study that was already analysed with the same series is not run twice.
- Uploads you make by hand never trigger a rule. Rules apply to studies that arrive from a gateway.
- If queueing fails, for example when credits run out, LAMINATE tries again for about a day and tells your administrators, who can also retry from a list.
- Automatic runs use your workspace defaults. Lesion segmentation is off unless an administrator has turned it on, and runs use credits.
What leaves your site
- Pseudonymised images (the patient ID is an ANON code, dates are shifted)
- Scanner and protocol details, and series and study descriptions
- Pixel data, including any text burned into the image
What stays
- Patient name, ID, birth date and real study dates
- The key that protects the original identity
- The log of who looked an identity up (written on site, tamper-evident)
Privacy, in plain terms
- The original identity is kept only at your site, encrypted with a key your hospital holds. The cloud keeps an unreadable copy and has no key.
- Someone signed in to your workspace, on your hospital network, can see the real identity next to the pseudonym. Each lookup is written to a tamper-evident log on site.
- Data in the cloud is pseudonymised, not anonymous. Dates are shifted and some descriptions are kept so that rules and comparisons work.
- Text burned into image pixels is not removed. Scanners should be set up so none is added.
- The gateway starts every connection to the cloud. Your hospital still allows DICOM in on the gateway’s port, and your staff’s browsers reaching the gateway over the hospital network.
- The gateway’s key is shown once when it is created, can be rotated at any time, and stops working the moment it is.
Notifications
| When | How | Who |
|---|---|---|
| An analysis completes or fails | In the app and by email | The person who started it, and workspace admins |
| An analysis is cancelled | In the app | The same people |
| An upload from the gateway finishes or fails | In the app | The uploader and workspace admins |
| An automatic run could not be queued | In the app | Workspace admins, with a Failed dispatches list to retry from |
| Credits are running low | In the app and by email | Workspace admins and owners |
Each person chooses which analysis emails they get, on a notification settings page. There are no webhooks, and no alert yet when a gateway goes offline.
Good to know
- The gateway uses classic DICOM (storage, query, retrieve). DICOMweb, HL7 and FHIR are not supported.
- Only single timepoint results return to PACS today, as one derived FLAIR* series. The lesion mask and the report stay in LAMINATE.
- If the cloud is unreachable, the gateway keeps the study and retries for days. If the gateway is offline when a result is ready, the result is sent when it reconnects, for a limited number of attempts.
- The pixel data of every study is stored in the cloud workspace, so hosting location and access should be agreed with your data protection team.
This application is intended for research and educational purposes only. It is not FDA-approved and should not be used for clinical diagnosis, treatment decisions, or patient care.
What a demo can show, end to end
With a test gateway and synthetic or public data, we can walk through the whole path in one sitting.
- Open the gateway page: it shows online, its AE title and counters for studies received, matched and queued.
- Show a PACS node and what it is allowed to do: receive from, send results to, or both.
- Switch on the standard automation rule, and test it against a DICOM file in the browser.
- Send a test study from a DICOM sender and watch the original patient name on the sending side.
- See the study arrive in the workspace under a pseudonym, moving through its upload stages.
- Watch the analysis go from queued to completed in the Analysis Studio.
- Open the result in the Single Timepoint Viewer.
- See the FLAIR* series appear in the PACS, under the original patient and study.