Self-hosted medical imaging
PACSinfra is a self-hosted medical imaging platform built on Orthanc and OHIF. Upload DICOM studies, review them, and write reports, all on your own server, with patient data that never leaves your control.
The demo is a limited preview: each demo company is deleted after 24 hours and holds up to 10 studies (500 MB). Some features, such as the DICOM node and worklist, are not included.
Security
Each organisation sees only its own studies, verified by an automated isolation test suite. The image server is never exposed: every image passes through an API that checks permissions and records an audit entry for each request.
How it works →Patient identifiers are removed from DICOM metadata, including free-text fields and nested sequences, before anything reaches the image archive. Clinics can keep real identities in a separate application database. Research teams can run on pseudonyms only.
Limitation: De-identification covers DICOM metadata, not text burned into image pixels.
What's included
Studies are shared through projects, where each person is an owner, contributor or reader. Readers can be view-only.
Sign-in supports two-factor authentication (2FA).
Upload single files or ZIP archives, with validation and automatic ingest.
View every series in the browser with the built-in OHIF viewer. Study lists with search, filters and previews.
Draft, finalise, version and export reports to PDF. Dictate them by voice, recognised on the doctor’s own computer.
Send a study to a patient behind a password, and chat with them in real time on the same link.
Restricted studies can still be opened when needed, but access is always recorded with a reason.
An append-only record of every action, exportable to CSV.
Each user can query and retrieve only from the DICOM machines an admin has allowed them to use.
A full REST API with Swagger docs, for developers and integrations.
Architecture
Studies travel from the hospital to PACSinfra over an encrypted connection, through a DICOM node inside the hospital network or by web upload. Inside PACSinfra, the API checks permissions and records an audit entry before any image reaches a browser or the mobile app. We can set up this whole path for you on our own stack: the PACSinfra server, the DICOM node inside your network, and the encrypted link between them.
Hospital or clinic
Your local network
Modalities
CT · MR · X-ray · ultrasound
Existing PACS / RIS
Optional
DICOM Node
Or: web upload
DICOM files or ZIP archives from a browser
PACSinfra cloud
Your cloud account or your own server
Gateway
The only public entry point
PACSinfra
Private network, no public ports
DICOM worker
Validate · de-identify · ingest
PostgreSQL
App data, audit, job queue
Garage
S3 upload staging
Your team
Anywhere with a login
Web app + Viewer
Study list, viewer, reports, admin
Mobile app
Study list, image previews and reports on the go
Capacity and scaling
PACSinfra sets no limit on the number of studies or the amount of storage. Capacity comes from the server you run it on, and each part of the stack can be scaled on its own.
5,000
radiologists at once, zero errors
Each simulated reader opens the worklist, opens a study and loads every image of a series, just as the viewer does. Studies open in under 200 ms (p95).
~600
images per second, de-identified and searchable
The full pipeline, end to end: upload, de-identification, storage and indexing. A typical 300-image MR series is ready to read in about 5 seconds.
50,000+
studies per organisation, pages in ~3 ms
Tested with 50,000 studies in one organisation: the study list opens in about 3 ms. Every page is an index lookup, never a full scan, so it stays this fast as your archive grows.
Load and ingest figures measured with k6 on a single 5-CPU, 8 GB machine, with the load generator running on the same machine. No cluster, no CDN.
Growing past one server? I have done this before.
I have run PACS and medical imaging infrastructure in production for more than four years, for a university healthcare research group in Australia. As you grow, I can scale your installation with you: more DICOM workers, the database and image storage on their own servers, and more sites connected through DICOM nodes.
Product
Built-in medical image viewer
Scroll through every series, adjust window and level, measure and compare, right in the browser.
Find any study fast
Search and filter every study by patient, date, modality, project or uploader, then open it in one click.
Every action on record
See who opened which study, when, and under which permission. Entries can never be changed or deleted.
Projects decide who sees what
Group studies into projects and add colleagues as owner, contributor or reader. Each card shows how many studies it holds.
The demo is a limited preview. To keep it safe for public sign-ups, each demo company is deleted after 24 hours and holds up to 10 studies (500 MB in total), and some main features, such as the DICOM node and worklist, are not included. Book a call and we will walk you through the full product.
Book a callGetting started
How your team starts using PACSinfra once it is set up on your server.
Once PACSinfra is set up, you sign in as your organisation’s administrator, with two-factor sign-in.
Add colleagues by email. Each one gets a link to sign in, or signs in with Google.
Group studies the way you work: a clinic, a department, a trial or a reading list.
Choose what each person may do in that project: owner, contributor or reader.
Upload DICOM files or ZIP archives into the project, or let the DICOM node send them from your scanners.
Open studies in the viewer, write reports, by voice if you like, and share them with patients.
Owner
Manages the project: its people, their roles and which studies it holds. Can do everything a contributor can.
Contributor
Uploads studies, writes and finalises reports, and shares studies with patients.
Reader
Views studies and finalised reports. Can be set to view-only, without downloading files.
Studies are seen only through projects. An administrator manages people, but opens no study unless they are added to its project.
Technology
A lean, modern stack built for speed and safety: a compiled Go 1.27 API, a Python 3.12 imaging engine and PostgreSQL 17.
API
Go 1.27
One static binary, images streamed with zero buffering
DICOM engine
Python 3.12
De-identifies in memory, before anything is stored
Web app and viewer
Next.js 16 · React 19 · TypeScript 7
Bilingual, with the OHIF v3 viewer built in
Orthanc 25.6
A proven DICOM server, used as the image archive
PostgreSQL 17
Application data, the audit trail, the job queue and the image index
Garage 1.0
S3-compatible object storage for upload staging
Pricing
One organisation per licence. You may modify the code. You may not resell or redistribute it.
$2,000one-time
The full source code and a lifetime licence, with every feature and the first year of maintenance free.
✓ DICOM viewer, reports, voice dictation, share links and live chat, all for $2,000
Maintenance (debugging and fixes for any error): free for the first year, then $1,000/year
+$1,000one-time
We install and configure PACSinfra on your own server, cloud or Kubernetes cluster, and make sure it runs, so your team can start straight away.
+$1,000
A DICOM node inside your hospital network that receives studies from your modalities and existing PACS and forwards them over an encrypted connection, plus a modality worklist. Set up and connected for you, with its full source code.
| What’s included | Self-Hosted$2,000 | Setup+$1,000 | Add-on+$1,000 |
|---|---|---|---|
| The software | |||
| DICOM viewer in the browser | – | – | |
| DICOM upload and ingest | – | – | |
| Radiology reports, exported to PDF | – | – | |
| Voice dictation for reports | – | – | |
| Password-protected share links | – | – | |
| Live chat with patients | – | – | |
| Projects and permissions | – | – | |
| Two-factor sign-in | – | – | |
| Break-the-glass access | – | – | |
| Audit trail with CSV export | – | – | |
| DICOM gateway: each user queries and retrieves only from allowed machines | – | – | |
| REST API with Swagger docs | – | – | |
| The service | |||
| Full source code, handed over to your team | – | – | |
| First year of maintenance, debugging and fixes (free) | – | – | |
| Setup on your own server | – | – | |
| Migration of your existing studies | Quoted per project | ||
| Connecting your scanners | |||
| DICOM node connector | – | – | |
| Modality worklist | – | – | |
| Source code of the DICOM node and worklist | – | – | |
About
Bang Tran
陳業邦
I’m Bang Tran, the engineer who builds PACSinfra and writes its code. I have more than eight years of experience in the field, working across React, DuckDB, FastAPI and Kubernetes.
For more than four years I built and ran medical imaging and PACS systems in production for a university healthcare research group in Australia. PACSinfra is built on that experience, and on what running those systems every day taught me.
Medical imaging is a field I find genuinely interesting. When you choose PACSinfra, I set it up on your server, hand the full source code over to your team, help move your existing studies in, and look after it. Having run these systems in production, I can help you scale it as your sites, users and study volume grow.
4+ years
running PACS medical imaging infrastructure in production for a university healthcare research group in Australia
8+ years
in the software engineering industry
FAQ
On your own server. PACSinfra is self-hosted: studies, reports and the audit trail stay on infrastructure you control.
Patient identifiers are removed from DICOM metadata, including free-text fields and nested sequences, before anything reaches the image archive. De-identification covers DICOM metadata, not text burned into image pixels.
Yes. Clinics can keep real identities in a separate application database, apart from the image archive. Research teams can run on pseudonyms only.
Tenant isolation is verified by an automated isolation test suite, and every image request passes through an API that checks permissions and records an audit entry.
Through the DICOM Node, which sits inside the hospital network, receives studies from your modalities and existing PACS, and sends them to PACSinfra over an encrypted connection. You can also upload DICOM files or ZIP archives from a browser.
There is no built-in limit on studies or storage. Images live on a disk volume you can grow, and the API and DICOM worker hold no local state, so you can add capacity as you need it.
A single server, in your building or your own cloud. We can set it up on your server for $1,000 and hand the full source code over to your team; moving your existing studies in is quoted separately. I have run PACS in production long enough to help you scale it as your volume grows.
Studies are visible through projects. Each person has a role in each project (owner, contributor or reader), readers can be view-only without download, and emergency access is always recorded with a reason.
PACSinfra supports two-factor authentication (2FA) at sign-in.
Yes. PACSinfra has a full REST API with Swagger docs.
Questions about buying? See the pricing FAQ.
Try the demo, compare plans, or talk to us about deploying PACSinfra on your server.