Inside the Salnus Surgeon Portal: A Browser DICOM Viewer with Integrated Orthopedic AI

In Brief

Most orthopedic imaging workflows make the surgeon move between tools: one program to view the scan, another to outline the bones, a third to plan, with the patient's images copied between them each time. The Salnus Surgeon Portal brings the whole sequence into a single browser tab. You open a CT, the AI outlines the bones on your own computer, an interactive 3D model builds in the same view, and the planning tools work on the same scan. No upload, no installation, nothing to switch between. This is an honest walkthrough of what is live today and what is still ahead. Salnus is for research use only and is not a diagnostic device.

One Portal, Not a Toolchain

The idea behind Salnus is that the steps belong together. A surgeon's preoperative routine usually crosses several disconnected programs, a viewer, a bone-outlining tool, a planning suite, each with its own login and file export, and the scan travels between them. Every hand-off costs time, and every copy of the images is one more place patient data has to be looked after.

Salnus runs the whole chain, view, outline, reconstruct, measure, plan, in one place, in the browser, with the images staying on the surgeon's computer. The fact that it is one connected workflow is the point. Here is what each part actually does.

What You See: The Viewer

The foundation is a proper DICOM viewer. It opens a standard CT series in the browser and shows it in the correct orientation and scale, so the reformatted views are anatomically true rather than stretched.

• Three planes from one scan: axial, coronal, and sagittal views, with linked crosshairs, so a point you click in one plane is shown in the other two.
• Measurement tools: length, angle, Cobb angle, area, and point tools, reached by right-click or a keyboard shortcut.
• Controls that remember you: scroll speed and layout settings that persist, because a trackpad and a mouse wheel do not scroll the same way.

If you want the wider landscape, our survey of browser-based DICOM viewers covers it. The point here is that the viewer is the surface the AI works on, not a separate product.

What the AI Does

This is where keeping everything together pays off. With a CT open, one click runs the bone-outlining AI, and the result comes back as a coloured overlay on the very slices you were already reading.

Bone outlining is the core. Done by hand, separating bone on a CT is slow and varies between operators; an AI-first pass the surgeon then checks turns a long manual task into a quick review. And because it runs on your computer rather than a remote server, the data-protection burden stays small, the scan is never uploaded.

From Scan to 3D Model

An outline on the slices is useful, but surgeons think in three dimensions. The portal turns the outlined bones into an interactive 3D model in the same panel as the slices, so you can rotate the anatomy, show or hide a single bone, and save the model for 3D printing, without leaving the case.

Doing this smoothly in a browser, next to three live views, takes care: the model has to look like bone rather than a blocky cast, and it has to stay stable on an ordinary clinic laptop. Those details are what separate a demo from something usable on a real case, and we treat them as part of the job.

Planning on the Anatomy

On top of the 3D anatomy sits planning. The first planning tool supports multi-ligament knee reconstruction: the surgeon marks the relevant anatomical points, and the tool helps lay out graft-tunnel placement on that anatomy.

One honest detail: marking those points today is manual, the surgeon places them, and the tool does the rest. Automatic point detection is in development but not yet reliable enough to put in front of a surgeon, so we have deliberately left it out rather than ship something that drops points in plausible-but-wrong places. Planning sits on the bone outlining, which sits on the viewer, one workflow, one scan.

What Is Live, and What Is Not

Keeping the workflow together is the strength; being clear about what it does and does not do is what keeps it trustworthy.

• Research use only. Salnus is a research tool, not a regulated diagnostic device, and is not a substitute for clinical judgement.
• Knee CT bone outlining is the validated core. MRI cartilage and soft-tissue outlining are in active development, not finished features.
• Full-limb alignment needs the right scan. A knee-only CT supports rotation and sizing measurements; full mechanical-axis angles such as HKA need imaging that includes the hip and ankle, which is a planned workflow, not something we claim from a knee scan.
• We sit beside robotics, not against it. Where an intraoperative robotic system fits, and where independent software planning sits next to it, is covered in our piece on robotic arthroplasty and complementary planning.

Saying plainly what a tool does not do is part of earning the right to be used on a real patient.

What Is Yet to Come

Today's portal is the foundation. The direction from here, all research-stage, is toward a fuller preoperative planning surface:

1. Outlining that waits for no one. The next step is to outline the bones the moment a scan arrives, so that by the time the surgeon opens the case, the result is already there. The aim is to remove the wait entirely, not just shorten it.
2. The multi-scan case builder. Bringing hip, knee, and ankle CT into one planning context unlocks true full-limb alignment, LDFA, MPTA, HKA, and the alignment frameworks that depend on the whole leg rather than a single joint.
3. Implant-agnostic arthroplasty planning. A knee- and hip-replacement planning workflow that treats the implant as data rather than a lock-in, so the plan is not tied to one manufacturer's catalogue.
4. MRI bone and cartilage outlining. Extending the on-device story from CT bone to MRI bone, cartilage, and soft tissue, the harder modality, in development now.
5. Automatic landmark detection. Replacing today's manual point-marking with detection the surgeon can trust, so planning gets faster without becoming a black box.
6. Evidence-linked case library. A teaching and reference layer that ties each case to its imaging, its AI output, and the relevant literature.

None of this comes with a promised date; it is the direction of a tool being built in the open, alongside the surgeons using it. Each piece ships when it is trustworthy, not before.

FAQ

Do I need to install anything? No. The portal runs in a standard browser. There is nothing to install and no server to maintain.

Are the patient's images uploaded somewhere? No. The portal is built so that viewing, outlining, and AI all run on your own computer; the images stay on the machine.

Which scans does the AI support today? Knee CT bone outlining, femur, patella, tibia, and fibula, is the validated core today. MRI and additional regions are in development.

Can I save the 3D model? Yes. The 3D bone model can be exported for 3D printing from within the case.

Is Salnus a diagnostic device? No. Salnus is for research use only and does not replace clinical judgement or a regulated diagnostic device.