Solutions · Rehab and digital health

For rehab and digital health: models that keep patient data where it is

Movelytics builds custom on-device movement models and on-premises language models for rehabilitation and digital health teams. The patient's video is analysed on the device and is not uploaded. A domain language model can run inside your own infrastructure. PoseTracker is not a certified medical device; we say so first.

Last updated:

Overhead reach · shoulder angle and trunk lineIllustration. Authored pose sequence, not live inference.

01Your problem

A programme followed at home is a programme nobody sees

Between two sessions, a patient is alone with a sheet of exercises. The practitioner does not know whether they were done, how many, or how. Remote follow-up only works if the tool can see the movement.

And in health, the usual way to see, sending video to a server, is the one thing many organisations cannot accept. Health data rules, hosting requirements and patients' trust all point the same way: the data should not leave.

02What becomes possible

What the technology makes possible

Movement measured on the patient's phone
Joint angles, range of movement, repetitions and holds, computed in real time on the device.
Keypoints that fit a protocol
A shoulder protocol and a lumbar protocol do not need the same points. With our own model, the lab can add the keypoints a protocol needs, including trunk and spine points.
A language model that stays inside
A domain-specific language model for rehabilitation coaches and physiotherapists, running on a device or on your premises, so that a patient note is never sent to a third party.
Older phones included
Patients do not all own recent phones. We optimise for low-end and five-year-old Android devices.

The pipeline is described on the technology page. For the use case seen from an app developer's side, PoseTracker has a guide on virtual physical therapy and one on measuring range of motion.

03What it is not

What PoseTracker is not

This matters for how you use it. A wellness or follow-up feature that shows a patient their own movement is one thing. A measurement used for a clinical decision is another, and it carries regulatory obligations that are yours as the manufacturer of the final product. We will tell you plainly which side of that line a planned feature is on, as far as the technology is concerned.

04Which route

Product route or custom route

  1. 01

    PoseTracker as a component

    For follow-up and engagement features, integrated by your developers, with our technical documentation.

  2. 02

    A custom pose model

    Keypoints and movements specific to your protocols, trained by the lab and measured on your target devices.

  3. 03

    An on-premises language model

    A domain model for your practitioners, deployed on infrastructure you control.

Projects go through a scoping call, a proof of concept and production. See work with the lab.

05Privacy

Privacy first, by architecture

The strongest privacy measure is not to have the data. On-device inference means the patient's video is not uploaded to Movelytics. We collect usage metering only: a number of API calls. We do not store videos, keypoints or analysis history, and we never sell data.

What your product stores, it stores in your system, under your responsibility as data controller, with the hosting your sector requires. Our data commitments describe our side.

06FAQ

Questions from rehab and digital health teams

Is PoseTracker a medical device?

No. PoseTracker is not a certified medical device and has no validation study against motion capture. Movelytics can supply technical documentation for its integration as a third-party component in a customer's product.

Does patient video leave the phone?

No. Pose estimation runs on the device and the video is not uploaded for inference. Movelytics collects usage metering only.

Can a language model run on our own servers?

Yes. The lab builds domain-specific language models, for example for rehabilitation coaches and physiotherapists, that run on a device or on premises, inside infrastructure the customer controls.

Can you add keypoints for a specific protocol?

Yes. Because Movelytics trains its own pose model, the lab can add custom keypoints and fine-tune the model for a protocol, then measure it on the target devices.

Rehab and digital health

Describe the protocol. We will say what can be measured.