Surgical computer vision

Tracking what
imaging hides.

Real-time positional tracking across changing imaging conditions in surgical video. Multiple regions of interest, continuous through lighting transitions, with calibrated confidence. Built to integrate into medical devices.

FLUORESCENCE WHITE LIGHT CONFIDENCE 100% 88% 66% 54% 82% 99% landmark tracked region Illustrative concept only

What we build

SymmetryLab develops computer vision modules for surgical navigation systems. Our focus is on a class of problems that has been consistently difficult for endoscopic and minimally invasive surgical video: maintaining spatial reference when imaging conditions change.

Features of interest are identified in the imaging mode where they are most visible, then continuously tracked as the surgeon switches between modalities, across lighting changes, contrast washout, and repeated transitions during a procedure. Multiple regions are tracked simultaneously, with confidence reported per region.

The technology is designed as a software component for integration into regulated medical devices by qualified manufacturers. It is not a standalone clinical product.

Detect & reference all regions, blue light Anchor to anatomy stable landmarks Track & reconstruct white light, all regions Calibrated confidence shape vs reference lighting switch Multi-region overlay + per-region confidence Illustrative. Not actual output

Multi-region tracking

Multiple regions of interest detected and tracked in parallel, each with independent positional state and confidence. The number of regions is determined by what is found in the source modality, not fixed in advance.

Continuous through transitions

Tracking persists across imaging mode switches and continues frame-to-frame through extended sequences. Repeated lighting transitions during a single procedure are handled without manual reacquisition.

Geometrically grounded confidence

Per-region confidence is derived from geometric consistency between the live reconstruction and the reference captured at the point of detection. Not a black-box score, but a measurable signal the integrating system can act on.

How it works

Conventional tracking fails when imaging conditions change. We separate what is tracked from what is inferred. This makes the system robust to exactly the transitions where others break.

01

Detect in optimal conditions

All features of interest are identified in the imaging mode where they are most visible. Position and shape of every region are captured at peak signal quality and stored as a reference set, not approximated.

02

Anchor to stable anatomy

Each region's position is expressed relative to anatomical landmarks visible across imaging modes. Landmark geometry is what we track, not pixel appearance, which makes the system robust to the visual change at the modality boundary.

03

Reconstruct and report

When conditions change, landmarks are re-identified and tracked continuously. Every region is reconstructed in parallel. Confidence is computed by comparing each live reconstruction to its reference, giving the device a calibrated signal per region.

Where it applies

The underlying methods generalise across any surgical context where imaging conditions vary or features become intermittently visible.

Fluorescence-guided procedures

Tumour and tissue localisation persisting through illumination mode changes. Multi-region positions maintained from fluorescence detection through white-light resection.

Contrast-enhanced imaging

Feature reference maintained as contrast washes out. Spatial awareness preserved through the full contrast cycle without requiring manual reacquisition.

Endoscopic and laparoscopic navigation

Anatomical reference maintained under variable lighting, viewpoint change, and instrument occlusion. Robust to the non-linear motion of flexible scopes.

Multi-modality video

Spatial correspondence across imaging modes within a single procedure. Supports devices that switch between two or more imaging modalities intraoperatively.

Built for regulated environments

SymmetryLab software is developed as a supplier-grade component, intended to sit inside CE-marked and FDA-cleared medical devices manufactured by our clients. Our development practices and deliverables are designed for the expectations of regulated medical software.

We deliver software components, integration support, and documentation. Not consumer-facing products. Our clients remain the device manufacturer of record; we support their technical file and quality management processes as a qualified supplier.

Live demonstrations on real surgical video, technical briefings, and integration discussions are available under NDA following an initial qualification call.

Request technical brief →

Start a conversation

For technical briefings, integration discussions, or evaluation requests, reach us directly. Live demonstrations available under NDA.

Get in touch

contact@symmetrylab.co.uk

SymmetryLab Ltd. registered in the UK. All rights reserved.
The technology described on this site is a software component intended for integration into medical devices by qualified manufacturers under their own regulatory frameworks. SymmetryLab software is not itself a medical device, is not CE-marked or FDA-cleared as a standalone product, and is not intended for direct clinical use.