Work · Medtech · MED/02 / 2019 → 2023
Seeing clearly at gigapixel scale
A pathology imaging platform deployed in leading laboratories worldwide. A digitised slide weighs billions of pixels; the pathologist must navigate it like a microscope, without latency or artefacts. Fluidity is not interface comfort: it is a condition of diagnosis.
The context
Digital pathology applies machine learning to slide-based diagnosis: biopsies digitised into images of billions of pixels, examined by pathologists whose every conclusion commits a patient. The platform equips leading laboratories worldwide; the screen replaces the microscope, provided it concedes nothing to it.
The work
Development of the WebGL / Three.js viewer: multi-resolution tile streaming, fluid deep-zoom navigation, colour fidelity. Rendering pipeline optimisation (+40% on high-resolution scans), decomposition of the monolith into microservices to control load, a GraphQL layer and a visual workflow editor to compose analysis pipelines, including the ML models that pre-annotate slides.
What remains
A daily working tool for teams that diagnose on screen what they used to diagnose under the microscope. And a transferable lesson: in critical systems, perceived performance is a clinical requirement, not a refinement.
- Three.js / WebGL gigapixel slide viewer with tile streaming
- Optimised rendering pipeline: +40% performance on high-resolution scans
- Microservice decomposition and load control
- GraphQL layer and visual editor for analysis workflows