• Nederlands

Sector

Healthcare & life sciences

From image analysis in the hospital to monitoring in research barns. Eight of our eighteen cases sit in this domain, alongside academic hospitals, healthcare companies and Wageningen University.

A light laboratory bench with a microscope, a rack of slides and a teal wash bottle in daylight

The situation

Here the data is image, signal and variation

In this domain the data rarely arrives in tidy rows and columns. It is an IHC image, a retinal photograph, a picture of an infant skull, or a series of measurements from a barn. Two recordings of the same phenomenon also look different: staining varies, lighting varies, and image size varies with it.

That variation is the actual work. In the tumour cell detection for the Antoni van Leeuwenhoek, adaptive colour analysis sits in the system precisely because IHC staining differs per specimen, and image filtering runs first to reduce the chance of errors. A model that works on laboratory images and stumbles in practice is not a rarity here but the normal starting point.

Why explainability comes first here

What a model delivers here is a preliminary assessment, not a verdict. In the tumour cell detection the judgement stays with the pathologist; what the system provides is a consistent assessment of images that would otherwise be reviewed one by one. Reproducibility therefore matters more than the last few percent of accuracy.

We believe technology should be reliable, transparent and explainable, particularly where decisions affect people. In healthcare that is not a principle we add; it is a requirement the sector already sets.

Roles

What we do here

Twentynext works in four ways: advice, people, projects and maintaining what runs. In this domain these three occur most often. What advice means here, you will find under data consultancy in healthcare.

Turning image and signal into a repeatable assessment

Models that assess images or measurement series in a way that is repeatable and stays traceable: which processing was applied, and on what the decision was based.

Research with clinical and academic partners

Part of this work is research, with hospitals, healthcare companies and universities. Part of our R&D is exactly that: not finished yet. That is a status, not an interim phase that resolves itself.

An algorithm that runs inside someone else’s product

Sometimes the result is not a system of our own but a component: the algorithm Twentynext developed for Skully Care runs inside their app.

Evidence

Where this has already been done

Case

Measuring skull deformation

Skully Care measures skull deformation in babies from a single photo. Twentynext developed the algorithm that automates that measurement; their app is used by more than 900 therapists in 30 countries.

Read the case

Case

Detecting and classifying tumour cells

Research with the Antoni van Leeuwenhoek: analysing IHC images automatically and assessing clusters for severity and aggressiveness. The judgement stays with the pathologist.

Read the case

Solution

Medical diagnostics

Assessing images with AI: where this helps and where it does not, and what is needed before a model becomes usable in practice.

View the solution

Scope

What we do and do not claim here

We build algorithms and systems. The clinical judgement stays with the healthcare professional, and that is not a formality: our systems deliver a preliminary assessment that a doctor or researcher then weighs.

We are not a manufacturer of medical devices and do not present ourselves as one. Certification, clinical validation and market approval sit with the party bringing the product to market; we deliver the data and AI work underneath.

Part of what is described here is research rather than a delivered product. We prefer to say that beforehand rather than afterwards. If you are working on a question in this domain and the data or AI side is missing, that is the conversation we are glad to have.

View all cases and projects

Want to know where your organisation stands? Take the free Data Maturity Scan: sixteen questions, immediate result.

Work with us

Realise your project together?

The people who build it also run it afterwards. Eindhoven, since 2014.

Martijn van Grieken

Martijn van Grieken

Director Data & AI

Get in touch