Improving Endoscopy with AI

Dr Hagen Wenzek and his team at CorporateHealth International (CHI) have collaborated with Intel to use Artificial Intelligence (AI) to optimise the process by which they conduct video capsule endoscopies: ensuring patient safety and freeing up valuable time for nurses and doctors.

Many patients, and their families, will be familiar with the rigours and demands of medical examinations, and the stress and disruption these can cause to daily life. For medical professionals too, each procedure takes valuable time away from analysing medical results, and other front-line patient support.

Any way to decrease the time spent on such procedures is massively beneficial to everyone involved. CHI is an organisation doing just that.

Video capsule endoscopy is a technology that is used to take pictures of the small and large bowel, avoiding diagnostic colonoscopies – which are invasive and time consuming. CHI is pioneering the use of AI in the procedure, to make it even more efficient.

How it works

That small but powerful tool at the heart of CHI’s solution is a video camera the size of a vitamin pill that’s swallowed by the patient. As it travels through the digestive tract, it takes up to 400,000 images of the lining of the patient’s mucosa (the inner part of your bowel).

Unlike a traditional colonoscopy, it’s minimally-invasive and does not require a doctor present to conduct. The patient simply swallows the capsule and wears a belt that contains an antenna to receive all the data from the capsule. The equipment is then returned to CHI, who then begin analysing the video data captured.

The process of analysing the video is:

  1. Video footage captured by the capsule is reviewed by a team of specialist nurses, who document their findings
  2. The nurses send a pre-reading over to a doctor, who validates the findings
  3. Validated findings are then sent to a gastrointestinal (GI) specialist, who makes a diagnosis for the patient

Introducing AI into the endoscopy process

Healthcare is an area that’s rife with opportunity for AI implementation: the wealth of patient data available in the industry means there’s plenty of information to train models on.

Like many organisations experimenting, trialling, and iterating with AI, CHI has adopted a ‘toe in the water’ approach. This exploratory strategy is aimed at proving the value step-by-step, rather than automating every stage of the process immediately. Currently, CHI is developing a solution that helps to confirm and add further precision to the findings of the specialised nurses at stage 1 of the process outlined above, in a project funded by InnovateUK with the National Health Service (NHS) in Scotland and England, and the Catapult S.A., Wolfram Research, Medilogik, and Universities in Barcelona.

Over the next few years CHI is hoping to optimise the diagnostic process further: for the majority of procedures within two years it aims to remove the need for the specialised nurses at stage one, and within four years the doctor at stage two. Doing so will save these busy professionals a huge amount of time – analysing the video for a single patient takes hours of attention. Time not spent analysing video can be spent on patient care instead.

Once we have proven that AI works on the pre-reading level we can start introducing it so that maybe in a year or two after that the doctor can directly look at an AI generated report, and then take her or his medical expertise to validate the video, the outcomes, and create a report that goes to the referring gastroenterologist.

Dr Hagen Wenzek, Managing Director, CorporateHealth International

Regulations concerning the movement of patient data

Patient data, naturally, is anonymised by CHI. But the Scottish NHS also requires that patient data remain within the UK – which ruled out any processing at CHI’s Danish headquarters and would make using a cloud solution complicated.

Through the Intel® AI Builders program, Intel provided CHI with its own on-premise infrastructure.

Two servers fitted with Intel® Xeon® processors now run in the CHI Inverness offices, meaning the patient data never leaves the UK. This resulted in significant cost savings for the start-up organisation, as well as offering impressive performance: the Intel servers installed have cut the time to process one video from two hours to 20 minutes.

Reece Moyes, Data Engineer, CHI

Spotlight on: Intel® Xeon® Scalable processors

Most AI today, including deep learning training, already runs on Intel® Xeon® processors. With this in mind, Intel® Xeon® Scalable processors have been rebuilt to amplify the capabilities of existing enterprise IT infrastructures.

Intel® Xeon® Scalable processors are built specifically to run high-performance AI workloads alongside data centre and cloud workloads.

For more information, click here.

Intel provides invaluable AI expertise

But it’s not just hardware – Intel also supports CHI in an advisory role. According to Dr Wenzek:

The Intel® AI Builders Programme has really helped us because they were able to provide us not just with hardware, but more importantly with some of the knowledge on how to set up that hardware, how to make sure it's secure, how to make sure that it's embedded in the network the right way - because it holds patient data (even if anonymised). For us to hire all these people to have that knowledge in house or pay consultants would have been quite prohibitive.

To sign up for the Intel® AI Builders program today

Intel technologies' features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system configuration. No product or component can be absolutely secure. Check with your system manufacturer or retailer or learn more at

Intel does not control or audit third-party data. You should review this content, consult other sources, and confirm whether referenced data are accurate.

Intel technologies' features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system configuration. No product or component can be absolutely secure. Check with your system manufacturer or retailer or learn more at

Intel, the Intel logo, and other Intel Marks are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries.

Other names and brands may be claimed as the property of others.

© Intel Corporation