What are the practical applications of AI in healthcare? Dr Christoph Zindel, member of the Managing Board at Siemens Healthineers, responsible for the Imaging and Advanced Therapies business segments, has a clear vision.

Healthcare IT News (HITN): Dr Zindel, you joined the Managing Board of Siemens Healthineers at the beginning of October. One of your stated goals is to champion digitisation and the use of AI in healthcare. From buzzword to applied technology?

Zindel: AI ceased to be a buzzword a long time ago, at least as far as Siemens Healthineers is concerned. For 20 years we've been working on ways to integrate AI - that's to say, algorithm-based intelligence - into our technology. But, and this is certainly true, today this is taking place on different levels of complexity and integration.

HITN: What do these levels look like? How far has AI progressed?

Zindel: The vanguard is clearly its application in everyday clinical practice, namely in the automation of clinical processes. The objective here is to increase efficiency using AI, not just by reducing the error rate, but also by raising the level of diagnostic precision, for instance through improved image quality.

HITN: How do you do that?

Zindel: Here's one example to illustrate more specifically: using a 3D camera, we can already automatically and optimally position patients in a CT scanner in the centre of the system so that the anatomical and functional images are subsequently more accurate, and we can do this with less radiation exposure. This obviously reduces costs, because examinations do not have to be repeated.

HITN: So what's next on the AI development agenda?

Zindel: That would be AI-based diagnostics. The issue we need to address here is: how can we integrate AI so that scanners and algorithms interact with each other, for example so that scan results are prepared in the best way possible for the physician? There are some promising approaches: AI algorithms can measure entire structures in the body, organs as well as vessels, and they can even identify abnormalities on clinical images. Essentially, entire clinical diagnostic processes can be digitally integrated.

HITN: Can you give us any other examples?

Zindel: We have developed a platform based on algorithms that compares individual CT scans with millions of other images, and tells the radiologist: look again more closely, there's an abnormality here. The Klinikum Großhadern of the Ludwig Maximilian University of Munich and the Klinikum Nürnberg are already using such a software solution for CT scans of the chest - we call this AI-powered radiology assistant the 'Rad Companion'.

A similar technological development adds AI-based decision support to clinical treatment pathways: patient data from various sources - image data, lab results, pathology data - are compiled in sensible form and coupled with algorithms that can recognise patterns. All of this is then compared with current treatment guidelines. This allows data to be evaluated and objectified and the next stage in the treatment process to be recommended. I myself have a medical background - surgery, to be precise - and let me tell you: I foresee excellent and relevant areas of application for developments like these, for example on tumour boards.

HITN: Cynics might say that relying on AI to help make diagnoses and treatment decisions carries a high element of risk...

Zindel: Systems like these have to be reliable, that goes without saying. Doctors and patients have to be able to rely on the data. That's why it makes sense for solutions like these to undergo a comprehensive approval process. The aforementioned Rad Companion for radiology, for example, has already received FDA approval and CE certification. The clinical decision support system - introduced as the AI Pathway Companion for the first time at the RSNA Congress - is currently being tested under clinical conditions, including at the University Hospital of Basel.

HITN: But your research goes a step further: you are working on a digital twin...

Zindel: Together with partners like the University Hospital Heidelberg, we are currently working on combining intelligent algorithms with data to create a virtual copy of an organ, that's correct. In this case, 'self-optimising networks' are used, where the aim is to be able to better understand medical and physical conditions. To a certain extent, this should make it possible to predict therapeutic viability and plan complex treatments.

Right now, we are researching a digital twin of the heart, but this technology can also be applied to other organs, in future perhaps to the entire human body - in the context of a lifelong, multidimensional individual health record. This would then be a step towards patient-centred individual medicine. The long-term vision of the future, based on big data analysis, goes even further: by merging millions of data sets into cohorts, we can identify and mitigate risks longitudinally. For example, a patient who is deemed to show a high risk of developing bowel cancer based on an analysis of all of their test results could be promptly recommended for a colonoscopy. This would be a giant step forward and would have a great impact.

HITN: So we've addressed the possibilities, the visions. But what do clinicians think about such developments?

Zindel: Right now opinion is heavily divided. Many are enthusiastic and want to pursue this avenue, not least because, in the context of staff shortages in their own clinics, they can see how AI could support their workflows. But compared with countries like China, where the healthcare sector can't implement AI fast enough, developments in Germany are much more restrained.

HITN: What in your opinion is the reason for such restraint?

Zindel: There are some very traditional fears, fears for one's own profession...

HITN: You mean radiologists are frightened of being replaced by machines?

Zindel: Exactly. Incidentally, a fear that I believe to be completely unfounded. Radiologists will always be needed to skilfully evaluate and classify the results of AI. Today - given all the advanced technology described - it really doesn't take a highly trained radiologist to count the lesions in a clinical image, for example. Their exceptional skills and expertise can be put to much better use elsewhere.

HITN: How do you hope to alleviate the anxiety?

Zindel: We want to make people more familiar with AI; the medical profession, sure, as well as people in general. We want to remove the buzzword label and turn AI into something that everyone can understand, whose value everyone can recognise. Let's forget the presentation slides, where AI is repeatedly theorised, and embed it into the minds of the people. That's why we're out there at trade shows and events, championing AI. We need to make AI accessible, that's the only way to ensure it receives what it absolutely deserves: trust.

The chief executive of Siemens Healthineers, Dr Bernd Montag, will talk about AI on 5 November in Berlin at  the HEALTH - The Digital Leaders event, the annual meeting of executives and decision makers in German healthcare. More information about the event can be found here.