Diabetes: AI skin scanner detects disease progression

Progressive diabetes is accompanied by gradual changes that occur at the micro scale of nerves and blood vessels. This leads to damage that goes unnoticed for a long time. A team from TU Munich applied the optoacoustic imaging method Raster-Scan Optoacoustic Mesoscopy (RSOM) together with AI to measure the severity of diabetes by assessing microvascular changes in the skin.

In this interview with MEDICA-tradefair.com, Prof. Vasilis Ntziachristos describes the impact the procedure can have for those affected – who could eventually use it by themselves at some point.

Prof. Ntziachristos, how does your method work?

Prof. Vasilis Ntziachristos: RSOM is a new technology we developed for clinical examinations. It uses visible light to excite ultrasound waves within tissue. The ultrasound waves detected are of very broad frequency response (about 100 Megahertz) so that for the first time, we can visualize rich optical contrast with high resolution inside the tissue, much deeper than optical microscopy.

This new ability allows us to recognize features of the tissue structure that were previously not achievable with other non-invasive imaging methods. Therefore, RSOM opens new possibilities for using the skin as a window to many diseases. For example, it is well known that diabetes affects the micro vessels in the skin. We have now shown that by obtaining detailed three-dimensional images of the skin's microvasculature using RSOM, we can non-invasively detect the progression of diabetes.

What is the benefit of such scanners in medicine?

Ntziachristos: Many people visit their doctor once a year or less, often not until symptoms appear. Therefore, current medicine assesses disease in a rather infrequent way. It is also often difficult to understand if there is improvement or progression of a disease in response to changing one’s lifestyle or in response to drugs. Using non-invasive sensors, we can monitor disease more frequently and quantify gradual changes over time. So, we can monitor the evolution of disease in a more precise way than is possible today.

There is also evidence in the literature that microvasculature changes due to diabetes occur very early, typically confirmed by invasive tests. With the new sensor technology, we are therefore not only opening up opportunities for improving the precision of progression monitoring, but also early detection.

How often can this method be used?

Ntziachristos: Patients can take measurements on a daily basis because we work with a non-invasive sensor. For example, one can take a quick measurement in the morning, or once a week. The measurements only take a few seconds. Such frequency should be sufficient to build a consistent baseline and detect even small changes.

We are also working towards a mobile solution with a pocket-size portable sensor. We see that the ability to monitor your health status in a precise and quantitative way can be a key motivating factor for consistently engaging with intervention processes, for example exercise, healthy eating or taking medication.

What are other potential uses?

Ntziachristos: It is possible to use the sensor to detect various other diseases. We now apply it to diagnostics and monitoring of cardiovascular disease and in dermatological conditions, from the detection of skin cancer to inflammatory diseases such as psoriasis.

Another important aspect is that in the digital age, many people are interested in taking control in monitoring their health. They want to see the status of their health themselves and possibly decide when to go to the doctor. With the help of this technology, they can do so.

What role does AI play in the operation of the new sensor?

Ntziachristos: AI is important for a thorough and precise analysis. In this study, we trained AI to automate image analysis tasks. This allows us to reliably extract features from the images that doctors or individuals can understand, such as vessel density and thickness. There are many different ways to apply AI to the analysis of RSOM data and its use will be essential as we move on to more complex measurements and to monitoring more diseases.