Analysis of health-related data collected and measured by digital devices is becoming increasingly prevalent and desirable in preventative medicine. The ubiquity of smartphones and wearable devices has seen the generation of large volumes of health-related data on a continuous basis from across large populations.

There has been an explosion in analyzing these health-related data in hitherto unexplored ways in order to enhance the diagnosis, treatment and prevention of disease and other medical conditions. Innovation in this field has been happening on an unprecedented scale[1].

One key objective is to identify so-called digital biomarkers in the health-related data – these are flags in a data set for a given individual correlating to particular health conditions. Digital biomarkers can thus serve as digital “fingerprints” of possible diseases and other conditions, which might have previously gone undetected. Once identified for an individual, digital biomarkers can then be analysed to perform a medical diagnosis earlier, or more accurately than existing diagnostic techniques. The digital biomarkers can also be used to determine or enhance the form of medical treatment given.

With the large number of data sources now available, tech and healthcare companies have been turning to artificial intelligence (AI) and machine learning technologies to identify and analyse digital biomarkers from the vast pools of available data.

The attractions of obtaining patent protection in this fledgling, but fast-growing, field are self-evident – if a previously unknown digital biomarker for a known disease can be protected, then exclusivity conferred via patent protection can be invaluable. Often, however, the correlation of a particular biomarker to a given disease or medical condition may already be known and therefore the biomarker per se cannot be protected directly. Valuable patent protection, however, can often still be obtained for the innovation which relates to the identification of digital biomarkers from the large volumes of health-related data acquired.

Accordingly, there are three main areas of focus for patent protection:

  1. Protection of digital biomarker(s) per se and their identified novel correlation(s) to a given medical condition.
  2. Protection of novel techniques and data processing used to identify digital biomarker(s) and their correlation(s) to particular medical conditions from a large volume of health-related data obtained from across a population.
  3. Protection of the techniques and data processing used to identify particular digital biomarker(s) and associated medical conditions for a given individual based on their own health-related data sets.

As mentioned, it can be difficult to protect digital biomarker(s) per se if their correlation to particular disease(s) is already known, even at a high level of generality, based on existing medical knowledge. If this is the case, focusing on areas 2 and 3 above can still provide valuable patent protection. It is in these two areas where AI and machine learning are now frequently deployed.

When focusing on areas 2 and 3, it is worthwhile noting that medical diagnostic methods practised on the human body are excluded from patent protection in Europe[2]. However, this exclusion has been interpreted narrowly by the EPO’s Enlarged Board of Appeal[3] to relate to diagnostic methods which expressly include in vivo diagnostic steps. Ex vivo diagnostic steps, and thus automated data processing steps based on data already acquired are generally not excluded under Article 53(c) EPC.

Protecting AI Inventions for Digital Biomarkers

In our previous articles focused on protecting AI and machine learning, we identified a number of AI-related areas which could be the focus for obtaining patent protection in Europe. These areas are also applicable to the field of digital biomarkers as follows.

If the technology used for identifying digital biomarkers is based on new configurations of AI processes and methods (e.g. when implemented in software), then these could be patentable on the basis that a technical effect of identifying the digital biomarker would provide for enhanced treatment and diagnosis of a range of medical conditions. A defined range of likely medical conditions which could be diagnosed via the new forms of AI may well have to be set out in the patent application, and data should ideally be provided demonstrating the effectiveness of the AI in identifying the digital biomarkers. It may not be necessary to expressly claim the diagnostic steps per se in order to demonstrate technical contribution and avoid technical character objections, since the Enlarged Board of Appeal has confirmed that implied technical effects may suffice for technical contribution[4].

If the AI technology being used is well known, even if it is not known for identifying digital biomarkers, then it may be difficult to obtain patent protection at a general level for the use of the known AI technology in identifying digital biomarkers; this would likely be viewed as an obvious application of known AI technology by the skilled AI programmer. However, it might be that the type of health-related data acquired and used to identify digital biomarkers (with the known AI) is itself not known in the art, and thus claims focussed on particular types of health-related input data, or the particular form of training data utilised may be allowable. Again, the technical use for a defined range of likely medical conditions should be set out in the patent application, but there may be no need to disclose the computational models used in detail on the basis that they would be known to the skilled AI programmer.

Products using Digital Biomarkers in Treatment

Alongside the above areas of focus for specific AI-based solutions deployed in identifying digital biomarkers and associated medical conditions, it may also be the case that valuable patent protection can be obtained for medicinal products using AI techniques in treating medical conditions relating to the identified biomarkers, for example products where a particular dosing regimen can be defined by the presence (or lack) of one or more particular digital biomarkers identified by a given AI

technique. Moreover, products themselves identified or defined by the AI technique and the corresponding digital biomarkers identified could be directly claimed. For claims focused on these types of medicinal product, careful consideration needs to be given to the exact form of claim language used.


The considerations highlighted above for AI inventions involving digital biomarkers represent a number of areas of focus by patent practitioners involved with digital biomarkers. The exact form of claim language deployed when drafting patent applications and addressing objections at the EPO will depend on the specifics of the technology involved and commercial objectives. Our team of AI patent specialists has many years’ experience in working with healthcare companies to develop and deploy commercially relevant patenting strategies in this fast-growing area.