Patenting AI for Medical Diagnostics
Artificial intelligence (AI) and machine learning (ML) have revolutionised the field of data analysis. Trained AI and ML models can now outperform humans in many analysis tasks, including those where humans require years of study. One example where AI and ML models have particularly helped is in the field of healthcare and medical diagnosis.
Patients can present with a huge amount of medical data, for example, blood pressure measurements, X-rays and CT scans, histology images and biochemical blood tests data. It can take years of training for a human to assimilate these data in just one of these areas to enable them to provide a diagnosis and recommend treatment for each new case they see. This situation can often result in a bottleneck where results from a patient are not seen by an appropriately trained specialist for weeks.
Al and ML models which are trained to look for specific patterns in the data, for example, a set of CT scan data, can take new CT scan data from a patient and classify it into a likely diagnosis quickly and efficiently. This classification can then be used as an aid for medical professionals when assessing a patient’s clinical needs. From the perspective of European patent law, the application of AI to healthcare is particularly interesting because on the surface it seems to relate to several different types of patent exclusions. Mathematical methods, computer programs and methods of treatment and diagnoses performed on a human body are all excluded from patent protection in European patent law.
Specifically, Article 52(2) of the European Patent Convention (EPC) lists a number of subjects that are not regarded as inventions by the European Patent Office (EPO). Among them are programs for computers and mathematical methods. You may think that AI and ML models fall into both of these categories, being a mathematical model implemented on a computer. However, Article 52(3) of the EPC goes on to state that applications related to these subjects are only excluded “to the extent to which a European patent application or European patent relates to such subject-matter or activities as such”.
What this means in practice is that in order to obtain patent protection for a mathematical model like an AI or ML algorithm or model, the algorithm or model needs to contribute to the technical character of the invention, for example by being applied to a specific ‘technical’ application area. A variety of examples of applications which are considered ‘technical’ are provided in the EPO Guidelines for Examination and this list includes “providing a medical diagnosis by an automated system processing physiological measurements”.
Therefore, AI and ML methods which achieve a diagnosis, like categorising a histology sample or a radiograph as either showing a potential cancer or not, would be considered to be technical by the EPO. Depending on the invention, it is therefore possible to gain patent protection for AI and ML systems which achieve similar outcomes.
For some medical diagnostic methods, there is a second hurdle to patentability that also needs to be overcome – that of methods of treatment and diagnoses performed on a human or animal body.
European patent law has been written to prevent medical physicians and veterinarians from being impeded by patent rights. This is clearly a policy in the public interest as no one wants to reduce the freedom of physicians or veterinarians performing their job. Article 53 of the EPC thus states that patents shall not be granted for “diagnostic methods practised on the human or animal body”.
However it is important to note that this exception to patentability only applies to method claims and so does not exclude protection for a medical device or computer program which carries out a diagnostic method.
Even for method claims, in our experience it is still often possible to overcome the Article 53 EPC exclusion. The EPO Guidelines provide four phases which need to be performed for a method to fall within the Article 53 exclusion. These are: (i) the examination phase, involving the collection of data; (ii) the comparison of these data with standard values; (iii) the finding of any significant deviation; and i.e. a symptom, during the comparison, (iv) the attribution of the deviation to a particular clinical picture, i.e. the deductive medical or veterinary decision phase.
Furthermore, the Article 53 exclusion relates to diagnostic methods “practised on the human or animal body”. To determine this, it must be ascertained whether an interaction with the human or animal body takes place. AI and ML methods performing a technical part of the claim are usually performed on a device which is not interacting with the human or animal body. Even with the advent of wearable devices such as smart watches which are attached to our bodies for long periods, it is arguable that underlying an AI or ML model itself is sufficiently removed from any direct interaction with the body such that the method exclusion does not bite.
In conclusion, armed with knowledge of the issues, the exclusions of Article 52(2) and 53 of the EPC can often be avoided by careful drafting. Of course, the usual criteria of novelty and non-obviousness will still need to be met-but for the right invention, it is certainly possible to gain patent protection for the use of AI in healthcare applications.
Venner Shipley has experts in medical devices and AI and our team is perfectly placed to advise on protection on the issue of AI for medical diagnostics. If you would like any advice in this area, please let us know.