What was once only seen in science fiction movies, is now becoming reality. A surge in the applicability of artificial intelligence (AI) to mainstream life has occurred in the past 30 years, particularly in medicine. Techniques including artificial neural networks, fuzzy expert systems, evolutionary computation, and hybrid intelligent systems are being applied more than ever before to assist the medical community in patient diagnosis, with a similar or better success rate to their human counterparts.

The speed of innovation in the field of AI diagnostics is rapid. One concern for innovators is – can patent law keep up? There are patent eligibility challenges aplenty in major innovation destinations for both computer-based technologies and diagnostics. This article reviews examples of AI diagnostics, recent changes to patent subject matter eligibility and best practice in navigating these changes to protect medical AI in Australia and USA.

What is Medical AI?

According to the English Oxford Living Dictionary AI is “the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.”

The World Intellectual Property Organisation (WIPO) recently reported AI as one of the ‘Technology Trends of 2019’[1], providing detailed data on the rise of AI-related patents and key players seeking AI-related patent protection.

Of the top 20 fields of application identified in the analysis, life and medical sciences ranked only behind telecommunications (15%), with 12% of filed AI patents containing related subject matter.

It is no surprise that large multinationals like Siemens, Philips and Samsung hold the largest patent portfolios in the life and medical sciences. Universities and public research organisations (such as CAS, University of California and Zhejiang University) are also represented among the top 20 players, particularly in the field of neurosciences and neurorobotics.

Late last year, Siemens announced the release of AI-Rad Companion Chest CT just one of their latest AI diagnostics in medical imaging. The product assists radiologists in the interpretation of chest images. While typically, radiological images display a wide variety of information, due to time constraints and skill shortfalls, a radiologist’s analysis is often limited to the primary medical indication. The software algorithm of AI-Rad Companion Chest CT can differentiate between various structures in the chest region including lungs, heart, aorta, and coronary arteries, identifying potential abnormalities outside the primary concern including lung lesions and total calcium volume in the coronary arteries. AI-Rad Companion then automatically generates a standardised, reproducible, and quantitative report based on these findings.

The technology comes with the promise not to replace radiologists of the future, but rather to enhance their work. The same potential exists for most AI technology: the benefit of intelligent automation is an increase in efficiency, accuracy and productivity as well as a reduction in potential human error.

Is patent law moving at the same pace as medical AI innovation?


For many years, certain subject matter has been patent ineligible in Australia, including laws of nature, abstract ideas and mathematical algorithms. Recent seminal cases like Myriad[2], Research Affiliates[3] and RPL Central[4] have added isolated naturally occurring nucleic acid molecules, cDNA, RNA and the implementation of a method by a computer or the internet to the list of ineligible subject matter.

It is interesting to note that claims directed to diagnostic methods have not directly been considered by an Australian court just yet, however judges in the Myriad case made brief comments on the patentability of Myriad’s methods of diagnosis, opining:

“[i]t is not disputed that a process or method of detecting the increased likelihood of certain kinds of malignancy…..may be patentable subject matter as a process.”[5]

Whilst diagnostics per se are safe from the patentability chopping block, where does this leave diagnostics deriving the benefit of ‘computer-implemented invention’ like AI technology?

Algorithms themselves and their integration with a computer are not necessarily off limits. Recent evidence suggests the most likely successful way forward is the careful drafting of a patent specification with an emphasis on detailed technical description of the applicability of AI to the diagnostic outcome, and how the use of AI is transformative rather than simply a substitute for extended human endeavour.


Turning to the US, the situation becomes a bit more complex. The US Supreme Court has invalidated patents directed to diagnostics and computer-implemented inventions in the following seminal cases: Mayo[6], Alice[7] and in the US equivalent, Myriad[8].

In Mayo the plaintiff – specialty pharmaceutical and diagnostics company Prometheus Laboratories – had patented a suite of tests for assessing the correct dosages of certain Crohn’s disease medications. When Mayo Clinic developed its own similar tests and began using them, Prometheus sued Mayo for infringement. The US Supreme Court invalidated Prometheus’ method patent and provided a framework for determining patent eligibility in two steps. In 2014, the US Supreme Court went on to clarify Mayo in Alice which resulted in the so-called Alice/Mayo test.

As part of the two-step test, it is first determined whether the claims are a law of nature, natural phenomenon or an abstract idea (known as ‘the judicial exceptions’). What constitutes an ‘abstract idea’ is relevant when considering computer-implemented inventions based, for example, on AI because US courts have invalidated patent claims covering subject matter that could be performed through an ‘ordinary mental process,’ ‘in the human mind’ or by ‘a human using a pen and paper’. When one considers the development of AI technology, particularly medical AI, this is in fact what the invention is – the supplementation of the human mind with intelligent software.

Cases including Ex Parte Kirshenbaum[9] can be informative of approaches to achieving patent protection in AI based technologies. Specifically, the US Patent Appeal Board stated that eligible inventions are those that have a feature or limitation that produces a:

“useful, concrete and tangible result without pre-empting an abstract idea like a mathematical algorithm”

Further, patent eligible inventions that can be described in terms of structure (“what it is”) rather than functionally (“what it does”) will have a better chance of allowance.

Since the Alice decision, the USPTO has issued several iterations of practical guidance for evaluating subject matter eligibility. Throughout this period, concerns have been expressed that different examiners within and between technology centres have applied inconsistent standards due to a lack of clarity in how the ‘judicial exception’ to subject matter eligibility should be applied. This has created a challenging environment for inventors and companies trying to reliably predict the outcomes of patent examination. Public demand for clarity and consistency has led to the latest ‘Revised Patent Subject Matter Eligibility Guidance[10]’, released by USPTO Director Andrei Iancu in January this year.

The 2019 Guidance is just that – a set of guidelines with no changes made to the law. However, an optimistic view suggests that the recent changes will do what they have been established to do – provide consistency and clarity to what was essentially a roll of the dice in the approach a US examiner might take in claim interpretation, especially in the area of AI diagnostics. Early indications are that the Guidelines have constructively softened the approach to patenting of computer implemented inventions, and that many will now pass the first of the two step Alice/Mayo test.


While patent law is perhaps not moving at the same pace as the growth of AI techniques in diagnostic testing, the signs at least in the USA are positive for the patentability of this subject matter. Since the US courts at least in the past have led the world in advancing the technology cover afforded by patent law, and since Australian Courts have tended to follow their US brethren, although we might be waiting a while, Australian innovators in the field of artificially intelligent diagnostics can have optimism.

Kindly published in AusBiotech’s Australasian Biotechnology Journal April 2019