A Tale Of Two Jurisdictions: Sufficiency Of Disclosure For Artificial Intelligence (AI) Patents In The US And The EPO – Intellectual Property – United…

Posted: November 5, 2021 at 10:00 pm

PatentNext Summary: In order to prepareapplications for filing in multiple jurisdictions, practitionersshould be cognizant of claiming styles in the various jurisdictionsthat they expect to file AI-related patent applications in, anddraft claims accordingly. For example, different jurisdictions,such as the U.S. and EPO, have different legal tests that canresult in different styles for claiming artificialintelligence(AI)-related inventions.

In this article, we will compare two applications, one in theU.S. and the other in the EPO, that have the same or similarclaims. Both applications claim priority to the same PCTApplication (PCT/AT2006/000457) (the "'427 PCTApplication"), which is published as PCT Pub. No.WO/2007/053868.

As we shall see, despite the application having the same orsimilar claims, prosecution of the applications in the twojurisdictions nonetheless resulted in different outcomes, with theU.S. application prosecuted to allowance and the EPO applicationending in rejection.

****

Pertinent to our discussion is an overview of AI. A briefdescription of AI follows before analysis of the AI-related claimsat issue.

Artificial Intelligence (AI) is fundamentally a data-driventechnology that takes unique datasets as input to train AI computermodels. Once trained, an AI computer model may take new data asinput to predict, classify, or otherwise output results for use ina variety of applications.

Machine learning, arguably the most widely used AI technique,may be described as a process that uses data and algorithms totrain (or teach) computer models, which usually involves thetraining of weights of the model. Training typically involvescalculating and updating mathematical weights (i.e., numeralvalues) of a model based on input that can comprise hundreds,thousands, millions, etc. sets of data. The trained model allowsthe computer to make decisions without the need for explicit orrule-based programming.

In particular, machine learning algorithms build a model ontraining data to identify and extract patterns from the data andtherefore acquire (or learn) unique knowledge that can be appliedto new data sets.

For more information, see Artificial Intelligence & the IntellectualProperty Landscape

AI inventions are fundamentally software-related inventions. Inthe U.S., as a practical rule, software-related patents shoulddisclose an algorithm by which the software-related invention isachieved. An algorithm provides support for a software-relatedpatent pursuant to 35 U.S.C. 112(a) including (1) byproviding sufficiency of disclosure for the patent's"written description" and (2) by "enabling" oneof ordinary skill in the art (e.g., a computer engineer or computerprogrammer) to make or use the related software-related inventionwithout "undue experimentation." Without such support, apatent claim can be held invalid. For more information regardinggeneral aspects of the sufficiency of disclosure in the U.S. forsoftware-related inventions, see Why including an "Algorithm" isImportant for Software Patents (Part 2)

U.S. Patent 8,920,327 (the "'327 Patent") issuedfrom the '457 PCT Application. The ''327 Patent is anexample of an AI patent that did not experiencesufficiency issues in the U.S. The below provides an overview ofthe '327 Patent.

The '327 Patent is titled "Method for DeterminingCardiac Output" and includes a single independent claimregarding a method for cardiac output from an arterial bloodpressure curve. The method is implemented via a cardiac device, asillustrated in Figure 1 (copied below):

Fig. 1 illustrates device 1 for implementing the invention ofthe '327 patent, where measuring device 2 measures theperipheral blood pressure curve, and where related measurement datais fed into device 1 via line 3, and stored and evaluated there.The device further comprises optical display device 4, input panel5, and keys 6 for inputting and displaying information.

The claimed method includes an AI aspect, i.e., namely the useof "an artificial neural network having weightingvalues that are determined by learning."

Claim 1 is copied below (with the AI aspectbolded):

1. A method for determiningcardiac output from an arterial blood pressure curve measured at aperipheral region, comprising the steps of:

measuring the arterial bloodpressure curve at the peripheral region; arithmeticallytransforming the measured arterial blood pressure curve to anequivalent aortic pressure; and

calculating the cardiac outputfrom the equivalent aortic pressure,

wherein the arithmetictransformation of the arterial blood pressure curve measured at theperipheral region into the equivalent aortic pressure is performedby the aid of an artificial neural networkhaving weighting values that are determined bylearning.

Figure 3 of the '327 patent (copied below) is a schematicillustration of the artificial neural network, as recited in claim1.

The specification of the '327 patent describes that"FIG. 3 illustrates the structure of the neural network...,and it is apparent that the neural network ... is comprised ofthree layers 14, 15, 16." The specification discloses that asupervised learning algorithm is used to train the weights of themodel, e.g., "[t]he weights and the bias for the latter twolayers 15 and 16 are determined by supervised learning."

The input data fed to the supervised learning algorithm to trainthe AI model includes "associated blood pressure curve pairsactually determined by measurements in the periphery or in theaorta, respectively, are used." The measurements used for theinput data may come "from patients of different ages, sexes,constitutional types, health conditions and the like."

No issues with respect to sufficiency were raised during theprosecution of the application in the U.S. that was issued as the'327 patent.

More generally, issues of sufficiency in the U.S. typicallyarise in litigation, and result in expert testimony, i.e., "abattle of the experts," where expert witnesses (e.g.,typically university professors or industry consultants) fromopposing sides opine on the knowledge of a person of ordinary skillin the art and sufficiency of disclosure in view of thatperson.

The EPO has developed its own, yet similar, stance on AI-relatedinvention when compared with the U.S. Nonetheless, outcomes ofprosecution can be different. The below provides a cursory overviewof developments in the EPO with respect to AI-related inventionsand analyzes the treatment of an EPO application as filed based onthe PCT Application '457 (which is the same PCT Application asfor the '327 patent discussed above).

Generally, artificial intelligence inventions may be patented inthe European Patent Office (EPO). For example, in its Guidelinesfor Examination, the EPO defines AI and machine learning as"based on computational models and algorithms forclassification, clustering, regression and dimensionalityreduction, such as neural networks, genetic algorithms, supportvector machines, k-means, kernel regression and discriminantanalysis." Section 3.3.1 (Artificial intelligence and machinelearning).

As such, the EPO dubs AI and machine learning as "per se ofan abstract mathematical nature," irrespective of whether suchmodels may be trained with training data. Id. Thus, simplyclaiming a machine learning model (e.g., such as a "neuralnetwork") does not, alone, necessarily imply the use of a"technical means" in accordance with EPO law.

Nonetheless, the Guidelines for Examination at the EPO recognizethat the use of an AI model, when claimed as a whole with theadditional subject matter, may demonstrate a sufficient technicalcharacter. Id. As an example, the Guidelines forExamination at the EPO states that "the use of a neuralnetwork in a heart-monitoring apparatus for the purpose ofidentifying irregular heartbeats makes a technicalcontribution." Id. As a further example, the EPOGuidelines for Examination further states that "[t]heclassification of digital images, videos, audio or speech signalsbased on low-level features (e.g. edges or pixel attributes forimages) are further typical technical applications ofclassification algorithms." Id.

In a decision in 2020, the EPO Board of Appeals rejected amachine learning-based patent application that claimed an"artificial neural network" because the patentspecification failed to sufficiently disclose how the artificialneural network was trained. See T0161/18 (Equivalent aortic pressure / ARCSEIBERSDORF). The application in question claimed priority to thePCT Application '457, which is the same parent application asthe '327 patent, as discussed above.

The claims were the same or similar as to those in the U.S.,where the claims-at-issue directed to determining cardiac outputfrom an arterial blood pressure curve measured at a periphery, andrecited, in part (with respect to AI), that the "bloodpressure curve measured on the periphery is converted into theequivalent aortic pressure with the help of anartificial neural network, the weighting values ??ofwhich are determined bylearning."

Claim 1 is reproduced below (in English based on a machinetranslation of the original opinion German):

1. A method for determining thecardiac output from an arterial blood pressure curve measured atthe periphery, in which the blood pressure curve measured at theperiphery is mathematically transformed to the equivalent aorticpressure and the cardiac output is calculated from the equivalentaortic pressure, characterized in that the transformation of theblood pressure curve measured on the periphery is converted intothe equivalent aortic pressure with the help of anartificial neural network, the weighting values ??ofwhich are determined by learning.

The Board analyzed the claim in view of the specificationpursuant to Article 83 EP (Sufficient disclosure). As described bythe Board, Article 83 EPC requires that the invention be disclosedin the European patent application so clearly and completely that aperson skilled in the art can carry it out. For this, thedisclosure of the invention in the application must enable theperson skilled in the art to reproduce the technical teachinginherent in the claimed invention on the basis of his generalspecialist knowledge.

The Board then turned to the specification to determine whetherit disclosed enough support to meet these requirements in view ofthe claimed "artificial neural network." However, thespecification was found lacking because it failed to"disclose which input data aresuitable for training the artificial neural network according tothe invention, or at least one data set suitable for solving thetechnical problem at hand."

Instead, the Board found that the specification "merelyreveals that the input data should cover a broad spectrum ofpatients of different ages, genders, constitution types, healthstatus and the like."

Therefore, the Board found that the training of the artificialneural network could therefore not be reworked by the personskilled in the art, and the person skilled in the art can thereforenot carry out the invention.

Because of these deficiencies, the Board found that thespecification failed to provide sufficient disclosure pursuant toArticle 83 EPC.

For similar reasons, the Board further found that the claimedsubject matter lacked an "inventive step" pursuant toArticle 56 EPC. Specifically, the Board found that the claimed"artificial neural network" was not adapted for thespecific, claimed application because the specification failed todisclose how the artificial neural network was trained, andspecifically failed to disclose weight values that resulted fromsuch training. For this reason, the claimed "artificial neuralnetwork" could not be distinguished from the cited prior art,which resulted in failure to demonstrate the requisite inventivestep.

As the Board described:

In the present case, the claimedneural network is therefore not adapted for the specific, claimedapplication. In the opinion of the Chamber, there is therefore onlyan unspecified adaptation of the weight values, which is in thenature of every artificial neural network. The board is thereforenot convinced that the claimed effect will be achieved in theclaimed method over the entire range claimed. This effect cannot,therefore, be taken into account in the assessment of inventivestep in the sense of an improvement over the prior art.

Accordingly, at least with respect to patent applications filedin the EPO, and where an AI or machine learning model is to bedistinguished from the prior art, then a patent applicant may wantto include an example training data set, example trained weights,or at least sufficiently describe the input used to train the modelon a specific, claimed application or end-use. For example, atleast one example of data can be provided (or claimed) to show theinputs used to train specific weights, which may allow for theclaim to have sufficient disclosure, and, at the same time allowthe claim to cover a spectrum of AI models trained with aparticular set of data.

For the time being, such disclosure for an EPO case could beconsidered as additional when compared with the sufficiency ofdisclosure in the U.S. However, it is to be understood that theU.S. Patent office has also indicated the importance of includingtraining data or specific species of data used to train a model inits example guidance. See How to Patent an Artificial Intelligence (AI)Invention: Guidance from the U.S. Patent Office (USPTO). In anyevent, while there have been few court cases on AI-relatedinventions in the U.S. (see How the Courts treat Artificial Intelligence (AI)Patent Inventions: Through the Years since Alice), future casesmay indicate whether the U.S. will trend towards the EPO'sdecision in T0161/18 with respect to the sufficiency ofdisclosure.

The content of this article is intended to provide a generalguide to the subject matter. Specialist advice should be soughtabout your specific circumstances.

See original here:

A Tale Of Two Jurisdictions: Sufficiency Of Disclosure For Artificial Intelligence (AI) Patents In The US And The EPO - Intellectual Property - United...

Related Posts