The EPO grants many patents for inventions involving machine learning. However, a recent EPO Board of Appeal decision (T0161/18) highlights the importance of drafting such applications carefully so that sufficient details of the training dataset are included. A failure to do so can result in refusal of the application on the grounds of both sufficiency and lack of inventive step.

The invention and claims

The invention concerned a device in which aortic heart pressure were calculated from blood pressure measurements using a neural network.

Independent method claim 1 contained the following characterising portion:

" ..... characterized in that the transformation of the blood pressure curve measured at the periphery is converted into the equivalent aortic pressure with the help of an artificial neural network, the weighting values of which are determined by learning."

A corresponding independent device claim was included comprising a measuring device for detecting blood pressure and a computing unit. This claim contained the following similar characterising portion:

" ..... characterized in that the computing unit for transforming the measured blood pressure curve (7) has an artificial neural network (8) whose weighting values were determined by learning."

Sufficiency

The Board of Appeal first considered whether the invention was sufficiently disclosed. "Sufficient disclosure" is a requirement of patent applications which arises from Article 83 of the European Patent Convention (EPC).

Article 83 is a short article, which states very simply:

The European patent application shall disclose the invention in a manner sufficiently clear and complete for it to be carried out by a person skilled in the art.

The board stated that, for this purpose, "the disclosure of the invention in the application must enable the person skilled in the art to reproduce the technical teaching inherent in the claimed invention on the basis of his general specialist knowledge".

With regard to the training of the neural network, the present application stated that the input data should cover a wide range of patients of different ages, genders, constitutional types, health status and the like so that the network did not become specialized.

However, the application did not disclose which input data were suitable for training the artificial neural network according to the invention, and the application did not disclose at least one data set suitable for solving the technical problem of the invention.

The board stated that, "The training of the artificial neural network cannot therefore be reworked by the person skilled in the art, and the person skilled in the art cannot therefore carry out the invention."

The board therefore concluded that the invention was not sufficiently disclosed and thus did not meet the requirements of Article 83 EPC above.

Inventive step

Interestingly, the board also considered the invention to lack an inventive step for the same reason, i.e. the invention was not sufficiently disclosed.

The appellant saw the problem solved by the invention as creating a method and a corresponding device which guaranteed a precise determination of the cardiac output, whereby the computational effort was kept within reasonable limits thus enabling integration into a mobile and appropriately handy device.

The board was not convinced that the artificial neural network according to claim 1 provided the advantages asserted by the applicant, since neither the claim nor the description contained details regarding the training of the artificial neural network.

In particular, the board stated that, "The mere indication that weight values are determined by learning does not go beyond what the expert understands by an artificial neural network."

The board therefore considered that the claimed neural network was not adapted for the specific, claimed application. The board stated that there was therefore, "only an unspecified adaptation of the weight values, which is the nature of every artificial neural network."

Conclusion

Patent applications in the field of machine learning must be drafted carefully to ensure that sufficient details of the training dataset are included, and must clearly disclose which input data are suitable for training the artificial neural network.