T 1910/20 (Displaying cluster centres/ROCHE) 03-02-2023
Download and more information:
SYSTEMS AND METHODS FOR AUTOMATICALLY DISPLAYING PATTERNS IN BIOLOGICAL MONITORING DATA
Roche Diabetes Care GmbH
F. Hoffmann-La Roche AG
I. The appeal is against the examining division's decision to refuse the application on the grounds that the main request and auxiliary requests 1 to 3 did not meet the requirements of Article 56 EPC with regard to the following document:
D1: WO 2010/075350 A1
II. In their statement setting out the grounds of appeal, the appellants requested that the decision be set aside and that a patent be granted on the basis of one of the requests on which the contested decision was based. They requested oral proceedings as an auxiliary measure.
III. The board summoned the appellants to oral proceedings. In its preliminary opinion pursuant to Article 15(1) RPBA, the board raised objections under Article 56 EPC.
IV. Oral proceedings were held before the board.
V. Claim 1 of the main request reads as follows:
"A collection system for automatically displaying patterns in glucose data, the collection system comprising:
- one or more processors;
- an electronic display communicatively coupled to the one or more processors; and
- machine readable instructions that are executed by the one or more processors, wherein the machine readable instructions cause the one or more processors to execute an algorithm comprising:
- receiving a glucose data signal indicative of ambulatory glucose levels sampled over time, wherein the glucose data signal is associated with a time index indicative of a point in time, a date, or both that the glucose data signal was sampled; and
- dividing the glucose data signal into segments of interest according to the time index, wherein the segments of interest are each a time window of data;
characterized in that said algorithm is a pattern enhancement algorithm for enhancing patterns that exist within the glucose data signal, the pattern enhancement algorithm further comprising:
- extracting, automatically, a set of features according to a mathematical algorithm from each of the segments of interest, thereby, automatically transforming each of the segments of interest into the set of features that is a reduced representation of the segments of interest;
- determining, automatically, a distance metric between each of the segments of interest based at least in part upon the set of features of the segments of interest by applying a function for determining the distance metric to the set of features, wherein the distance metric is indicating a degree of similarity between each of the segments of interest;
- clustering, automatically, the segments of interest into a number of groups of clustered segments based at least in part upon the distance metric, according to a clustering algorithm, thereby, grouping similar segments of interest in each of the clustered segments;
- calculating a cluster center for each of the groups of clustered segments, wherein the cluster center is based upon a mean of one of the groups of clustered segments; and
- presenting, automatically, the cluster centers on the electronic display."
Claim 1 of auxiliary request 1 differs from claim 1 of the main request as follows (with the additions underlined):
"[...]
- extracting, automatically, a set of features according to a mathematical algorithm from each of the segments of interest, thereby, automatically transforming each of the segments of interest into the set of features that is a reduced representation of the segments of interest, wherein the mathematical algorithm comprises a Principal Component Analysis (PCA), a Kernel PCA, a wavelet analysis, a frequency analysis, or a combination thereof; in particular, wherein the clustering algorithm comprises an optimizer;
- determining, automatically, a distance metric between each of the segments of interest based at least in part upon the set of features of the segments of interest by applying a function for determining the distance metric to the set of features, wherein the distance metric is indicating a degree of similarity between each of the segments of interest, wherein the function for determining the distance metric is based upon a sum of squared distance, a sum of absolute distance, a Mahalanobis distance, a Manhattan distance, a maximum norm, or a combination thereof;
- clustering, automatically, the segments of interest into a number of groups of clustered segments based at least in part upon the distance metric, according to a clustering algorithm, thereby, grouping similar segments of interest in each of the clustered segments, wherein the clustering algorithm comprises a K-means algorithm, a Hierarchical clustering algorithm, a Gaussian mixture modeling algorithm, a Normalized Cuts algorithm, or a combination thereof;
[...]"
Claim 1 of auxiliary request 2 differs from claim 1 of auxiliary request 1 as follows (with the deletions [deleted: struck-through]):
"[...]
- extracting, automatically, a set of features according to a mathematical algorithm from each of the segments of interest, thereby, automatically transforming each of the segments of interest into the set of features that is a reduced representation of the segments of interest, wherein the mathematical algorithm comprises a Principal Component Analysis (PCA), a Kernel PCA, a wavelet analysis, a frequency analysis, or a combination thereof; [deleted: in particular, wherein the clustering algorithm comprises an optimizer;]
- determining, automatically, a distance metric between each of the segments of interest based [deleted: at least in part] upon the set of features of the segments of interest by applying a function for determining the distance metric to the set of features, wherein the distance metric is indicating a degree of similarity between each of the segments of interest, wherein the function for determining the distance metric is based upon a sum of squared distance, a sum of absolute distance, a Mahalanobis distance, a Manhattan distance, a maximum norm, or a combination thereof;
- clustering, automatically, the segments of interest into a number of groups of clustered segments based [deleted: at least in part ]upon the distance metric, according to a clustering algorithm, thereby, grouping similar segments of interest in each of the clustered segments, wherein the clustering algorithm comprises a K-means algorithm, a Hierarchical clustering algorithm, a Gaussian mixture modeling algorithm, a Normalized Cuts algorithm, or a combination thereof;
[...]"
Claim 1 of auxiliary request 3 differs from claim 1 of auxiliary request 2 as follows (with the additions underlined):
"[...]
- presenting, automatically, the cluster centers on the electronic display, comprising presenting at least one of the following
- summary statistics calculated for each group of clustered segments including one of the following: the mean, median, standard deviation, mean absolute difference, percentage of time in hyperglycemia, percentage of time within a target range, percentage of time below a threshold, and percentage of time above a specific threshold;
- regions of importance selected from the following: hyperglycemia, hypoglycemia, and glucose target ranges; and
- contextual data associated with the clustered data segments selected from the following: meal tags, carbohydrate intake, and insulin injections."
1. The contested decision held that the features in the characterising portion of claim 1 of each of the requests indicating the details of the pattern enhancement algorithm were distinguishing features with regard to D1. The appellants did not contest the distinguishing features but did dispute the examining division's conclusion that the distinguishing features did not contribute to the technical character of the invention. In particular, the examining division had held that the purpose of the distinguishing features was a presentation of information, which did not credibly assist a user in performing a technical task by means of a continued and guided human-machine interaction process (the requirement in T 336/14 and T 1802/13).
In their statement setting out the grounds of appeal, the appellants argued that the technical effect of the distinguishing features was enhancing patterns in measured glucose data which may otherwise not be recognised or may be overlooked, namely making said patterns visible or recognisable. They argued that in the case in hand the presented information, in particular the cluster centres, contributed to the technical character of the invention because the presented clusters would help an experienced and skilled doctor to better understand the measurement data and thus assist them in building a more appropriate diabetes therapy plan.
The board is not convinced by the appellants' arguments. Building a therapy plan, which should not be confused with practising a method of therapy on the human body, is a purely intellectual exercise, not a technical task. It is analogous to the purely intellectual deductive decision phase in diagnosis (see T 1091/17, point 1.8 of the reasons). For want of any technical task, the argument that the information presented in this case allegedly assists a user in performing a technical task can be dismissed straight away.
2. At the oral proceedings, the appellants argued that the pattern enhancement algorithm as indicated in the distinguishing features was above all a mathematical method which contributed to the technical character of the invention, not only by virtue of the resultant presentation of information but also on account of the derivation of a patient's physiological state. They referred to the Guidelines for Examination in the European Patent Office, G-II, 3.3, which states the following: "If steps of a mathematical method are used to derive or predict the physical state of an existing real object from measurements of physical properties, as in the case of indirect measurements, those steps make a technical contribution regardless of what use is made of the results." The appellants argued that a patient was also an existing real object and their physiological state, in terms of patterns in their blood glucose data, was a physical state. They further referred to T 1785/14 as support for their argument.
The appellants' argument seemed to build on a misleading blanket statement they had cited from the Guidelines. When the board reminded the appellants that the Guidelines are not binding on the boards (see Case Law of the Boards of Appeal, 10th edition, July 2022, III.W.1), the appellants indicated that they had cited this statement as it also represented their own opinion on the contribution of mathematical method steps to the technical character of the invention. For this reason, and also in view of Article 20(2) RPBA, the board deems it necessary to explain why this statement is erroneous. In the preceding sentence of the Guidelines, referring to relevant case law, it is correctly stated that "[d]efining the nature of the data input to a mathematical method does not necessarily imply that the mathematical method contributes to the technical character of the invention". In particular, to contribute to technical character, it is not sufficient that the quantities processed by a mathematical method represent physical parameters. As arguably all physical parameters represent a physical state of an existing real object, it follows that the subsequent sentence in the Guidelines, at least at its level of generality ("those steps make a technical contribution regardless of what use is made of the results"), is wrong.
Moreover, it cannot be derived from T 1785/14 (to which the appellants referred) that defining the input to a mathematical method as physiological parameters would suffice to contribute to the technical character of the invention. The board in that case found that the distinguishing features solved the technical problem of allowing for better accuracy when determining the sampling time for an individual animal (see T 1785/14, point 3.1.2). That is not comparable with the case in hand, in which the appellants have been unable to demonstrate any technical effect of the presentation of cluster centres produced by the mathematical steps of claim 1 of any of the requests.
3. The appellants also referred to two early decisions of the boards of appeal on the possible technical contribution of the presentation of information, namely T 115/85 and T 362/90. These decisions have been discussed in detail in subsequent case law. As summarised in T 336/14 (see point 1.2.3 of the reasons), for any such contribution, it is relevant to determine whether the presented information constitutes an operation state, a condition or an event internal to the underlying technical system. In the case in hand, the presented cluster centres are not an internal state of the underlying technical system, namely the collection system. Instead, they arguably relate to the internal state of a patient, as the appellants themselves asserted (see point 2. above). Therefore, these decisions have no relevance to the case in hand.
4. Consequently, claim 1 does not involve an inventive step (Article 56 EPC) in any of the requests.
For these reasons it is decided that:
The appeal is dismissed.