WO2021209436A1

WO2021209436A1 - Drug prediction using a machine learning model

Info

Publication number: WO2021209436A1
Application number: PCT/EP2021/059535
Authority: WO
Inventors: David L. Duke; Tony Huschto; Christian Ringemann
Original assignee: F. Hoffmann-La Roche Ag; Roche Diabetes Care Gmbh; Roche Diabetes Care, Inc.
Priority date: 2020-04-16
Filing date: 2021-04-13
Publication date: 2021-10-21

Abstract

The present invention relates to a method for generating a software-implemented module configured to determine a drug dose, to a method for training a machine learning model to generate a software- implemented module and to a method for determining a drug dose. The present invention furthermore refers to a system for generating a software-implemented module, to a system for training a machine learning model, and to a system for determining a drug dose. Additionally, the present invention relates to a computer program product.

Description

Drug prediction using a machine learning model

Description

Technical Field

The present invention relates to a method for generating a software-implemented module configured to determine a drug dose, to a method for training a machine learning model to generate a software- implemented module and to a method for determining a drug dose. The present invention, furthermore, refers to a system for generating a software -implemented module, to a system for training a machine learning model, and to a system for determining a drug dose. Additionally, the present invention relates to a computer program product.

Background Art

Machine Learning (ML) is a branch of computer science that is used to derive algorithms driven by data. Instead of using explicit formulas, ML algorithms employ real-world training data to generate models that are more accurate and sophisticated than models traditionally conceived by humans.

US 7,651,845 discloses a method for insulin prediction based on a postprandial glucose pattern recognition.

US 10,391,242 discloses a bolus calculator, which calculates an insulin bolus based on user entered event data and a recommendation of adjustment of several values is given.

The methods disclosed in the prior art are not able to incorporate additional information into a drug dose calculation. Furthermore, the methods disclosed in the prior art are too complex and/or not accurate enough.

Problem to be solved

It is therefore an objective of the present invention to provide improved systems, computer program products and methods for determining a drug dose avoiding at least in part the drawbacks of the prior art, in particular with regard to accuracy of determination of the drug dose.

Summary of the Invention This problem is solved by the subject matter provided in the independent claims and throughout the specification and the drawings. Moreover, the problem is solved by the method for generating a software-implemented module configured to determine a drug dose, the method for training a machine learning model to generate a software -implemented module, and by a method for determining a drug dose. Furthermore, for solving the problem a system for generating a software -implemented module, a system for training a machine learning model, and a system for determining a drug dose are provided. Additionally, a computer program product is provided. Preferred embodiments of the invention which may be realized in an isolated way or in any arbitrary combination are disclosed in the dependent claims and throughout the specification.

As used in the following, the terms “have”, “comprise”, or “include” or any arbitrary grammatical variations thereof are used in a non-exclusive way. Thus, these terms may both refer to a situation in which, besides the feature introduced by these terms, no further features are present in the entity described in this context and to a situation in which one or more further features are present. As an example, the expressions “A has B”, “A comprises B” and “A includes B” may both refer to a situation in which, besides B, no other element is present in A (i. e. a situation in which A solely and exclusively consists of B) and to a situation in which, besides B, one or more further elements are present in entity A, such as element C, elements C and D or even further elements.

Further, it should be noted that the terms “at least one”, “one or more” or similar expressions indicating that a feature or element may be present once or more than once typically will be used only once when introducing the respective feature or element. In the following, in most cases, when referring to the respective feature or element, the expressions “at least one” or “one or more” will not be repeated, notwithstanding the fact that the respective feature or element may be present once or more than once.

Further, as used in the following, the terms “preferably”, “more preferably”, “particularly”, “more particularly”, “specifically”, “more specifically” or similar terms are used in conjunction with optional features, without restricting alternative possibilities. Thus, features introduced by these terms are optional features and are not intended to restrict the scope of the claims in any way. The invention may, as the skilled person will recognize, be performed by using alternative features. Similarly, features introduced by “in an embodiment of the invention” or similar expressions are intended to be optional features, without any restrictions regarding alternative embodiments of the invention, without any restrictions regarding the scope of the invention and without any restrictions regarding the possibility of combining the features introduced in such way with the optional or nonfunctional features of the invention. According to a first aspect, a method for generating a software-implemented module configured to determine a drug dose is provided, the method comprising, in one or more data processing devices the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time, b) determining if the second analyte value is within a pre-selected range, c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range, d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model, wherein the software-implemented module is configured to, when loaded into a data processing device having one or more processors, determine a drug dose from analyzing a second data set which comprises a third analyte value which was received at a third time and at least one second context data value.

In a second aspect, a method for training a machine learning model to generate a software-implemented module is provided, the method comprising in one or more data processing devices the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time, b) determining if the second analyte value is within a pre-selected range, c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range, d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model.

In a third aspect, the present invention provides a method for determining a drug dose, the method comprising in one or more data processing devices the steps: al) providing a second data set, the second data set comprising a third analyte value which was received at a third time, at least one second context data value, bl) providing a software-implemented module comprising a prediction algorithm representing a machine learning model generated in a machine learning process, cl) determining the drug dose comprising analyzing the second data set by the prediction algorithm, wherein generating the software-implemented module in a machine learning process comprises the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time, b) determining if the second analyte value is within a pre-selected range, c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range, d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model.

In another aspect of the present invention, a system for generating a software-implemented module configured to determine a drug dose is provided, the system comprising one or more data processing devices, the one or more data processing devices configured to:

A) provide at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time,

B) determine if the second analyte value is within a pre-selected range,

C) generate a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range,

D) generate the software -implemented module comprising a prediction algorithm representing the machine learning model, wherein the software-implemented module is configured to when loaded into a data processing device having one or more processors, determine a drug dose from analyzing a second data set which comprises a third analyte value which was received at a third time and at least one second context data value. In still a further aspect, the present invention provides a system for training a machine learning model to generate a software-implemented module, the system comprising one or more data processing devices, the one or more data processing devices configured to:

B) determine if the second analyte value is within a pre-selected range,

D) generate the software -implemented module comprising a prediction algorithm representing the machine learning model.

In a further aspect, the present invention provides a system for determining a drug dose, the system comprising one or more data processing devices, the one or more data processing devices configured to:

Al) provide a second data set, the second data set comprising a third analyte value which was received at a third time, at least one second context data value,

Bl) provide a software -implemented module comprising a prediction algorithm representing a machine learning model generated in a machine learning process,

Cl) determine the drug dose comprising analyzing the present data set by the prediction algorithm, wherein generating the software -implemented module in a machine learning process comprises

A) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time,

B) determining if the second analyte value is within a pre-selected range,

C) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range,

D) generating the software-implemented module comprising a prediction algorithm representing the machine learning model. Further, a computer program product, comprising program code configured to, when loaded into a computer having one or more processors, perform at least one of the inventive methods is provided.

The software-implemented module generated in the method according to the present invention allows a more accurate determination of the drug dose of a user. Further, the inventive method allows for incorporation of additional information, such as activity data or the health status of a user, to the determination of the drug dose.

In step a) of the inventive method, at least one first data set is provided. The at least one first data set can be provided by any known method. The first data set may, for example, be provided from a remote device, from a cloud device or from the one or more data processing devices. The remote device may be a controller device including a personal computer, a tablet computer, or a phone, the remote device preferably is able to communicate, preferably wirelessly, with another device such as a medical device including a medication delivery pump, a continuous glucose monitoring device and/or a blood glucose monitoring device. As the case may be, the medication delivery pump may be a durable pump or a body worn patch pump or a drug delivery pen.

All values comprised in the first data set may be linked to the same user.

The term “linked to the same user” within the context of the present invention means that all values are received from one user. They may be user specific. This means that the values comprised in the first data set may be different for different users.

The first data set may be pre-processed in a step aa), which can be carried out before or after step a). Step aa) is in particular carried out before step b). Methods for pre-processing the first data set are known as such. Pre-processing comprises for example outlier detection and/or outlier elimination.

Thus, in an embodiment the following step aa) is carried out before or after step a) and before step b): aa) pre-processing the first data set to obtain a pre-processed first data set.

If step aa) is carried out, then in the steps following step aa) the pre-processed first data set is used as first data set.

The first data set may be a first training data set for training the machine learning model in the machine learning process of step c). Thus, preferably the first data set provided in step a) is a first training data set.

The first data set comprises a first analyte value, which was received at a first time, at least one first context data value and a second analyte value, which was received at a second time.

The first analyte value may be any value of an analyte, such as a concentration, or an absolute amount. In an embodiment, the first analyte value is a first analyte concentration.

As further used herein, the term "analyte" refers to an arbitrary element, component or compound which may be present in a body fluid and the concentration of which may be of interest for a user. Preferably, the analyte may be or may comprise an arbitrary chemical substance or chemical compound which may take part in the metabolism of the user, such as at least one metabolite. As an example, the analyte may be selected from the group consisting of glucose, cholesterol, triglycerides, lactate. Preferably, the analyte is glucose. The analyte may be detected and the analyte concentration may be determined in an isolated sample obtained from the user or in vivo using any of the generally known in vitro or in vivo analyte sensor measurement systems.

Thus, preferably the first analyte value is a first glucose concentration.

The first analyte value may be received by any method. For example, the first analyte value may be received by a continuous glucose sensor, by a flash glucose sensor or by a blood glucose meter.

The first analyte value was received at a first time. The first time may be any time of the day. Preferably, the first time is a point in time immediately before or immediately after at least one of the at least one first context data value is obtained.

“Immediately before or immediately after at least one of the at least one first context data value is obtained” within the context of the first context data value means immediately before or immediately after the first context data value is created. This is not, on necessity, the time at which the first context data value is received in the first data set. This means that if the first context data value is, for example a first insulin bolus, then “immediately before or immediately after at least one of the at least one first context data value is obtained” means immediately before or immediately after the first insulin bolus is given to the user. If the at least one first context data value is for example an activity index, then “immediately before or immediately after at least one of the at least one first context data value is obtained” means immediately before the activity is started or immediately after the activity is finished. For example, if one of the at least one first context data values is a first medication dose, such as a first insulin bolus, or a first carbohydrate intake, such as a meal, then the first time is preferably a point in time immediately before or immediately after the a first medication dose is given to the user or the first carbohydrate intake is taken by the user. In an embodiment, the first analyte value is a first glucose value which is measured immediately before a first insulin bolus is given to the user.

“Immediately before” within the context of the present invention means at most 15 minutes before, preferably at most 10 minutes before and most preferably at most 5 minutes before. For example, “immediately before” means at a point in time in the range of 5 minutes to 0.5 minutes before. “Immediately after” within the context of the present invention means at most 15 minutes after, preferably at most 10 minutes after and most preferably at most 5 minutes after. For example, “immediately after” means at a point in time in the range of 0.5 minutes to 5 minutes after.

As generally used the term “first context data value” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art and is not to be limited to a special or customized meaning. A first context data value may be a value relating to a physiological state of a user or a value being able to influence a physiological state of a user.

A “value” within the context of the present invention does not only mean one precise numerical quantity but also covers, for example a range of numbers or a sequence of values or numbers, such as, for example, in a glucose pattern.

Values relating to a physiological state of a user are, for example, the body temperature, an analyte pattern, an illness, a stress level and/or a hormonal status.

The analyte pattern is preferably a pattern of the same analyte as the analyte of the first analyte value. Therefore, if, for example, the first analyte is glucose, then the analyte pattern is preferably a glucose pattern, such as a hypoglycemia or a hyperglycemia.

A hormonal status is for example the status of the female menstrual cycle.

Values being able to influence a physiological state of a user are, for example activity indices, a first medication dose and/or a first carbohydrate intake.

As generally used the term “medication” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art. A medication is typically used to diagnose, cure, treat, or prevent disease. Thus, the medication may for example be selected from the group consisting of antibiotics, chemotherapy drugs, pain relievers, insulin and glucagon, preferably selected from insulin and glucagon.

As generally used the term “dose” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art. Thus, a dose is a quantity of a medication or a drug. A dose may be a quantity of the medication or the drug to be administered to a user. The dose may be a quantity to be administered at one time, such as an insulin bolus, or may be a quantity to be administered over a time range, such as an insulin basal rate.

A first medication dose may be any first medication dose, such as a first insulin dose (for example an insulin bolus) or a first glucagon dose.

An activity index may be an index of any activity of a user, such as an athletic activity. The value of the activity index may, for example be a dichotomous value. A dichotomous value may only state if there is/was activity or not. It is also possible that the activity index is more detailed, giving, for example, information on the intensity of the activity.

The first context data value may in particular be selected from at least one of the group consisting of a first medication dose, a first carbohydrate intake, an activity index, and an analyte pattern.

The first context data value may be received in a time range between 10 hours before and 10 hours after the first time at which the first analyte value was received, for example. In an embodiment, the first context data value is received in a time range of from 4 hours before the first time at which the first analyte value was received and 30 minutes after the first time at which the first analyte value was received. It is also possible that the first context data values are received during different time ranges. For example, it is possible that at least one of the at least one first context data values is received in a time range of 0.5 minutes to 15 minutes after the first time (immediately after the first time), whereas another of the at least one first context data values is received in a time range between 0.5 hours and 2 hours before the first time. For example, an activity index may be received in a time range between 0.5 hours to 2 hours before the first time, a first insulin bolus may be received in a time range between 0.5 minutes to 10 minutes after the first time and a first carbohydrate intake may be received in a time range between 10 minutes and 30 minutes after the first time.

Thus, in a preferred embodiment, the at least one first context data value was received in a time range of from 4 hours before the first time and 30 minutes after the first time. The first context data value may be received by manual input of the user. It may alternatively or additionally be received from a device such as an activity tracker and/or a medication delivery pump, such as an insulin delivery pump.

Different first data sets may comprise different types of first context data values.

The second analyte value may be any value of an analyte, such as a concentration, or an absolute amount. In an embodiment, the second analyte value is the same value of an analyte as the first analyte value. Therefore, if the first analyte value is an analyte concentration, then the second analyte value preferably is an analyte concentration, as well. Thus, in an embodiment, the second analyte value is a second analyte concentration.

In an embodiment, the analyte of the second analyte value is the same analyte as the analyte of the first analyte value.

Thus, preferably the second analyte value is a second glucose concentration.

The second analyte value may be received by any method. The second analyte value may be received by the same method as the first analyte value. For example, the second analyte value may be received by a continuous glucose sensor, by a flash glucose sensor or by a blood glucose meter.

The second analyte value is received at a second time, which is different from the first time. The second time may be any time of the day. In an embodiment, the second time is later than the first time. Preferably the second time is at least 30 minutes after the first time. More preferably, the second time is at least 1 hour after the first time and most preferably at least 2 hours after the first time. The second time is usually not more than 8 hours, preferably not more than 6 hours and particularly preferably not more than 4 hours after the first time. For example, the second time is in a time range from 2 hours after the first time to 4 hours after the first time.

It is, thus, preferred that the second time at which the second analyte value was received is at least one hour after the first time at which the first analyte value was received.

In step b) it is determined if the second analyte value is within a pre-selected range.

As generally used, the term “pre-selected range” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art. The pre-selected range may be pre-selected by a user or for a user, for example by a healthcare professional. In an embodiment, the pre-selected range is a standard value range of the second analyte value. Within the context of the present invention, the standard value range is the range of values which is expected for a healthy user. For example, if the second analyte value is a second glucose concentration, then the pre-selected range may be a euglycemic range. Thus, if the second analyte value is a second glucose concentration, then the pre-selected range may be in the range from 70 mg/dl to 180 mg/dl glucose, preferably in the range from 80 mg/dl to 140 mg/dl glucose.

It is, thus, preferred that the second analyte value is a second glucose concentration and that the pre selected range is in the range from 80 mg/dl to 140 mg/dl glucose.

In step c) a machine learning model is generated in a machine learning process. The machine learning process comprises training the machine learning model by at least one first data set which comprises the second analyte value which is within the pre-selected range.

If the second analyte value is outside the pre-selected range, then the first data set comprising this second analyte value is in a preferred embodiment discarded. “Discarded” within the context of the present invention means that the first data set is not used as a training data set and that it is not used to train the machine learning model in step c).

Therefore, it is preferred that the first data set is discarded if in step b) the second analyte value is outside the pre-selected range.

In a further embodiment, the following step bb) is carried out before step c): bb) if the second analyte value is outside the pre-selected range then the first data set comprising the second analyte value which is outside the pre-selected range is discarded.

Step bb) if preferably carried out after step b) and before step c).

In an embodiment, before step c) a step be) is carried out: be) pre-processing the first data set comprising the second analyte value which is within the pre selected range to obtain a pre-processed first data set comprising the second analyte value which is within the pre-selected range.

Step be) is preferably carried out after step b) and before step c). Step be) may be carried out before or after optional step bb), preferably step be) is carried out after optional step bb). For the pre-processing in step be), the embodiments and methods described above for optional step aa) hold true.

If step be) is carried out then in the steps following step be) the pre-processed first data set comprising the second analyte value which is within the pre-selected range is used.

Step be) may not be carried out if step aa) is carried out. Step aa) may not be carried out if step be) is carried out.

In step c) a machine learning model is generated in a machine learning process. The machine learning process comprises training the machine learning model by the at least one first data set which comprises the second analyte value which is within the pre-selected range.

As generally used the term “machine learning model” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art. The machine learning model generated in step c) may be any machine learning model, such as an artificial neural net and/or decision trees.

In the machine learning process the machine learning model is then trained. It is trained by the at least one first data set which comprises the second analyte value for which in step b) it was determined that it is within a pre-selected range. The training of machine learning models as such is known to the skilled person and depends, among others, on the used machine learning model.

In step d) the software-implemented module comprising a prediction algorithm representing the machine learning model is generated.

As generally used the term “software-implemented module” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art and is not to be limited to a special or customized meaning. A software-implemented module may be a functionality of a software program which is independent from other functionalities of the software program.

The software-implemented module is configured to when loaded into a data processing device having one or more processors, determine a drug dose from analyzing a second data set.

In particular, the prediction algorithm comprised in the software-implemented module is configured for determining a drug dose from analyzing a second data set. The second data set comprises a third analyte value which was received at a third time and at least one second context data value.

The second data set may be pre-processed to obtain a pre-processed second data set before it is analyzed.

The third analyte value may be any value of an analyte, such as a concentration, or an absolute amount. In an embodiment, the third analyte value is the same value of an analyte as the first analyte value and/or the second analyte value. Therefore, if the first analyte value is an analyte concentration and/or the second analyte value is an analyte concentration then the third analyte value is preferably an analyte concentration, as well. Thus, in an embodiment, the third analyte value is a third analyte concentration.

In an embodiment, the analyte of the third analyte value is the same analyte as the analyte of the first analyte value and as the analyte of the second analyte value.

Thus, preferably the third analyte value is a third glucose concentration.

In a particularly preferred embodiment, the first analyte value is a first glucose concentration, the second analyte value is a second glucose concentration and the third analyte value is a third glucose concentration.

The third analyte value may be received by any method. For example, the third analyte value may be received by a continuous glucose monitoring sensor, by a flash glucose sensor or by a blood glucose meter. The third analyte value may be received by the same method as the method by which the first analyte value and the second analyte value were received.

The third analyte value was received at a third time. The third time may be any time of the day. Preferably, the third time is a point in time immediately before or immediately after at least one of the at least one second context data value is obtained.

“Immediately before or immediately after at least one of the at least one second context data value is obtained” within the context of the second context data value means immediately before or immediately after the second context data value is created. This is not, on necessity, the time at which the second context data value is received in the second data set. This means that if the second context data value is, for example a second insulin bolus, then “immediately before or immediately after at least one of the at least one second context data value is obtained” means immediately before or immediately after the second insulin bolus is given to the user. If the at least one second context data value is for example an activity index, then “immediately before or immediately after at least one of the at least one second context data value is obtained” means immediately before the activity is started or immediately after the activity is finished.

For example, if one of the at least one second context data value is a second medication dose, such as a second insulin bolus, or a second carbohydrate intake, such as a meal, then the third time is preferably a point in time immediately before or immediately after the a second medication dose is given to the user or the second carbohydrate intake is taken by the user. In an embodiment, the second analyte value is a second glucose value which is measured immediately before a second carbohydrate intake is taken by the user.

The third time is usually in a time range after the first time at which the first analyte data was received and after the second time at which the second analyte data was received. The third time may be in a time range of hours, days or even weeks or months after the first time and/or the second time. The third time is preferably less than one year after the first time at which the first analyte data was received and less than one year after the second time at which the second analyte data was received.

In an embodiment, the third time at which the third analyte value was received is at least 1 hour after the second time at which the second analyte value was received.

As generally used the term “second context data value” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art and is not to be limited to a special or customized meaning. A second context data value may be a value relating to a physiological state of a user or a value being able to influence a physiological state of a user.

Values relating to a physiological state of a user are, for example the body temperature, an analyte pattern, an illness, a stress level and/or a hormonal status.

The analyte pattern is preferably a pattern of the same analyte as the analyte of the second analyte value and/or the first analyte value. Therefore, if, for example, the second analyte is glucose, then the analyte pattern is preferably a glucose pattern, such as a hypoglycemia or a hyperglycemia. A hormonal status is for example the status of the female menstrual cycle.

Values being able to influence a physiological state of a user are, for example activity indices, a second medication dose and/or a second carbohydrate intake.

A second medication dose may be any second medication dose, such as a second insulin dose (insulin bolus) or a second glucagon dose.

An activity index may be an index of any activity of a user, such as an athletic activity.

The second context data value may in particular be selected from at least one of the group consisting of a second medication dose, a second carbohydrate intake, an activity index, and an analyte pattern.

The second context data value may be received in a time range between 10 hours before and 10 hours after the third time at which the third analyte value was received, for example. In an embodiment, the second context data value was received in a time range of from 4 hours before the third time at which the third analyte value was received and 30 minutes after the third time at which the third analyte value was received. It is also possible that the second context data values are received during different time ranges. For example, it is possible that at least one of the at least one second context data values is received in a time range of 0.5 minutes to 15 minutes after the third time, whereas another of the at least one second context data values is receives in a time range between 0.5 hours and 2 hours before the third time. For example, an activity index may be received in a time range between 0.5 hours to 2 hours before the third time, an insulin bolus may be received in a time range between 0.5 minutes to 10 minutes after the third time and a carbohydrate intake may be received in a time range between 10 minutes and 30 minutes after the third time.

Thus, in a preferred embodiment, the at least one second context data value was received in a time range of from 4 hours before the third time and 30 minutes after the third time.

The second context data value may be received by manual input of the user. It may alternatively or additionally be received from a device such as an activity tracker and/or a medication delivery pump, such as an insulin delivery pump.

The second context data may be the same context data as the first context data of at least one of the first data sets. A drug dose is determined from analyzing the second data set.

As generally used the term “drug” is a broad term and is to be given its ordinary and customary meaning to a person of ordinary skill in the art. A drug is typically used to diagnose, cure, treat, or prevent disease . Thus, the drug may for example be selected from the group consisting of antibiotics, chemotherapy drugs, pain relievers, insulin and glucagon, preferably selected from insulin and glucagon.

Thus, in an embodiment, the drug dose is an insulin bolus and/or an insulin basal rate.

The prediction algorithm comprised in the software -implemented module determines the drug dose from analyzing the second data set.

In the method for determining a drug dose according to the present invention, in step al) a second data set is provided. The second data set comprises a third analyte value which was received at a third time and at least one second context data value.

The second data set can be provided by any known method. In particular, the second data set may be provided by the methods described for the provision of the first data set. The second data set may, for example be provided from a remote device, from a cloud device or from the one or more data processing devices.

The second data set may be pre-processed in a step ala), which is carried out before or after step al). Step ala) is in particular carried out before step b 1). Methods for pre-processing the second data set are known as such. Pre-processing comprises for example outlier detection and outlier elimination.

Thus, in an embodiment the following step ala) is carried out before or after step al) and before step bl): ala) pre-processing the second data set to obtain a pre-processed second data set.

If step ala) is carried out, then in the steps following step ala) the pre-processed second data set is used.

For the third analyte value which was received at a third time and the at least one second context data value, the embodiments and preferences described above hold true.

In step bl) a software-implemented module is provided. The software-implemented module comprises a prediction algorithm representing a machine learning model. For the software -implemented module, the embodiments and preferences described above hold true. The software -implemented module may be provided by any method. In an embodiment, the software- implemented module is provided according to steps a) to d) as described above.

In step cl) the drug dose is determined. Determining the drug dose comprises analyzing the second data set by the prediction algorithm. For the determining of the drug dose the embodiments and preferences described above hold true.

In a step dl), the drug dose which was determined may be send to a medication delivery pump. In an embodiment, in a step el), the medication delivery pump will then administer the drug dose to a user.

Thus, in an embodiment, a step dl) is carried out after step cl), wherein the drug dose which was determined in step cl) is sent to a medication delivery pump.

In a further embodiment, after step dl), a step el) is carried out, wherein the medication delivery pump will administer the drug dose which was sent to the medication delivery pump in step dl) to a user.

If, for example, the drug dose is an insulin dose, then the medication delivery pump may be an insulin delivery pump and the determined insulin dose will then be sent to the insulin delivery pump. In an embodiment, the insulin delivery pump may then administer the insulin dose to a user.

The method for determining a drug dose may for example be carried out on a smart phone, on a computer, on a tablet, on a receiver of a blood glucose meter or of a continuous glucose monitoring device, on a remote control of a medication delivery pump or on a medication delivery pump itself. Thus, the system for determining a drug dose preferably comprises a smart phone, a computer, a tablet, a receiver of a blood glucose meter or of a continuous glucose monitoring device, a remote control of a medication deliver pump and/or a medication delivery pump.

The embodiments disclosed above and below with respect to at least one of the methods may apply to at least one of the systems mutatis mutandis.

Description of the Figures

Further optional features and embodiments of the invention will be disclosed in more detail in the subsequent description of preferred embodiments, preferably in conjunction with the dependent claims. Therein, the respective optional features may be realized in an isolated fashion as well as in any arbitrary feasible combination, as the skilled person will realize. The scope of the invention is not restricted by the preferred embodiments. The embodiments are schematically depicted in the Figures. Therein, identical reference numbers in these Figures refer to identical or functionally comparable elements.

In the Figures

Figure 1 is a flow chart depicting the method steps for generating a software -implemented module and the method steps for training a machine learning model.

Figure 2 is a flow chart depicting the method steps for determining a drug dose.

Detailed Description of Embodiments

In figure 1 a flow chart of the method for generating a software-implemented module configured to determine a drug dose and of the method for training a machine learning model to generate a software- implemented module is depicted. The method comprises the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time (110), b) determining if the second analyte value is within a pre-selected range (111), c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range (112), d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model (113).

The software implemented module generated in step d) (113) is configured to, when loaded into a data processing device having one or more processors, determine a drug dose from analyzing a second data set which comprises a third analyte value which was received at a third time and at least one second context data value.

Figure 1 also shows optional step aa) pre-processing the first data set to obtain a pre-processed first data set (110a). In the method shown in figure 1, optional step aa) (110a) is carried out after step a) (110) and before step b) (111). It is clear to the skilled person that if step aa) (110a) is carried out after step a) (110) and before step b) (111), then in step b) (111) and in all steps following step b) (111), such as step c) (112) and step d) (113), the pre-processed first data set is used instead of the first data set. A flow chart of the method for determining a drug dose according to the present invention is shown in figure 2. The method comprises the following steps: al) providing a second data set, the second data set comprising a third analyte value which was received at a third time, at least one second context data value (220), bl) providing a software -implemented module comprising a prediction algorithm representing a machine learning model generated in a machine learning process (221), cl) determining the drug dose comprising analyzing the second data set by the prediction algorithm

(222).

Figure 2 also shows optional step dl) (223), which is carried out after step cl) (222), wherein the drug dose which was determined in step cl) (222) is sent to a medication delivery pump. Optional step e 1) (224) is shown in figure 2, as well. Optional step el) (224) is carried out after optional step dl) (223). In optional step el) (224) the medication delivery pump administers the drug dose which was sent to the medication delivery pump in step dl) (223) to a user.

The software-implemented module is provided in step bl) (221) according to steps a) (110), b) (111), c) (112) and d) (113) of the method for generating the software-implemented module as shown in figure 1.

List of reference numbers

110 step a)

110a optional step aa) 111 step b)

112 step c)

113 step d)

220 step al)

221 step bl) 222 step cl)

223 optional step dl)

224 optional step el)

Claims

1. A method for generating a software-implemented module configured to determine a drug dose, the method comprising, in one or more data processing devices the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time, b) determining if the second analyte value is within a pre-selected range, c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range, d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model, wherein the software-implemented module is configured to, when loaded into a data processing device having one or more processors, determine a drug dose from analyzing a second data set which comprises a third analyte value which was received at a third time and at least one second context data value.

2. A method for training a machine learning model to generate a software-implemented module, the method comprising in one or more data processing devices the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time, b) determining if the second analyte value is within a pre-selected range, c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range, d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model.

3. A method for determining a drug dose, the method comprising in one or more data processing devices the steps: al) providing a second data set, the second data set comprising a third analyte value which was received at a third time, at least one second context data value, bl) providing a software-implemented module comprising a prediction algorithm representing a machine learning model generated in a machine learning process, cl) determining the drug dose comprising analyzing the second data set by the prediction algorithm, wherein generating the software-implemented module in a machine learning process comprises the steps: a) providing at least one first data set, the first data set comprising: a first analyte value which was received at a first time, at least one first context data value, a second analyte value which was received at a second time, which is different from the first time, b) determining if the second analyte value is within a pre-selected range, c) generating a machine learning model in a machine learning process comprising training the machine learning model by the at least one first data set, which comprises the second analyte value which is within the pre-selected range, d) generating the software-implemented module comprising a prediction algorithm representing the machine learning model.

4. The method according to any one of claims 1 to 3, wherein the first data set provided in step a) is a first training data set.

5. The method according to any one of claims 1 to 4, wherein the first context data value is selected from at least one of the group consisting of a first medication dose, a first carbohydrate intake, an activity index, an analyte pattern and/or the second context data value is selected from at least one of the group consisting of a second medication dose, a second carbohydrate intake, an activity index, and an analyte pattern.

6. The method according to any one of claims 1 to 5, wherein the second time at which the second analyte value was received is at least 1 hour after the first time at which the first analyte value was received.

7. The method according to any one of claims 1 to 6, wherein the second analyte value is a second glucose concentration and wherein the pre-selected range is in the range from 80 mg/dl to 140 mg/dl glucose.

8. The method according to any one of claims 1 to 7, wherein the at least one first context data value was received in a time range of from 4 hours before the first time and 30 minutes after the first time and/or wherein the at least one second context data value was received in a time range of from 4 hours before the third time and 30 minutes after the third time.

9. The method according to any one of claims 1 to 8, wherein the first data set is discarded if in step b) the second analyte value is outside the pre-selected range.

10. The method according to any one of claims 1 or 3 to 9, wherein the third time at which the third analyte value was received is at least 1 hour after the second time at which the second analyte value was received.

11. A system for generating a software-implemented module configured to determine a drug dose, the system comprising one or more data processing devices, the one or more data processing devices configured to:

B) determine if the second analyte value is within a pre-selected range,

D) generate the software -implemented module comprising a prediction algorithm representing the machine learning model, wherein the software-implemented module is configured to when loaded into a data processing device having one or more processors, determine a drug dose from analyzing a second data set which comprises a third analyte value which was received at a third time and at least one second context data value.

12. A system for training a machine learning model to generate a software-implemented module, the system comprising one or more data processing devices, the one or more data processing devices configured to:

B) determine if the second analyte value is within a pre-selected range,

13. A system for determining a drug dose, the system comprising one or more data processing devices, the one or more data processing devices configured to:

Bl) provide a software implemented module comprising a prediction algorithm representing a machine learning model generated in a machine learning process,

B) determining if the second analyte value is within a pre-selected range,

D) generating the software-implemented module comprising a prediction algorithm representing the machine learning model.

14. A computer program product, comprising program code configured to, when loaded into a computer having one or more processors, perform the method according to at least one of claims 1 to 10.