SYSTEMS AND METHODS FOR OBTAINING AND PRESENTING CLINICAL RESPONSE INFORMATION USING MACHINE-LEARNED MODELS TRAINED ON IMPLANTED NEUROSTIMULATOR DATA

1 . A method of informing of a clinical response of a patient as a function of an event of interest that occurred at an associated time, the method comprising:

processing an input dataset to obtain a pre-event window of model inputs and a post-event window of model inputs, wherein the input dataset is selected from a subject-patient dataset comprising a plurality of data types that are based on electrical activity of the patient's brain sensed and stored by an implanted neurostimulation system, and at least one feature of the patient;

applying a clinical response estimator (CRE) model to the pre-event window of model inputs to derive one or more pre-event CRE biomarkers, wherein the clinical response estimator is trained on datasets across a patient population;

applying the CRE model to the post-event window of model inputs to derive one or more post-event CRE biomarkers; and

displaying as a function of time, the one or more pre-event CRE biomarkers, the one or more post-event CRE biomarkers, and an occurrence of the event of interest.

2 . The method of claim 1 , wherein processing an input dataset to obtain a pre-event window of model inputs and a post-event window of model inputs comprises applying a first machine-learned model to a first subset of the input dataset to obtain one or more of the model inputs, wherein the first machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as ictal records, and exclude data types of records of electrical activity of the brain classified as interictal records.

3 . The method of claim 1 , wherein processing an input dataset to obtain a pre-event window of model inputs and a post-event window of model inputs comprises applying a second machine-learned model to a second subset of the input dataset to obtain one or more of the model inputs, wherein the second machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as interictal records, and exclude data types of records of electrical activity of the brain classified as ictal records.

4 . The method of claim 1 , wherein the event of interest comprises one of a replacement of a component of the implanted neurostimulation system, a change in operation of the implanted neurostimulation system, a change in therapy for the patient, and a change in patient habits.

5 . The method of claim 1 , wherein the input dataset comprises at least one data type that is sensed and stored over a time period, and that is characterized by a value, the method further comprising:

displaying the values of the at least one data type as a function of time.

6 . The method of claim 1 , further comprising:

applying a machine-learned model to the input dataset to identify a plurality of key inputs; and

displaying values of the key inputs as a function of time.

7 . An apparatus for informing of a clinical response of a patient as a function of an event of interest that occurred at an associated time, the apparatus comprising:

a display;

a memory having a plurality of modules; and

a processor coupled to the memory and the display, and configured to execute operations based on the plurality of modules to:

process an input dataset to obtain a pre-event window of model inputs and a post-event window of model inputs, wherein the input dataset is selected from a subject-patient dataset comprising a plurality of data types that are based on electrical activity of the patient's brain sensed and stored by an implanted neurostimulation system, and at least one feature of the patient;

apply a clinical response estimator (CRE) model to the pre-event window of model inputs to derive one or more pre-event CRE biomarkers, wherein the clinical response estimator is trained on datasets across a patient population;

apply the CRE model to the post-event window of model inputs to derive one or more post-event CRE biomarkers; and

display as a function of time, the one or more pre-event CRE biomarkers, the one or more post-event CRE biomarkers, and an occurrence of the event of interest.

8 . The apparatus of claim 7 , wherein to process the input dataset, the processor is configured to:

apply a first machine-learned model to a first subset of the input dataset to obtain one or more of the model inputs, wherein the first machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as ictal records, and exclude data types of records of electrical activity of the brain classified as interictal records.

9 . The apparatus of claim 7 , wherein to process the input dataset, the processor is configured to:

apply a second machine-learned model to a second subset of the input dataset to obtain one or more of the model inputs, wherein the second machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as interictal records, and exclude data types of records of electrical activity of the brain classified as ictal records.

10 . The apparatus of claim 7 , wherein the event of interest comprises one of a replacement of a component of the implanted neurostimulation system, a change in operation of the implanted neurostimulation system, a change in therapy for the patient, and a change in patient habits.

11 . A method of modifying an operation of an implanted neurostimulation system of a patient, the method comprising:

determining a clinical response estimate (CRE) biomarker by:

deriving an input dataset from a subject-patient dataset comprising a plurality of data types that are based on electrical activity of the patient's brain sensed and stored by the implanted neurostimulation system, and at least one feature of the patient, wherein the input dataset is derived based on a plurality of key inputs of the subject-patient dataset,

processing the input dataset to obtain a plurality of model inputs, and

applying a machine-learned CRE model to the plurality of model inputs to determine the CRE biomarker, wherein the machine-learned CRE model is trained on datasets across a patient population;

comparing the CRE biomarker to a criterion; and

responsive to the criterion not being met, adjusting a parameter of the operation, or triggering the operation.

12 . The method of claim 11 , wherein processing the input dataset comprises applying a first machine-learned model to a first subset of the input dataset to obtain one or more of the plurality of model inputs, wherein the first machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as ictal records, and exclude data types of records of electrical activity of the brain classified as interictal records.

13 . The method of claim 11 , wherein processing the input dataset comprises applying a second machine-learned model to a second subset of the input dataset to obtain one or more of the plurality of model inputs, wherein the second machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as interictal records, and exclude data types of records of electrical activity of the brain classified as ictal records.

14 . The method of claim 11 , wherein processing the input dataset comprises one or a combination of:

combining a plurality of data types to obtain a corresponding model input;

combining a plurality of patient features to obtain a corresponding model input; and

applying a machine-learned model to records of electrical activity of the brain to obtain a model input corresponding to one of a brain activity type and a numeric value.

15 . The method of 11 , wherein the operation comprises electrographic event detection and the adjusted parameter is a detection parameter.

16 . The method of 11 , wherein the operation comprises stimulation delivery and the adjusted parameter is a stimulation parameter.

17 . The method of claim 11 , wherein the triggered operation comprises stimulation delivery.

18 . The method of 11 , wherein determining is one of:

continuous, and

triggered by an event of interest corresponding to one of a replacement of a component of the implanted neurostimulation system, a change in operation of the implanted neurostimulation system, a change in therapy, and a change in patient habits.

19 . An implantable neurostimulation system comprising:

a detection subsystem;

a therapy subsystem;

a memory having a plurality of modules; and

a processor coupled to the memory, the detection subsystem, and the therapy subsystem, and configured to execute operations based on the plurality of modules to:

determine a clinical response estimate (CRE) biomarker by being further configured to:

derive an input dataset from a subject-patient dataset comprising a plurality of data types that are based on electrical activity of a patient's brain sensed and stored by the implanted neurostimulation system, and at least one feature of a patient, wherein the input dataset is derived based on a plurality of key inputs of the subject-patient dataset,

process the input dataset to obtain a plurality of model inputs, and

apply a machine-learned CRE model to the plurality of model inputs to determine the CRE biomarker, wherein the machine-learned CRE model is trained on datasets across a patient population;

compare the CRE biomarker to a criterion; and

adjust a parameter of the operation or trigger the operation in response to the criterion not being met.

20 . The implantable neurostimulation system of claim 19 , wherein to process the input dataset, the processor is configured to:

apply a first machine-learned model to a first subset of the input dataset to obtain one or more of the plurality of model inputs, wherein the first machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as ictal records, and exclude data types of records of electrical activity of the brain classified as interictal records.

21 . The implantable neurostimulation system of claim 19 , wherein to process the input dataset, the processor is configured to:

apply a second machine-learned model to a second subset of the input dataset to obtain one or more of the plurality of model inputs, wherein the second machine-learned model is trained on datasets that include data types derived from records of electrical activity of the brain classified as interictal records, and exclude data types of records of electrical activity of the brain classified as ictal records.

22 . The implantable neurostimulation system of claim 19 , wherein the operation comprises electrographic event detection and the processor is configured to adjust a detection parameter of the detection subsystem.

23 . The implantable neurostimulation system of claim 19 , wherein the operation comprises stimulation delivery and the processor is configured to adjust a stimulation parameter of the therapy subsystem.

24 . The implantable neurostimulation system of claim 19 , wherein the triggered operation comprises stimulation delivery and the processor is configured to initiate stimulation delivery by the therapy subsystem.