Systems and methods for using federated learning for training centralized seizure detection and prediction models on decentralized datasets
Inventors
ARCOT DESAI, Sharanya • Tcheng, Thomas K.
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12333443-B2
Publication Date
2025-06-17
Expiration Date
Abstract
A server for updating a current version of a machine learning model resident in implanted medical devices includes an interface, a memory, and a processor. The interface is configured to receive a plurality of updated versions of the machine learning model from a plurality of remote sources remote from the server. The remote source may be, e.g., implanted medical devices and/or subservers. The processor is coupled to the memory and the interface and is configured to aggregate the plurality of updated versions to derive a server-updated version of the machine learning model, and to transmit the server-updated version of the machine learning model to one or more of the plurality of remote sources as a replacement for the current version of the machine learning model.
Core Innovation
The invention relates to replacing a first machine learning model of a first architecture resident in a plurality of implanted medical devices. The first machine learning model is generated using a first dataset and is configured to detect a neurological event in electrical activity of a brain.
The replacement is achieved by providing remote sources with information on a second dataset for generating a second machine learning model of a second architecture different than the first architecture. The second dataset includes at least one type of data not included in the first dataset and comprises records of electrical activity collected in response to detections of neurological events by the first machine learning model.
Each remote source generates a version of the second machine learning model based on its corresponding second dataset. The server receives a plurality of versions from the remote sources, aggregates the plurality of versions to derive a server-generated version, and transmits the server-generated version to one or more remote sources as a replacement for the first machine learning model.
The server architecture includes an interface, memory, and a processor configured to provide the second dataset information to remote sources, aggregate the received versions of the second machine learning model, and transmit the server-generated version as a replacement. In device-level embodiments, an implantable medical device includes the first machine learning model, receives information on the second dataset, extracts the second dataset from stored data, generates a version of the second machine learning model of a second architecture different than the first architecture, and provides the version to a server.
Claims Coverage
The independent claims include replacing an on-device brain-detection machine learning model via a server-coordinated workflow and an implantable medical device that generates and provides a second-model version to the server. Across the independent claims, the core inventive features include dataset distribution that reuses detection-linked electrical activity, remote generation of a different-architecture second machine learning model, server-side aggregation into a server-generated version, and transmitting that version back as a replacement for the first model, with an explicit server configuration and device-side model-version generation.
Replacing a first architecture machine learning model resident in implanted medical devices
A method of replacing a first machine learning model of a first architecture resident in a plurality of implanted medical devices, wherein the first machine learning model is configured to detect a neurological event in electrical activity of a brain.
Providing a second dataset to remote sources for a different architecture model
Providing, from a server to each of a plurality of remote sources remote from the server, information on a second dataset for generating a second machine learning model of a second architecture different than the first architecture, wherein the second dataset includes at least one type of data not included in the first dataset and comprises records of electrical activity collected in response to detections of neurological events by the first machine learning model.
Generating remote versions and aggregating at the server
Generating, at each of the plurality of remote sources, a version of the second machine learning model based on a corresponding second dataset; receiving, at a server, a plurality of versions of the second machine learning model from the plurality of remote sources; aggregating, at the server, the plurality of versions of the second machine learning model to derive a server-generated version of the second machine learning model.
Transmitting a server-generated replacement model to remote sources
Transmitting, at the server, the server-generated version of the second machine learning model to one or more of the plurality of remote sources as a replacement for the first machine learning model.
Server for replacing the first machine learning model by aggregating remote versions
A server comprising an interface configured to receive a plurality of versions of a second machine learning model from a plurality of remote sources remote from the server, and a processor configured to provide to each remote source information on a second dataset for generating a second machine learning model of a second architecture different than the first architecture, aggregate the plurality of versions to derive a server-generated version, and transmit the server-generated version to one or more remote sources as a replacement for the first machine learning model.
Implantable medical device generating a second-model version from a second dataset
An implantable medical device comprising a first machine learning model of a first architecture configured to detect a neurological event in electrical activity of a brain; an interface configured to receive information on a second dataset that includes at least one type of data not included in the first dataset and comprises records of electrical activity collected in response to detections of neurological events by the first machine learning model, and to provide to a server a version of a second machine learning model; and a processor configured to extract the second dataset from data stored in memory and generate the version of the second machine learning model based on the second dataset, wherein the second machine learning model is of a second architecture different than the first architecture.
Taken together, the independent claims define coordinated replacement of an on-device, first-architecture neurological event detection model using a second, different-architecture model. The second dataset is distributed to remote sources using detection-linked electrical activity and at least one additional data type, remote sources generate versions of the second model, a server aggregates those versions into a server-generated model, and the server transmits the server-generated model back as a replacement; in device-side embodiments, the implantable medical device extracts the second dataset and generates the second-model version for provision to the server.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?