Data aggregation, integration and analysis system and related devices and methods
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12288608-B2
Publication Date
2025-04-29
Expiration Date
Abstract
A system for recording, storing and processing diagnostic information, including: a computer implementing a computer-readable media including digital data and ground truth; a registry constructed and arranged to store and associate transactions or accesses on the data; and a machine learning system that considers each learning step modification a microtransaction for the data used in that step and which is recorded in the transaction registry. Other embodiments of this aspect include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the methods.
Core Innovation
The described invention registers a plurality of training images on a distributed ledger, where each training image is labeled with truth data. The distributed ledger records training activity in an association between each given training image and a sequence of model changes attributable to that image.
During training of a machine learning model using the plurality of training images, the training includes recording each of a plurality of modifications of the machine learning model that are attributable to a given training image as a microtransaction to the distributed ledger, in association with the given training image.
The invention addresses attribution of changes in a trained machine learning model’s performance to specific training data, particularly where medical diagnostics rely on training images and corresponding truth data. The described embodiments support associating truth data with diagnostic data and, optionally, an entity that determined the diagnostic data.
Claims Coverage
The independent claims cover three parallel aspects of the same attribution-by-ledger concept: a method, a non-transitory machine-readable medium, and a system. Across these independent claims, the coverage centers on two inventive features.
Registering labeled training images on a distributed ledger
Registering a plurality of training images on a distributed ledger, each training image labeled with truth data.
Recording attributable model modifications as microtransactions linked to training images
Training a machine learning model using the plurality of training images, where as the machine learning model is trained, recording each of a plurality of modifications of the machine learning model that are attributable to a given training image as a microtransaction to the distributed ledger in association with the given training image.
Non-transitory medium instructions to perform registration and microtransaction recording
A non-transitory machine-readable medium comprising memory with instructions encoded thereon, the instructions, when executed, causing one or more processors to perform operations including registering a plurality of training images on a distributed ledger, each training image labeled with truth data; and training a machine learning model using the plurality of training images, including recording each of a plurality of modifications of the machine learning model attributable to a given training image as a microtransaction to the distributed ledger in association with the given training image.
System performing registration and attribution recording
A system comprising memory with instructions encoded thereon and one or more processors that, when executing the instructions, are caused to perform operations comprising registering a plurality of training images on a distributed ledger, each training image labeled with truth data; and training a machine learning model using the plurality of training images, including recording each of a plurality of modifications of the machine learning model that are attributable to a given training image as a microtransaction to the distributed ledger in association with the given training image.
Across the independent claims, the inventive core is the ledger-based registration of training images with truth data and the recording, during training, of microtransactions that associate specific model modifications with the specific training image(s) to which they are attributable. Dependent claim coverage further specifies neural networks, diagnostic data, and optionally provenance about an entity that determined diagnostic data.
Stated Advantages
Enables post-training attribution of contributions to model performance changes assessed on test metrics.
Provides a record of model modifications attributable to a given training image as microtransactions on the distributed ledger.
Documented Applications
Medical diagnostic data attribution for a machine learning system trained using training images and ground truth, with performance assessment using test data set performance metrics.
Attribution of model performance changes to specific training images and labeled data elements during machine learning training in a distributed ledger-based workflow.
Interested in licensing this patent?