Platforms for conducting virtual trials
Inventors
SHRAGER, Jeffrey C. • Tenenbaum, Jay Martin • PORTER, Christopher Kelly • Hoos, William Arthur • SHAPIRO, Mark Adam
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11887738-B2
Publication Date
2024-01-30
Expiration Date
Abstract
The present disclosure provides platforms, systems, media, and methods for capturing clinical cases and expert-derived treatment rationales to facilitate biomedical decision making, which can include virtual clinical trials that continuously learn from the experiences of all patients, on all treatments, and all the time. Algorithms such as Bayesian machine learning methods can be applied to coordinate such virtual trials.
Core Innovation
The invention provides a computing platform for precision oncology that captures a clinical case using a clinical case capture tool with selectable clinical case templates, each including adaptive clinical case parameters with selectable values. The platform receives user input pertaining to at least one template, parameter, and value to capture a clinical case of a patient, and generates a summary of the captured clinical case that includes at least one treatment rationale for the patient. The platform further publishes the captured clinical case information into a clinician workflow for expert clinician network review.
A clinician survey application publishes the summary of a captured clinical case to an expert clinician network and collects peer review feedback pertaining to a treatment rationale. Peer review feedback is used to store a plurality of peer reviewed treatment rationales to form a knowledge base comprising clinical cases and treatment options and associated treatment rationales. In parallel, the platform updates the knowledge base using a registry that collects outcome data of the captured clinical case.
A clinical decision engine applies a machine learning algorithm utilizing the knowledge base to provide one or more ranked treatment protocols or treatment options based on predicted outcomes. The decision engine conducts a Bayesian decision process to prioritize the ranked treatment protocols or treatment options based at least in part on efficacy and variance or uncertainty of an equipoise set, and prioritizes the equipoise set based at least in part on a relative amount of information gained from the predicted outcomes. The platform coordinates treatment decisions across a plurality of patients by using the Bayesian decision process and training the machine learning algorithm with updated outcome data in the knowledge base.
Clinician feedback and captured rationales are incorporated into a continual learning loop by updating the knowledge base with outcome data from the captured clinical case. The system trains the machine learning algorithm using updated knowledge database training data comprising a plurality of input features and the outcome data, thereby enabling ongoing refinement of the ranked protocols or recommendations. The core platform design integrates clinical case capture, peer vetting, knowledge-base storage, Bayesian prioritization, and knowledge-base updating.
Claims Coverage
Two independent claims define the core platform. The inventive features include capturing adaptive clinical case templates and generating treatment-rationale summaries, publishing cases to an expert clinician network for peer review and storing peer reviewed rationales as a knowledge base, using a machine learning algorithm with a Bayesian decision process that prioritizes an equipoise set using efficacy, uncertainty, and relative information gain, and updating the knowledge base using outcome data from a registry and retraining with input features and outcomes.
Adaptive clinical case capture templates and treatment-rationale summary generation
A clinical case capture tool presenting a plurality of selectable clinical case templates, each template comprising a plurality of adaptive clinical case parameters with selectable values; receiving user input pertaining to at least one template, parameter, and value to capture a clinical case of a patient; and generating a summary of a captured clinical case comprising at least one treatment rationale for the patient.
Publishing to expert clinician network, peer review feedback collection, and knowledge-base storage
A clinician survey application publishing the summary of the captured clinical case to an expert clinician network; collecting peer review feedback pertaining to a treatment rationale of the summary; and storing a plurality of peer reviewed treatment rationales to form a knowledge base.
Machine learning clinical decision engine with Bayesian prioritization using efficacy, uncertainty, and information gain
A clinical decision engine receiving user input to identify a cohort and a treatment hypothesis; applying a machine learning algorithm utilizing the knowledge base to provide one or more ranked treatment protocols based on predicted outcomes; conducting a Bayesian decision process to prioritize the one or more ranked treatment protocols to coordinate treatment decisions across a plurality of patients, where the Bayesian decision process prioritizes the ranked treatment protocols based at least in part on efficacy and variance or uncertainty of an equipoise set, and prioritizes the equipoise set based at least in part on relative amount of information gained from predicted outcomes across the equipoise set.
Outcome registry updating of the knowledge base and training with updated outcomes
Providing a registry for collecting outcome data of the captured clinical case to update the knowledge base, wherein the machine learning algorithm is trained using the updated knowledge base comprising a training data set comprising a plurality of input features and the outcome data of the captured clinical case.
Knowledge database and clinician feedback driven treatment option recommendations with Bayesian coordination
A knowledge database comprising a plurality of clinical cases and treatment options and treatment rationales associated with the plurality of clinical cases; a clinical case capture tool for capturing a clinical case; a clinician survey application publishing the clinical case to an expert clinician network and collecting feedback for one or more treatment options and associated treatment rationales for the clinical case; and a clinical decision engine comprising a machine learning algorithm utilizing the knowledge database to provide a user with at least one recommendation of the one or more treatment options based on the feedback, wherein a Bayesian decision process prioritizes ranked treatment protocols based at least in part on efficacy and variance or uncertainty of an equipoise set, and prioritizes the equipoise set based at least in part on relative amount of information gained from predicted outcomes across the equipoise set, with the knowledge database updated with outcome data for the captured clinical case and the machine learning algorithm trained using the updated knowledge database comprising input features and the outcome data of the captured clinical case.
Across the two independent claims, the core claim coverage is a computing platform that captures adaptive clinical cases and generates treatment-rationale summaries, vets rationales through an expert clinician peer-review workflow to build a knowledge base, and uses a machine-learning-driven clinical decision engine that performs a Bayesian decision process. The Bayesian prioritization is explicitly tied to predicted efficacy, uncertainty or equipoise, and relative information gain, and the platform coordinates recommendations across a plurality of patients while updating the knowledge base with registry-collected outcomes and retraining on input features and outcomes.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?