Specialized health care system for selecting treatment paths
Inventors
Mitidis, Andonis • Ayers, Jeanine
Assignees
Publication Number
US-11270800-B1
Publication Date
2022-03-08
Expiration Date
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Methods for selecting treatment paths are disclosed generally comprising the steps of: (a) discovering a set of treatment path clusters based on latent patterns in historical patient trace data, (b) building a set of binary classifiers based on historical patient trace data, historical patient data and target outcomes, and (c) given the treatment path clusters, actual patient data, and a selected target outcome, applying the binary classifiers to predict a treatment path for a new patient that optimizing the selected target outcome. Processor based systems to implement the methods are also disclosed.
Core Innovation
Methods for selecting treatment paths are disclosed generally comprising the steps of: (a) discovering a set of treatment path clusters based on latent patterns in historical patient trace data, (b) building a set of binary classifiers based on historical patient trace data, historical patient data and target outcomes, and (c) given the treatment path clusters, actual patient data, and a selected target outcome, applying the binary classifiers to predict a treatment path for a new patient that optimizing the selected target outcome. Processor based systems to implement the methods are also disclosed.
Humans cannot create associations among hundreds of thousands of data, extract patterns and use that extracted information to change the standardize care map in minor or major ways to improve the outcome or to recommend a treatment path. The disclosed solution addresses shortcomings found in existing solutions by trying to solve the problem of optimizing the treatment recommendations by physicians so that the treatment meets desired outcomes.
The systems and methods create 'treatment path clusters' to identify the distinct groupings of clinical treatment paths to be applied to patients, where the treatment paths and treatment path clusters may be statistically determined from historical treatment paths, or patient traces. The solution builds a set of cluster centroids as potential treatment paths, builds a set of binary classifiers based on patient trace data, historical patient data and target outcomes, and, given cluster centroids, new patient data and a selected target outcome, applies the binary classifiers to determine a treatment path for the new patient that optimizes the selected target outcome.
Claims Coverage
Three independent claims are identified: claims 1, 12, and 15. The following inventive features are extracted from those independent claims.
Specialized health care system for selecting a treatment path
A system comprising: a processor; a first input device to receive historical patient trace data comprising a plurality of treatment actions each associated with a time; a data formatter module configured to transform historical patient trace data, historical patient data and new patient data into formatted data; a memory to store formatted data; treatment path modules including a pattern extractor module that extracts treatment patterns as a 3-dimensional array (actions × time × probability), a cluster maker module to create treatment path clusters, and a centroid maker module to create treatment path cluster centroids; a binary classifier module configured to receive formatted historical patient trace data, one or more target outcomes and historical patient data to define and train a binary classifier; predictor modules comprising a target outcome selector and a target outcome maximizer configured to receive treatment path cluster centroids, new patient data and the binary classifier to select a treatment path centroid; and a user interface to present the selected treatment path.
Computer readable medium implementing centroid and classifier selection
A non-transient computer readable medium for causing a processor based device to perform a method of: determining at least one treatment path cluster centroid utilizing a treatment topic extracting algorithm given historical patient trace data and historical patient data where treatment patterns comprise a 3-dimensional array (actions × time × probability); training a binary classifier given the historical patient trace data, the historical patient data, and at least one outcome; applying the binary classifier to determine a probability of the outcome for each treatment path cluster centroid given new patient data; selecting one outcome as the selected outcome; and identifying the treatment path cluster centroid with the highest probability as the selected treatment path.
Method of selecting a treatment path
A method comprising: determining at least one treatment path cluster centroid utilizing a treatment pattern extracting algorithm given historical patient trace data and historical patient data where treatment patterns are 3-dimensional arrays (actions × time × probability); training a binary classifier given historical patient trace data, historical patient data, and at least one outcome; applying the binary classifier to determine a probability of the outcome for each treatment path cluster centroid given new patient data and the centroid; selecting one outcome as the selected outcome; and identifying the centroid with the highest probability as the selected treatment path.
The independent claims relate to (1) a specialized system, (2) a computer-readable medium, and (3) a method for selecting treatment paths by extracting 3-dimensional treatment patterns from time-stamped historical patient trace data, clustering to form centroids, training binary classifiers on formatted historical trace and patient data with target outcomes, and using those classifiers with new patient data to select the centroid that maximizes a chosen outcome.
Stated Advantages
May be much faster, much more accurate and much cheaper than the current solutions.
Accurately make recommendations on patient treatment so that physicians can optimize treatment in light of treatment targets such as the desired outcome, speed and cost of the treatment.
Present scores to a physician for each treatment path indicating likelihoods for target outcomes to provide suggestions/recommendations that reflect how the physician and patient prioritize target outcomes.
Deliver information in real-time to patient interfaces such as the patient's smartphone, computer or other wireless devices and has the potential to influence patient behaviors.
Documented Applications
Selecting treatment paths for a new patient that optimize a selected target outcome (e.g., positive outcome, fast completion, inexpensive cost) by applying binary classifiers to treatment path cluster centroids and new patient data.
Providing suggestions and recommendations to physicians by presenting probability scores for possible treatment paths so physicians can prioritize outcomes such as outcome, cost and speed of treatment.
Serving as a diagnostic tool to aid in identifying and managing a myriad of medical situations by presenting selected treatment paths via a user interface.
Using medical sensors to provide historical patient trace data or new patient data to the system, including physiological and other medical sensors as described, to inform treatment path selection.
Deploying the system as a processor-based implementation or as a computer program product on computing devices including servers and client interfaces.
Interested in licensing this patent?