US 11,734,601 B2
Systems and methods for model-assisted cohort selection
Benjamin Edward Birnbaum, Brooklyn, NY (US); Joshua Daniel Haimson, New York, NY (US); Lucy Dao-Ke He, New York, NY (US); Katharina Nicola Seidl-Rathkopf, Brooklyn, NY (US); Monica Nayan Agrawal, Atlanta, GA (US); and Nathan Nussbaum, South Orange, NJ (US)
Assigned to Flatiron Health, Inc., New York, NY (US)
Filed by Flatiron Health, Inc., New York, NY (US)
Filed on May 3, 2019, as Appl. No. 16/403,475.
Application 16/403,475 is a continuation of application No. 15/951,614, filed on Apr. 12, 2018, granted, now 10,304,000.
Claims priority of provisional application 62/484,984, filed on Apr. 13, 2017.
Prior Publication US 2019/0258950 A1, Aug. 22, 2019
Int. Cl. G06N 20/00 (2019.01); G06N 5/046 (2023.01); G16H 10/60 (2018.01); G16H 10/20 (2018.01); G16H 50/70 (2018.01)
CPC G06N 20/00 (2019.01) [G06N 5/046 (2013.01); G16H 10/20 (2018.01); G16H 10/60 (2018.01); G16H 50/70 (2018.01)] 15 Claims
OG exemplary drawing
 
8. A method for selecting a cohort from among a population of individuals, the method comprising:
receiving, via a data interface, a selection of one or more relevant search terms from a user;
receiving, via the data interface, a plurality of medical records from a database storing records associated with selected individuals in a population of individuals, each of the selected individuals being associated with one or more of the medical records;
extracting, from the plurality of medical records, one or more snippets based on the one or more relevant search terms, each of the one or more snippets comprising a subset of text, within the plurality of records, surrounding the one or more relevant search terms and including a plurality of neighboring terms in addition to the one or more relevant search terms;
deriving, based on the one or more extracted snippets, one or more feature vectors associated with one or more of the plurality of medical records;
providing the one or more feature vectors to a trained machine learning model;
receiving outputs from the model, the outputs comprising scores for ones of the selected individuals, the one or more feature vectors being associated with one or more of the medical records associated with the scored individuals; and
determining whether the scored individuals are candidates for a cohort based on the received scores, wherein the determination is based on a comparison of each of the received scores to a predetermined threshold that is adjustable.