Back to results
US 12,670,974 B2Granted
Generating notes using machine learning
AssigneeBirth Model, Inc.
InventorAnish J. Shah
Life of this patent
Estimating the term…
Abstract
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a note for a subject. In one aspect, a method comprises: obtaining a note template that defines a format of a note for a subject; automatically querying one or more record databases to obtain a set of electronic record data for the subject; generating a model input to a generative machine learning model based at least in part on: (i) the note template, and (ii) the set of electronic record data for the subject; processing the model input using the generative machine learning model and in accordance with values of a set of generative machine learning model parameters to generate a model output that defines the note for the subject; and outputting the note for the subject.
Claims24
A method performed by one or more computers, the method comprising:
- obtaining a clinical note template that defines a format of a clinical note for a patient, comprising:
- presenting, to a user and on a display of a user device, a plurality of possible clinical note templates, wherein each of the plurality of possible clinical note templates comprises a sequence of text with a plurality of wildcard fields, wherein each wildcard field is designated to be replaced with a corresponding type of information related to the patient;
- and
- receiving, from the user, a selection that identifies the clinical note template from among the plurality of possible clinical note templates;
- wherein the clinical note template is a template for an operative note that describes a surgery performed on the patient or a template for a delivery note that describes a childbirth process for the patient;
- automatically querying one or more electronic medical record databases to obtain a set of electronic medical record data for the patient;
- automatically querying each of a plurality of sensors that are on or in proximity to the patient to obtain a waveform generated by the sensor that characterizes a physiological state of the patient, wherein the plurality of sensors comprise one or more of:
- a fetal heartrate sensor, or a dilation sensor, or a drug dosage sensor, or a capnography sensor, or an anesthesia depth sensor;
- processing, for each of the plurality of sensors, the waveform generated by the sensor using a respective classification machine learning model;
- generating a patient state classification based on the processed waveform, wherein the patient state classification classifies a state of the patient into a class from a set of classes;
- generating a model input to a generative neural network, the generating the model input comprising serializing and concatenating:
- (i) the clinical note template, and (ii) patient data comprising the set of electronic medical record data for the patient and the patient state classifications generated by processing the waveforms generated by the plurality of sensors;
- processing the model input using the generative neural network and in accordance with values of a set of generative neural network parameters to generate a model output that defines the clinical note for the patient,
- wherein the generative neural network has been trained by a machine learning training technique to perform a language modeling task;
- wherein the model output that defines the clinical note for the patient comprises a sequence of textual tokens;
- and
- wherein processing the model input using the generative neural network and in accordance with values of the set of generative neural network parameters to generate the model output that defines the clinical note for the patient comprises, for each of a plurality of positions in the sequence of textual tokens of the clinical note:
- processing:
- (i) the model input, and (ii) any textual tokens at preceding positions that precede the position in the sequence of textual tokens of the clinical note, using the generative neural network model to generate a score distribution over a set of possible textual tokens;
- and
- selecting a textual token to occupy the position in the sequence of textual tokens using the score distribution over the set of possible textual tokens;
- receiving, by a user interface and from the user, a free-form textual query that defines a request to modify the clinical note for the patient;
- generating a new model input to the generative neural network that comprises a concatenation of:
- (i) the clinical note for the patient, and (ii) the free-form textual query that defines the request to modify the clinical note for the patient;
- processing the new model input using the generative neural network and in accordance with the values of the set of generative neural network parameters to generate a new clinical note for the patient that reflects the request to modify the clinical note for the patient;
- and
- outputting the new clinical note for the patient.
24 more elements
A system comprising:
- one or more computers;
- and
- one or more storage devices communicatively coupled to the one or more computers, wherein the one or more storage devices store instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:
- obtaining a clinical note template that defines a format of a clinical note for a patient, comprising:
- presenting, to a user and on a display of a user device, a plurality of possible clinical note templates, wherein each of the plurality of possible clinical note templates comprises a sequence of text with a plurality of wildcard fields, wherein each wildcard field is designated to be replaced with a corresponding type of information related to the patient;
- and
- receiving, from the user, a selection that identifies the clinical note template from among the plurality of possible clinical note templates;
- wherein the clinical note template is a template for an operative note that describes a surgery performed on the patient or a template for a delivery note that describes a childbirth process for the patient;
- automatically querying one or more electronic medical record databases to obtain a set of electronic medical record data for the patient;
- automatically querying each of a plurality of sensors that are on or in proximity to the patient to obtain a waveform generated by the sensor that characterizes a physiological state of the patient, wherein the plurality of sensors comprise one or more of:
- a fetal heartrate sensor, or a dilation sensor, or a drug dosage sensor, or a capnography sensor, or an anesthesia depth sensor;
- processing, for each of the plurality of sensors, the waveform generated by the sensor using a respective classification machine learning model;
- generating a patient state classification based on the processed waveform, wherein the patient state classification classifies a state of the patient into a class from a set of classes;
- generating a model input to a generative neural network, the generating the model input comprising serializing and concatenating:
- (i) the clinical note template, and (ii) patient data comprising the set of electronic medical record data for the patient and the patient state classifications generated by processing the waveforms generated by the plurality of sensors;
- processing the model input using the generative neural network and in accordance with values of a set of generative neural network parameters to generate a model output that defines the clinical note for the patient,
- wherein the generative neural network has been trained by a machine learning training technique to perform a language modeling task;
- wherein the model output that defines the clinical note for the patient comprises a sequence of textual tokens;
- and
- wherein processing the model input using the generative neural network and in accordance with values of the set of generative neural network parameters to generate the model output that defines the clinical note for the patient comprises, for each of a plurality of positions in the sequence of textual tokens of the clinical note:
- processing:
- (i) the model input, and (ii) any textual tokens at preceding positions that precede the position in the sequence of textual tokens of the clinical note, using the generative neural network model to generate a score distribution over a set of possible textual tokens;
- and
- selecting a textual token to occupy the position in the sequence of textual tokens using the score distribution over the set of possible textual tokens;
- receiving, by a user interface and from the user, a free-form textual query that defines a request to modify the clinical note for the patient;
- generating a new model input to the generative neural network that comprises a concatenation of:
- (i) the clinical note for the patient, and (ii) the free-form textual query that defines the request to modify the clinical note for the patient;
- processing the new model input using the generative neural network and in accordance with the values of the set of generative neural network parameters to generate a new clinical note for the patient that reflects the request to modify the clinical note for the patient;
- and
- outputting the new clinical note for the patient.
27 more elements
One or more non-transitory computer storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
- obtaining a clinical note template that defines a format of a clinical note for a patient, comprising:
- presenting, to a user and on a display of a user device, a plurality of possible clinical note templates, wherein each of the plurality of possible clinical note templates comprises a sequence of text with a plurality of wildcard fields, wherein each wildcard field is designated to be replaced with a corresponding type of information related to the patient;
- and
- receiving, from the user, a selection that identifies the clinical note template from among the plurality of possible clinical note templates;
- wherein the clinical note template is a template for an operative note that describes a surgery performed on the patient or a template for a delivery note that describes a childbirth process for the patient;
- automatically querying one or more electronic medical record databases to obtain a set of electronic medical record data for the patient;
- automatically querying each of a plurality of sensors that are on or in proximity to the patient to obtain a waveform generated by the sensor that characterizes a physiological state of the patient, wherein the plurality of sensors comprise one or more of:
- a fetal heartrate sensor, or a dilation sensor, or a drug dosage sensor, or a capnography sensor, or an anesthesia depth sensor;
- processing, for each of the plurality of sensors, the waveform generated by the sensor using a respective classification machine learning model;
- generating a patient state classification based on the processed waveform, wherein the patient state classification classifies a state of the patient into a class from a set of classes;
- generating a model input to a generative neural network, the generating the model input comprising serializing and concatenating:
- (i) the clinical note template, and (ii) patient data comprising the set of electronic medical record data for the patient and the patient state classifications generated by processing the waveforms generated by the plurality of sensors;
- processing the model input using the generative neural network and in accordance with values of a set of generative neural network parameters to generate a model output that defines the clinical note for the patient,
- wherein the generative neural network has been trained by a machine learning training technique to perform a language modeling task;
- wherein the model output that defines the clinical note for the patient comprises a sequence of textual tokens;
- and
- wherein processing the model input using the generative neural network and in accordance with values of the set of generative neural network parameters to generate the model output that defines the clinical note for the patient comprises, for each of a plurality of positions in the sequence of textual tokens of the clinical note:
- processing:
- (i) the model input, and (ii) any textual tokens at preceding positions that precede the position in the sequence of textual tokens of the clinical note, using the generative neural network model to generate a score distribution over a set of possible textual tokens;
- and
- selecting a textual token to occupy the position in the sequence of textual tokens using the score distribution over the set of possible textual tokens;
- receiving, by a user interface and from the user, a free-form textual query that defines a request to modify the clinical note for the patient;
- generating a new model input to the generative neural network that comprises a concatenation of:
- (i) the clinical note for the patient, and (ii) the free-form textual query that defines the request to modify the clinical note for the patient;
- processing the new model input using the generative neural network and in accordance with the values of the set of generative neural network parameters to generate a new clinical note for the patient that reflects the request to modify the clinical note for the patient;
- and
- outputting the new clinical note for the patient.
24 more elements
Dependencies are inferred from the claim text.