AUTONOMOUS DIAGNOSIS OF EAR DISEASES FROM BIOMARKER DATA
A fully autonomous system is used to diagnose an ear infection in a patient. For example, a processor receives patient data about a patient, the patient data comprising at least one of: patient history from medical records for the patient, one or more vitals measurements of the patient, and answers from the patient about the patient's condition. The processor receives a set of biomarker features extracted from measurement data taken from an ear of the patient. The processor synthesizes the patient data and the biomarker features into input data, and applies the synthesized input data to a trained diagnostic model, the diagnostic model comprising a machine learning model configured to output a probability-based diagnosis of an ear infection from the synthesized input data. The processor outputs the determined diagnosis from the diagnostic model. A service may then determine a therapy for the patient based on the determined diagnosis.
1 . A method for training a diagnostic model for use in diagnosing an ear malady, the method comprising:
accessing a plurality of training examples that were generated by, for each of a plurality of patients:
receiving a set of biomarker features extracted from measurement data taken from an ear of a patient by:
obtaining an ear image of a portion of the patient's ear; and
extracting one or more biomarker features from the image by:
obtaining a set of samples of the ear image, each sample corresponding to a location in the ear;
for each of the set of samples, applying the sample to a trained feature detection model, the feature detection model comprising a neural network that is configured to output a likelihood of whether the sample contains an ear image object; and
storing a training example for the patient, the training example comprising at least the biomarker features for the patient and a label that indicates whether the patient has an ear malady; and
repeatedly applying given ones of the plurality of training examples to the diagnostic model.
2 . The method of claim 1 , further comprising updating parameters of the diagnostic model to improve an objective performance threshold when repeatedly applying the given ones of the plurality of training examples to the diagnostic model.
3 . The method of claim 1 , wherein the ear image is at least one of a two-dimensional image or a three-dimensional optical coherence tomography image.
4 . The method of claim 1 , wherein the portion of the patient's ear comprises at least one of a tympanic membrane, an anatomical structure adjacent to a tympanic membrane, an ear canal adjacent to the tympanic membrane, a malleus, an umbo, and a light reflex.
5 . The method of claim 1 , wherein receiving the set of biomarker features comprises:
applying a pressure stimulus to inside an ear of the patient;
receiving an acoustic response from the applied pressure stimulus;
extracting acoustic biomarker features from the received acoustic response; and
synthesizing, into input data, the acoustic biomarker features.
6 . The method of claim 5 , wherein the pressure stimulus is applied using at least one of pneumatic otoscopy or tympanometry.
7 . The method of claim 1 , wherein the biomarker features each further indicate a confidence value that reflects a confidence that an indicated anatomical feature of the ear was accurately determined.
8 . A computer program product for training a diagnostic model for use in diagnosing an ear malady, the computer program product comprising a computer-readable storage medium containing computer program code for:
accessing a plurality of training examples that were generated by, for each of a plurality of patients:
receiving a set of biomarker features extracted from measurement data taken from an ear of a patient by:
obtaining an ear image of a portion of the patient's ear; and
extracting one or more biomarker features from the image by:
obtaining a set of samples of the ear image, each sample corresponding to a location in the ear;
for each of the set of samples, applying the sample to a trained feature detection model, the feature detection model comprising a neural network that is configured to output a likelihood of whether the sample contains an ear image object; and
storing a training example for the patient, the training example comprising at least the biomarker features for the patient and a label that indicates whether the patient has an ear malady; and
repeatedly applying given ones of the plurality of training examples to the diagnostic model.
9 . The method of claim 1 , further comprising updating parameters of the diagnostic model to improve an objective performance threshold when repeatedly applying the given ones of the plurality of training examples to the diagnostic model.
10 . The method of claim 1 , wherein the ear image is at least one of a two-dimensional image or a three-dimensional optical coherence tomography image.
11 . The method of claim 1 , wherein the portion of the patient's ear comprises at least one of a tympanic membrane, an anatomical structure adjacent to a tympanic membrane, an ear canal adjacent to the tympanic membrane, a malleus, an umbo, and a light reflex.
12 . The method of claim 1 , wherein receiving the set of biomarker features comprises:
applying a pressure stimulus to inside an ear of the patient;
receiving an acoustic response from the applied pressure stimulus;
extracting acoustic biomarker features from the received acoustic response; and
synthesizing, into input data, the acoustic biomarker features.
13 . The method of claim 5 , wherein the pressure stimulus is applied using at least one of pneumatic otoscopy or tympanometry.
14 . The method of claim 1 , wherein the biomarker features each further indicate a confidence value that reflects a confidence that an indicated anatomical feature of the ear was accurately determined.
15 . A system for training a diagnostic model for use in diagnosing an ear malady, the system comprising:
memory with instructions encoded thereon; and
one or more processors that, when executing the instructions, are caused to perform operations comprising:
accessing a plurality of training examples that were generated by, for each of a plurality of patients:
receiving a set of biomarker features extracted from measurement data taken from an ear of a patient by:
obtaining an ear image of a portion of the patient's ear; and
extracting one or more biomarker features from the image by:
obtaining a set of samples of the ear image, each sample corresponding to a location in the ear;
for each of the set of samples, applying the sample to a trained feature detection model, the feature detection model comprising a neural network that is configured to output a likelihood of whether the sample contains an ear image object; and
storing a training example for the patient, the training example comprising at least the biomarker features for the patient and a label that indicates whether the patient has an ear malady; and
repeatedly applying given ones of the plurality of training examples to the diagnostic model.
16 . The system of claim 15 , the operations further comprising updating parameters of the diagnostic model to improve an objective performance threshold when repeatedly applying the given ones of the plurality of training examples to the diagnostic model.
17 . The system of claim 15 , wherein the ear image is at least one of a two-dimensional image or a three-dimensional optical coherence tomography image.
18 . The system of claim 15 , wherein the portion of the patient's ear comprises at least one of a tympanic membrane, an anatomical structure adjacent to a tympanic membrane, an ear canal adjacent to the tympanic membrane, a malleus, an umbo, and a light reflex.
19 . The system of claim 15 , wherein receiving the set of biomarker features comprises:
applying a pressure stimulus to inside an ear of the patient;
receiving an acoustic response from the applied pressure stimulus;
extracting acoustic biomarker features from the received acoustic response; and
synthesizing, into input data, the acoustic biomarker features.
20 . The system of claim 19 , wherein the pressure stimulus is applied using at least one of pneumatic otoscopy or tympanometry.