Rapid assessment and outcome analysis for medical patients
Machine learning is used to assess data for a patient in an emergency, providing rapid diagnosis based on a large amount of information. Assistance in triage may be provided. Given the large variety of patients and conditions that may occur, the machine learning may rely on synthetically generated images for more accurate prediction. The machine learning may accurately predict even with missing information and may be used to determine what missing information for a given patient is more or less important to obtain.
1. A method for assessment and/or outcome determination in a medical system, the method comprising:
acquiring medical scan data representing anatomical structures of patients being triaged in an emergency department of a hospital;
obtaining other data for the patients;
extracting a set of features from the medical scan data and the other data;
inputting, by a processor, the features to a machine-trained classifier, the machine trained classifier trained, at least in part, from synthetically generated images not specific to training data for existing patient cases;
determining, by the processor with application of the machine-trained classifier to the features, conditions of the patients and risks for the conditions;
providing an order for triage of the patients based on the determined risks;
identifying, by the processor, a missing source of information based on the features that are available for a first patient of the patients and the determined condition for the first patient, wherein the missing source of information is identified as more likely than other missing sources to result in a different condition or risk for the first patient; and
outputting a recommendation to acquire the missing source of information.
2. The method of claim 1 wherein acquiring comprises acquiring computed tomography data of cardiac regions of the patients.
3. The method of claim 1 wherein obtaining other data comprises obtaining symptoms, patient demographics, blood biomarkers, patient history, non-invasive measurements, extracts from a clinical report, or combinations thereof.
4. The method of claim 1 wherein extracting the features from the medical scan data comprises applying a machine-learnt classifier to the medical scan data, the machine-learnt classifier and machine-trained classifier comprising a cascade.
5. The method of claim 1 wherein determining comprises determining with the application of the machine-trained classifier where the machine-trained classifier is trained from a forward model producing the synthetically generated images from variations in an anatomical model.
6. The method of claim 5 wherein the production of the synthetically generated images is guided by sampling of conditions of the training data.
7. The method of claim 1 wherein determining comprises determining with the application of the machine-trained classifier where the machine-trained classifier is trained from the synthetically generated images with different degrees of statistical noise added.
8. The method of claim 1 wherein acquiring comprises acquiring from a past scan of the first patient, and wherein obtaining comprises obtaining current symptoms.
9. The method of claim 1 wherein determining the risk for the first patient comprises determining with a dictionary of patient data with corresponding outcomes.
10. The method of claim 1 wherein inputting comprises inputting the features for the first patient and additional information extracted from other patients for missing features of the first patient, and
wherein outputting comprises outputting the missing source based on a stochastic distribution resulting from the additional information.
11. The method of claim 1 further comprising identifying past cases similar to the first patient and indicating a distribution of treatments and outcomes.
12. The method of claim 1 further comprising identifying a sub-set of the medical scan data, other data, or both with a machine-learnt model and outputting the identified sub-set.
13. The method of claim 1 further comprising recommending a wearable sensor for the first patient based on a machine-learnt recommender applied to the other data, medical scan data, or both.
14. A method for assessment and/or outcome determination in a medical system, the method comprising:
acquiring medical scan data representing anatomical structures of a patient in an emergency;
obtaining other data for the patient;
extracting a set of features from the medical scan data and the other data;
determining, by the processor with application of the machine-trained classifier to the features, a condition of the patient, a risk for the patient, an outcome for the patient, a similar patient, or combinations thereof; and
outputting a recommendation to acquire a missing source of information based on missing features for the patient and the determined condition, the determined risk, or the determined outcome of the patient, wherein the missing source is identified as more likely than other missing sources to result in a different condition, risk, or outcome for the patient.
15. The method of claim 14 wherein determining comprises determining with application of the machine-trained classifier to the extracted features and stochastic sampling of the missing features over additional information extracted from other patients, and
wherein outputting comprises identifying the missing source based on results from the machine-trained classifier with the stochastic sampling.
16. The method of claim 14 , further comprising:
identifying, by the processor, the missing source of information based on missing features for the patient and the determined condition, the determined risk, or the determined outcome of the patient.