Learning or inferring medical concepts from medical transcripts using probabilistic models with words or phrases identification
View Patent ↗A medical concept is learned about or inferred from a medical transcript. A probabilistic model is trained from medical transcripts. For example, the problem is treated as a graphical model. Discrimitive or generative learning is used to train the probabilistic model. A mutual information criterion can be employed to identify a discrete set of words or phrases to be used in the probabilistic model. The model is based on the types of medical transcripts, focusing on this source of data to output the most probable state of a patient in the medical field or domain. The learned model may be used to infer a state of a medical concept for a patient.
1. A system for inferring a medical concept from a medical transcript, the system comprising:
an input operable to receive user identification of the medical transcript;
a processor operable to receive a text passage from the medical transcript and operable to apply a probabilistic model to the text passage of the medical transcript, the probabilistic model trained as a function of a discrete set of words or phrases, wherein the probabilistic model comprises a Bayes model having a summary node representing the presence or absence of at least one medical concept and observed during training but not at the time of inference for the text passage, a negation node, and a modifier node, the constructing comprising a function of discriminative learning; and
a display operable to output a state associated with a patient, the state being inferred as a function of an output from the probabilistic model applied to the text passage.
2. The system of claim 1 wherein the state is a disease, condition, symptom, or test result, the state being limited to true and false, or true, false and unknown.
3. The system of claim 1 wherein the medical transcript comprises unstructured, natural language information from a medical professional.
4. The system of claim 1 wherein the output from the probabilistic model comprises an indication of whether a medical concept is indicated in the medical transcript.
5. The system of claim 1 wherein the probabilistic model comprises learned statistics of words or phrases relative to the state.
6. The system of claim 1 wherein the probabilistic model includes probabilities associated with images, structured data, or combinations thereof.
7. The system of claim 1 wherein the probabilistic model comprises conditional and prior probability functions of words or phrases relative to the state.
8. The system of claim 1 wherein the words or phrases included within the discrete set are learned as a function of mutual information.
9. In a computer readable storage medium having stored therein data representing instructions executable by a programmed processor for inferring a medical concept from a medical transcript, the instructions comprising:
applying a probabilistic model to the medical transcript of a patient, the probabilistic model probabilistically associating different words of the medical transcript to a state of the medical concept, wherein the probabilistic model comprises a Bayes model having a summary node representing the presence or absence of at least one medical concept and observed during training but not at the time of inference for the text passage, a negation node, and a modifier node, the constructing comprising a function of discriminative learning; and
outputting the state as indicated by the medical transcript of the medical concept as a function of the probabilistic associations of the different words to the state.
10. The instructions of claim 9 wherein the probabilistic model comprises conditional and prior probability functions of words or phrases relative to the state.
11. A method for learning about a medical concept from a medical transcript, the method comprising:
receiving a plurality of labeled text passages from medical transcripts;
constructing, with a processor, a probabilistic model of the medical concept as a function of the labeled text passages of the medical transcripts , the constructing being a function of discriminative learning; and
outputting the probabilistic model;
wherein constructing the probabilistic model comprises learning with a Bayes model having a summary node representing the presence or absence of at least one medical concept and observed during training but not at the time of inference for the text passage, a negation node, and a modifier node.
12. The method of claim 11 wherein the medical transcripts comprise unstructured, natural language information from a medical professional, and wherein a state of the medical concept is limited to true and false, or true, false and unknown.
13. The method of claim 11 wherein construction comprises learning, with the processor, statistical distributions of words or phrases relative to a state of the medical concept.
14. The method of claim 11 further comprising determining, with the processor, words or phrases to be included within the probabilistic model, the determining being learned as a function of mutual information.
15. The method of claim 11 wherein constructing comprises learning conditional and prior probability functions of words or phrases relative to a state of the medical concept.