Prognosis Modeling From One or More Sources of Information

• US 20080033894A1
• Filed: 04/16/2007
• Published: 02/07/2008
• Est. Priority Date: 04/17/2006
• Status: Active Grant

• Current Assignee: Cerner Innovation, Inc. (Oracle Corporation)
• Original Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Inventors: Rao, R., Steck, Harald, Lambin, Philippe, Krishnan, Sriram, Dehing-Oberije, Cary
• Granted Patent: US 7,805,385 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

A predictor of medical treatment outcome is developed and applied. A prognosis model is developed from literature. The model is determined by reverse engineering the literature reported quantities. A relationship of a given variable to a treatment outcome is derived from the literature. A processor may then use individual patient values for one or more variables to predict outcome. The accuracy may be increased by including a data driven model in combination with the literature driven model.

Claims

1. A system for predicting medical treatment outcome, the system comprising:

an input operable to receive values for a plurality of patient specific characteristics of a patient;

a processor operable to apply the values to a first prognosis model, the first prognosis model relating a plurality of variables corresponding to the values to a treatment outcome, the relating being a function of medical knowledge collected from literature and incorporated into the first prognosis model; and

a display operable to output a patient specific prognosis for the patient as a function of the application of the values for the plurality of patient specific characteristics of the patient to the first prognosis model.

2. The system of claim 1 wherein the input comprises a data interface operable to receive the values mined from data by the processor or a different processor.

3. The system of claim 1 wherein the relating performed by the processor is a function of relationships elicited from medical doctors.

4. The system of claim 1 wherein the relating performed by the processor is a function of probabilities corresponding to individual predictors modeled for each variable.

5. The system of claim 1 wherein the first prognosis model is a function of medical knowledge collected from literature converted into a functional relationship for each variable independent of other variables.

6. The system of claim 1 wherein the relating is determined by modeling a continuous-time relationship from one or more discrete relationships from the literature.

7. The system of claim 1 wherein the processor is operable to apply the values or additional values of the patient to a second prognosis model, the second prognosis model being a statistical model derived from a database of patient data for a plurality of patients.

8. The system of claim 7 wherein the processor is operable to apply the values or additional values of the patient to a third prognosis model, the third prognosis model being a statistical model derived from the database of patient data for the plurality of patients, different values or additional values applied to the second prognosis model than for the third prognosis model, the different values or additional values for the second prognosis model being based on the literature and for the third prognosis model being based on the database.

9. The system of claim 7 wherein the patient specific prognosis is a function of outputs from the first and second prognosis models.

10. The system of claim 1 wherein the relating performed by the processor for one variable is a function of probabilities derived from a hazard ratio, a survival rate, and a fraction receiving a treatment option, the hazard ratio, the survival rate and the fraction provided in the literature, the first prognosis model incorporating a survival rate for the treatment option derived from the hazard ratio, the survival rate, and the fraction.

11. In a computer readable storage medium having stored therein data representing instructions executable by a programmed processor for predicting medical treatment outcome, the instructions comprising:

applying first and second prognosis models to individual patient information, the first and second prognosis models having different knowledge base sources;

predicting a first treatment outcome with the first prognosis model from the individual patient information and as a function of quantities from clinical study publications as a first knowledge base source; and

predicting a second treatment outcome with the second prognosis model from the individual patient information and as a function of model parameters from a database of past patient information for a plurality of patients as a second knowledge base source.

12. The instructions of claim 11 further comprising:

combining the first and second treatment outcomes.

13. The instructions of claim 11 wherein predicting with the first and second prognosis models comprises predicting with a combined prognosis model fusing the first and second prognosis models.

14. The instructions of claim 11 wherein predicting the first treatment outcome comprises:

inputting the individual patient information, the information comprising values for variables identified in the clinical study publications as correlated with the first treatment outcome;

applying each of the values to a corresponding survival rate as a function of time derived from a survival rate at a given time from the clinical study publication; and

probabilistically combining outputs from each of the applications for corresponding values, the combination being the first treatment outcome.

15. A method for developing a predictor of medical treatment outcome, the method comprising:

collecting quantitative medical knowledge of predictors for a disease from literature;

modeling a plurality of univariate relationships between a treatment and the predictors, respectively, from the quantitative medical knowledge;

combining models for the univariate relationships into a multivariate model; and

operating, with a processor, the combined models as a prognosis model for an individual patient as a function of values for the predictors, the values being for the individual patient.

16. The method of claim 15 further comprising:

deriving a data driven model by processing data for a plurality of patients in a database;

wherein operating comprises applying the data driven model and the multivariate model.

17. The method of claim 15 wherein collecting comprises obtaining a table, a graph, text, or combinations thereof;

and

wherein modeling comprises;

fitting a curve to a graph;

establishing logic statements as a function of a table;

determining a binary classification with a logistic or Cox regression as a function of a table, graph, text, or combinations thereof;

or

combinations thereof.

18. The method of claim 15 wherein modeling a first univariate relationship of the plurality of univariate relationships comprises:

identifying a variable predictive of an overall survival rate from the quantitative medical knowledge, the overall survival rate, a hazard ratio, and a fraction of study patients associated with one value of the variable to study patients associated with another value of the variable;

applying an exponential distribution of survival times; and

determining a first predictive survival rate from the study patients associated with the one value and a second predictive survival rate from the study patients associated with the other value, both as a function of the exponential distribution, the overall survival rate, and the fraction.

19. The method of claim 15 wherein combining comprises forming a Bayesian network as the multivariate model from the univariate relationships.

20. The method of claim 15 wherein modeling and combining comprise programming, and wherein operating comprises mining, with the processor, the values from a medical record of the individual patient, applying the values to the combined models, and outputting a survival rate for a treatment of the individual patient from the combined models.