PHYSIOLOGY-DRIVEN DECISION SUPPORT FOR THERAPY PLANNING

• US 20170071671A1
• Filed: 09/11/2015
• Published: 03/16/2017
• Est. Priority Date: 09/11/2015
• Status: Abandoned Application

• Current Assignee: Siemens Healthcare GMBH (Siemens AG)
• Original Assignee: Siemens Healthcare GMBH (Siemens AG)
• Inventors: Neumann, Dominik, Mansi, Tommaso, Passerini, Tiziano, Mihalef, Viorel, Pauly, Olivier, Georgescu, Bogdan, Ecabert, Olivier
• Application Number:
  US14/851,552
• Publication Number:
  US 20170071671A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: A61B 2034/101 Computer-aided simulation of surgical operations
• CPC Class Codes: A61B 34/10 Computer-aided planning, simulation or modelling of surgical operations
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders

Abstract

Using computational models for the patient physiology and the various therapy options, a decision support system presents a range of predicted outcomes to assist in planning the therapy. The models are used in various experiments for the many therapy options to determine an optimal approach.

Claims

1. A method for decision support for therapy, the method comprising:

segmenting organ data representing an organ of a first patient from scan data from a medical scanner, the scan data representing a volume of the first patient;

simulating, by a processor, a plurality of different therapies with a physiological model personalized to the organ based on the segmented data, the different therapies being for a therapy device with different parameters and/or for different therapy devices;

estimating, by the processor, uncertainties in the simulated outcomes of the different therapies; and

presenting on a display the simulated outcomes of the simulating of the different therapies and the estimated uncertainties.

2. The method of claim 1 wherein simulating comprises simulating with different physiological models.

3. The method of claim 1 wherein simulating comprises encoding a guideline as a Markov decision process.

4. The method of claim 3 further comprising learning the Markov decision process with reinforcement learning.

5. The method of claim 1 wherein simulating comprises identifying past patients similar to the first patient and using past patient results for the different therapies as the results of the simulation for the first patient.

6. The method of claim 5 wherein identifying comprises finding features of the first patient for comparison to data of the past patients with a first deep auto-encoder.

7. The method of claim 6 wherein finding with the first deep auto-encoder comprises finding using clinical information, hemodynamic information, and medical scan information reduced a low-dimensional, digital representation, and comparing the low-dimensional, digital representation to the data of the past patients.

8. The method of claim 7 wherein the hemodynamic information is estimated from the physiological model prior to simulation of the different therapies;

further comprising finding another set of past patients with a second deep auto-encoder with hemodynamic information resulting from the simulating of the different therapies;

wherein presenting comprises presenting the results based on the finding of the other set of past patients.

9. The method of claim 1 wherein estimating comprises estimating the uncertainties in the simulated outcomes of the simulating based on variability of the simulated outcomes.

10. The method of claim 1 wherein presenting comprises presenting the simulated outcomes and respective estimated uncertainties in a ranked order.

11. The method of claim 1 wherein segmenting the organ data comprises segmenting the organ data representing a vessel, wherein simulating comprises simulating with the physiological model providing biomechanical parameters with inverse modeling or computation fluid dynamics and the different therapies comprising variation in stent properties in a stent model, the simulating being interaction of the stent model with the physiological model.

12. A method for decision support for therapy, the method comprising:

inputting patient information from different sources to a first deep auto-encoder, the patient information specific to a first patient and a type of therapy device;

selecting similar patients to the first patient with an output of the first deep auto-encoder, the similar patients having been treated with the type of therapy device;

inferring a range of outcomes from a range of therapy devices of the type of therapy device from data for the similar patients; and

displaying the range of outcomes and the range of therapy devices for the first patient.

13. The method of claim 12 further comprising estimating uncertainties of the outcomes, wherein displaying comprises displaying the outcomes with the uncertainties.

14. The method of claim 12 wherein inputting comprises inputting the patient information as clinical data, hemodynamic factors, and imaging data.

15. The method of claim 12 wherein selecting comprises selecting with the output being a digital representation of a lower dimensional representation of the patient information.

16. The method of claim 12 wherein inferring comprises aggregating frequency and efficacy for each of the therapy devices.

17. The method of claim 12 further comprising simulating treatment of the therapy devices with a physiological model fit with at least some of the patient information, inputting the patient information and results of the simulation of the treatment to a second deep auto-encoder, and selecting another set of similar patients based on an output of the second deep auto-encoder.

18. The method of claim 12 wherein the type of therapy device comprises a stent, the patient information includes vessel information from a medical scanner, and the range of therapy devices comprises stents with different properties.

19. A method for decision support for therapy, the method comprising:

inputting patient information from different sources to a first deep auto-encoder, the patient information specific to a first patient and a type of therapy device;

selecting first similar patients to the first patient with an output of the first deep auto-encoder, the similar patients having been treated with the type of therapy device;

inferring a first range of first outcomes from a range of therapy devices of the type of therapy device from data for the first similar patients;

selecting at least one of the therapy devices based on the outcome;

simulating treatment by the selected at least one of the therapy devices using a physiological model personalized to the first patient and a model of the type of therapy device specific to the at least one of the therapy devices;

calculating hemodynamic factors resulting from the simulation of the treatment;

inputting the hemodynamic factors and at least some of the patient information to a second deep auto-encoder;

selecting second similar patients to the first patient with an output of the second deep auto-encoder;

inferring at least one second outcome from the at least one of the therapy devices from data for the second similar patients;

displaying the at least one second outcome and the at least one therapy device for the first patient.

20. The method of claim 19 further comprising simulating as a function of reinforcement learning.