Heterogeneous multi-core processing systems and data routing methods for high-throughput model predictive medical systems

Patent metadata

Field	Value
Publication number	US-9706963-B2
Application number	US-201414573802-A
Country	US
Kind code	B2
Filing date	Dec 17, 2014
Priority date	Dec 17, 2013
Publication date	Jul 18, 2017
Grant date	Jul 18, 2017

Abstract

A heterogeneous multiple core control system for predictive medical monitoring is provided. The system includes a first processor platform that is optimized for serialized computation and a second processor platform that is optimized for parallelized computation. A memory stores instructions that are executed by either the first processor platform or the second processor platform to provide the functionality of the system. By executing the instructions, the system determines whether a condition of a first patient is to be monitored using a primarily parallelized model or a primarily serialized model. The predicted condition is calculated using the primarily parallelized model by executing the primarily parallelized model on the second processor platform. The predicted condition is calculated using the primarily serialized model by executing the primarily serialized model on the first processor platform.

First claim

What is claimed is:

1. A heterogeneous multiple core control system for predictive medical monitoring comprising:
a first processor platform optimized for serialized computation;
a second processor platform optimized for parallelized computation; and
a memory storing instructions that, when executed by the first processor platform and the second processor platform, cause the system to
determine whether a condition of a first patient is to be monitored using a primarily parallelized model or a primarily serialized model,
calculate a predicted condition of the first patient using the primarily parallelized model by executing the primarily parallelized model on the second processor platform, and
calculate the predicted condition of the first patient using the primarily serialized model by executing the primarily serialized model on the first processor platform; and

wherein the instructions, when executed by the first processor platform and the second processor platform, further cause the system to
monitor a first change in average execution time resulting from adding a new patient monitored using the primarily serialized model on the first processor platform, and
monitor a second change in average execution time resulting from adding a new patient monitored using the primarily serialized model on the second processor platform, and
wherein the instructions, when executed by the first processor platform and the second processor platform, cause the system to calculate the predicted condition of the first patient using the primarily serialized model by executing the primarily serialized model on the first processor platform when the first change in average execution time is less than the second change in average execution time, and
executing the primarily serialized model on the second processor platform when the second change in average execution time is less than the first change in average execution time.

2. The heterogeneous multiple core control system of claim 1 , wherein the instructions, when executed by the first processor platform and the second processor platform, further cause the system to monitor a plurality of patients, wherein a first number of patients are monitored using the primarily parallelized model and a second number of patients are monitored using the primarily serialized model.

3. The heterogeneous multiple core control system of claim 2 , wherein the instructions, when executed by the first processor platform and the second processor platform, cause the system to calculate the predicted condition of the first patient using the primarily serialized model by executing the primarily serialized model on the first processor platform when the second number of patients monitored using the primarily serialized model is less than a threshold, and executing the primarily serialized model on the second processor platform when the second number of patients monitored using the primarily serialized model is greater than the threshold.

4. The heterogeneous multiple core control system of claim 1 , wherein the primarily serialized model is computationally less intensive than the primarily parallelized model and requires more iterations than the primarily parallelized model to reflect changes in the monitored condition.

5. The heterogeneous multiple core control system of claim 1 , further comprising sensors to monitor the concentration of an infused drug in the patient's body.

6. The heterogeneous multiple core control system of claim 1 , wherein the predicted condition is a future concentration of the infused drug in the patient's body.

7. The heterogeneous multiple core control system of claim 6 , wherein the instructions, when executed by the first processor platform and the second processor platform, further cause the system to determine an adjusted infusion amount based on the predicted future concentration, and provide a control instruction to an infusion pump causing the infusion pump to operate based on the adjusted infusion amount.

8. The heterogeneous multiple core control system of claim 1 , wherein the first processor platform includes a lower number of processing cores than the second processor platform and operates at a higher execution speed than the second processor platform.

9. The heterogeneous multiple core control system of claim 1 , further comprising a controller board, wherein the controller board includes the second processor platform.

10. The heterogeneous multiple core control system of claim 1 , wherein the primarily serialized model includes a pharmacokinetic drug diffusion model.

11. The heterogeneous multiple core control system of claim 1 , wherein the primarily parallelized model includes a Finite Domain Time Diffusion model.

12. The heterogeneous multiple core control system of claim 1 , wherein the primarily parallelized model includes a grid identifying drug concentrations at different anatomical locations of the first patient, wherein the grid is stored in a cache memory accessible by each processing core of the second processor platform, and wherein each iteration of the primarily parallelized model of the plurality of parallel processing cores accesses a plurality of predicted values from the grid and calculates a new value for a specific anatomical location.

13. A heterogeneous multiple core control system for predictive medical monitoring comprising:
a first processor platform optimized for serialized computation;
a second processor platform optimized for parallelized computation; and
a memory storing instructions that, when executed by the first processor platform and the second processor platform, cause the system to
determine whether a condition of a first patient is to be monitored using a primarily parallelized model or a primarily serialized model,
calculate a predicted condition of the first patient using the primarily parallelized model by executing the primarily parallelized model on the second processor platform, and
calculate the predicted condition of the first patient using the primarily serialized model by executing the primarily serialized model on the first processor platform

wherein the instructions, when executed by the first processor platform and the second processor platform, further cause the system to monitor a plurality of patients, wherein a first number of patients are monitored using the primarily parallelized model and a second number of patients are monitored using the primarily serialized model, and

wherein the instructions, when executed by the first processor platform and the second processor platform, further cause the system to balance a number of patient monitored by executing a model on the first processing platform and a number of patients monitored by executing a model on the second processing platform to reduce an average execution time.

14. The heterogeneous multiple core control system of claim 13 , wherein the primarily serialized model is computationally less intensive than the primarily parallelized model and requires more iterations than the primarily parallelized model to reflect changes in the monitored condition.

15. The heterogeneous multiple core control system of claim 13 , further comprising sensors to monitor the concentration of an infused drug in the patient's body.

16. The heterogeneous multiple core control system of claim 13 , wherein the predicted condition is a future concentration of the infused drug in the patient's body.

17. The heterogeneous multiple core control system of claim 16 , wherein the instructions, when executed by the first processor platfor

Assignees

• Gupta Sandeep
• Banerjee Ayan
• Univ Arizona State

Inventors

• Gupta Sandeep
• Banerjee Ayan

Classifications

• A61M2005/14208

  with a programmable infusion control system, characterised by the infusion program · CPC title

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title

• G16B50/00

  ICT programming tools or database systems specially adapted for bioinformatics · CPC title

• A61M5/1723

  using feedback of body parameters, e.g. blood-sugar, pressure (measurement of body parameters A61B5/00) · CPC title

• A61M2230/30

  Blood pressure (A61M2230/04 takes precedence) · CPC title