Edge Computing for Medical Control Applications

Medical monitoring for critical patients may include model predictive controllers (MPC).  MPC applications obtain a sample from a patients body and  based on their human physiology model either alarm the medical authority or change the drug infusion rate. The estimation of infusion rate is done every few seconds. Hence in a scenario of a multiple patients being monitored in a hospital, running such MPC application to estimate drug within the time constraints need computationally capable devices. This project evaluates  the use of many core platforms like GPU, Intel Xeon Phi (MIC) or Intel Core i7 to execute the MPC applications.

Model predictive controllers run to estimate the drug infusion rate for the patient individually.  Figure shows multiple patients being monitored in the hospital with MPC control algorithm running in data center with high computation capability.

Publications:

  1. Madhurima Pore, Ayan Banerjee and Sandeep K.S. Gupta, Heterogeneous many cores for medical control: Performance, Scalability, and Accuracy, International Conference on High Performance Computing Conference(HiPC14), Goa, India, December 2014 [PDF].
  2. Madhurima Pore, Ayan Banerjee, Sandeep K.S. Gupta, and Hari K Tadepalli, Performance trends of multicore system for throughput computing in medical application, International Conference on High Performance Computing Conference(HiPC13), Hyderabad, India, December 2013 [PDF].
  3. Madhurima Pore, Zahra Abbasi, Georgios Varsamopoulos, and Sandeep K.S. Gupta, Energy aware Colocation of Workload in Data centers, Workshop on Performance Engineering and Applications, Pune, India, December 2012 [PDF]. 

Researchers: Madhurima Pore,  Dr Ayan Banerjee, Dr. Sandeep K. S. Gupta

Funding Acknowledgement: 
Hari Tadepalli, Intel Corp and  NSF.