Batch Bayesian optimization design for optimizing a neurostimulator

Adam Kaplan; Thomas A. Murray

doi:10.1111/biom.13313

Batch Bayesian optimization design for optimizing a neurostimulator

Research output: Contribution to journal › Article › peer-review

Abstract

Recently, spinal epidural neurostimulation is being considered for rehabilitation of persons suffering from partial spinal-cord injury. The neurostimulator must be programmed by a neurosurgeon, yet little work has been done to develop rigorous methods for optimally programming the device. We propose an adaptive design to efficiently optimize programming of the neurostimulator based on specified interim evaluations of patient reported preferences. Preferences for the eligible device configurations are estimated after each interim analysis through a conditionally autoregressive model that assumes preference for one configuration is related to preferences for neighboring configurations. Using the adaptively updated preferences, a group of configurations is programmed into the device for the patient to evaluate during the next follow-up period. This selection is based on a balance of device exploration and preference maximization. We repeat this process until a specified stopping rule or the calibration end is reached. We show simulation studies to evaluate the overall quality of the adaptive calibration for various configuration selection strategies and the effects of stopping it early.

Original language	English (US)
Pages (from-to)	661-674
Number of pages	14
Journal	Biometrics
Volume	77
Issue number	2
Early online date	Jun 12 2020
DOIs	https://doi.org/10.1111/biom.13313
State	Published - Jun 24 2020

Bibliographical note

Funding Information:
We thank the AE and two anonymous reviewers for constructive suggestions. The work of TM was partially supported by NIH/NCI award P30-CA0077598 and the E-STAND Award from the Minnesota Office of Higher Education Spinal Cord Injury and Traumatic Brain Injury Grant Program. TM would also like to acknowledge the support of Medtronic Inc. in the form of a Faculty?Fellowship.

Funding Information:
We thank the AE and two anonymous reviewers for constructive suggestions. The work of TM was partially supported by NIH/NCI award P30‐CA0077598 and the E‐STAND Award from the Minnesota Office of Higher Education Spinal Cord Injury and Traumatic Brain Injury Grant Program. TM would also like to acknowledge the support of Medtronic Inc. in the form of a Faculty Fellowship.

Publisher Copyright:
© 2020 The International Biometric Society

Keywords

adaptive design
medical device
n-of-1 trial
personalized medicine
spinal-cord injury

Publisher link

10.1111/biom.13313

Find It

Find It at U of M Libraries

Cite this

@article{2865f38d7abb4b8e926bc1f4ca2c4b89,

title = "Batch Bayesian optimization design for optimizing a neurostimulator",

abstract = "Recently, spinal epidural neurostimulation is being considered for rehabilitation of persons suffering from partial spinal-cord injury. The neurostimulator must be programmed by a neurosurgeon, yet little work has been done to develop rigorous methods for optimally programming the device. We propose an adaptive design to efficiently optimize programming of the neurostimulator based on specified interim evaluations of patient reported preferences. Preferences for the eligible device configurations are estimated after each interim analysis through a conditionally autoregressive model that assumes preference for one configuration is related to preferences for neighboring configurations. Using the adaptively updated preferences, a group of configurations is programmed into the device for the patient to evaluate during the next follow-up period. This selection is based on a balance of device exploration and preference maximization. We repeat this process until a specified stopping rule or the calibration end is reached. We show simulation studies to evaluate the overall quality of the adaptive calibration for various configuration selection strategies and the effects of stopping it early.",

keywords = "adaptive design, medical device, n-of-1 trial, personalized medicine, spinal-cord injury",

author = "Adam Kaplan and Murray, {Thomas A.}",

note = "Funding Information: We thank the AE and two anonymous reviewers for constructive suggestions. The work of TM was partially supported by NIH/NCI award P30-CA0077598 and the E-STAND Award from the Minnesota Office of Higher Education Spinal Cord Injury and Traumatic Brain Injury Grant Program. TM would also like to acknowledge the support of Medtronic Inc. in the form of a Faculty?Fellowship. Funding Information: We thank the AE and two anonymous reviewers for constructive suggestions. The work of TM was partially supported by NIH/NCI award P30‐CA0077598 and the E‐STAND Award from the Minnesota Office of Higher Education Spinal Cord Injury and Traumatic Brain Injury Grant Program. TM would also like to acknowledge the support of Medtronic Inc. in the form of a Faculty Fellowship. Publisher Copyright: {\textcopyright} 2020 The International Biometric Society",

year = "2020",

month = jun,

day = "24",

doi = "10.1111/biom.13313",

language = "English (US)",

volume = "77",

pages = "661--674",

journal = "Biometrics",

issn = "0006-341X",

publisher = "Oxford University Press",

number = "2",

}

TY - JOUR

T1 - Batch Bayesian optimization design for optimizing a neurostimulator

AU - Kaplan, Adam

AU - Murray, Thomas A.

N1 - Funding Information: We thank the AE and two anonymous reviewers for constructive suggestions. The work of TM was partially supported by NIH/NCI award P30-CA0077598 and the E-STAND Award from the Minnesota Office of Higher Education Spinal Cord Injury and Traumatic Brain Injury Grant Program. TM would also like to acknowledge the support of Medtronic Inc. in the form of a Faculty?Fellowship. Funding Information: We thank the AE and two anonymous reviewers for constructive suggestions. The work of TM was partially supported by NIH/NCI award P30‐CA0077598 and the E‐STAND Award from the Minnesota Office of Higher Education Spinal Cord Injury and Traumatic Brain Injury Grant Program. TM would also like to acknowledge the support of Medtronic Inc. in the form of a Faculty Fellowship. Publisher Copyright: © 2020 The International Biometric Society

PY - 2020/6/24

Y1 - 2020/6/24

N2 - Recently, spinal epidural neurostimulation is being considered for rehabilitation of persons suffering from partial spinal-cord injury. The neurostimulator must be programmed by a neurosurgeon, yet little work has been done to develop rigorous methods for optimally programming the device. We propose an adaptive design to efficiently optimize programming of the neurostimulator based on specified interim evaluations of patient reported preferences. Preferences for the eligible device configurations are estimated after each interim analysis through a conditionally autoregressive model that assumes preference for one configuration is related to preferences for neighboring configurations. Using the adaptively updated preferences, a group of configurations is programmed into the device for the patient to evaluate during the next follow-up period. This selection is based on a balance of device exploration and preference maximization. We repeat this process until a specified stopping rule or the calibration end is reached. We show simulation studies to evaluate the overall quality of the adaptive calibration for various configuration selection strategies and the effects of stopping it early.

AB - Recently, spinal epidural neurostimulation is being considered for rehabilitation of persons suffering from partial spinal-cord injury. The neurostimulator must be programmed by a neurosurgeon, yet little work has been done to develop rigorous methods for optimally programming the device. We propose an adaptive design to efficiently optimize programming of the neurostimulator based on specified interim evaluations of patient reported preferences. Preferences for the eligible device configurations are estimated after each interim analysis through a conditionally autoregressive model that assumes preference for one configuration is related to preferences for neighboring configurations. Using the adaptively updated preferences, a group of configurations is programmed into the device for the patient to evaluate during the next follow-up period. This selection is based on a balance of device exploration and preference maximization. We repeat this process until a specified stopping rule or the calibration end is reached. We show simulation studies to evaluate the overall quality of the adaptive calibration for various configuration selection strategies and the effects of stopping it early.

KW - adaptive design

KW - medical device

KW - n-of-1 trial

KW - personalized medicine

KW - spinal-cord injury

UR - http://www.scopus.com/inward/record.url?scp=85087144173&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85087144173&partnerID=8YFLogxK

U2 - 10.1111/biom.13313

DO - 10.1111/biom.13313

M3 - Article

C2 - 32530495

AN - SCOPUS:85087144173

SN - 0006-341X

VL - 77

SP - 661

EP - 674

JO - Biometrics

JF - Biometrics

IS - 2

ER -

Batch Bayesian optimization design for optimizing a neurostimulator

Abstract

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this