Post on 05-Jul-2020
HARDWARE-IN-THE-LOOP SIMULATION AND ENERGY OPTIMIZATION OF CARDIAC PACEMAKERS
Nicola PaolettiDepartment of Computer Science, University of Oxford
Joint work with Chris Barker, Marta Kwiatkowska, Alexandru Mereacre and Andrea Patane’
EMBC 2015 – Milano – 28 Aug 15
MOTIVATION
• Cardiac pacemakers maintain a “correct” heart rhythm by sensing and stimulating heart beats
• One of the most common surgery procedures• Safety-critical
© Mayo Foundation for Medical Education and Research
Rigourous design methods needed for PM safety• Failures lead to device recalls, patient death
Energy efficiency• Battery depletion à re-implantation• How can we make it better (patient-specific)?
MOTIVATION
HW/SW co-design methods are required for energy optimization
• Models are not enough: need for real-time actual energy consumption data
• Hardware is not enough: need for (personalized) heart models to reproduce physiological conditions and verify safety
© Mayo Foundation for Medical Education and Research
MOTIVATION
HW/SW co-design methods are required for energy optimization
• Models are not enough: need for real-time actual energy consumption data
• Hardware is not enough: need for (personalized) heart models to reproduce physiological conditions and verify safety
Solution:HARDWARE-IN-THE-LOOP (HIL) SIMULATION Model
simulation
Execution onhardware
HIL ENERGY OPTIMISATION
COMPUTER
Heart ModelOptimization Algorithm(Gaussian Process Optimisation)
Online Energy model
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAVI
(s)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 108
MICROCONTROLLER
PacemakerModel
POWER MONITOR
Energy measurements
Energy readings
New parameters
Sensing andpacing
HEART AND PACEMAKER MODELS
PACEMAKER MODEL
LRI
VRP
AVI
PVARP
URI
Aget
Vget
AP
VP
AS
VS
AS
VS
AgetVP
VPVget
VP
VS
VSAPAS
VPVS
AP
VPCode
generation
HEART MODEL
Ventricle
Pacemaker
Atrium
SANode
AVJOut
AVJ
AP
Abeat
VgetAget
VP
VbeatAbeatAEctopic
AAVConductor AVJAnteIn AVJRetroIn
AVJAnteOutAVJRetroOut
AtrRetroIn
AtrAnteOutVtrAnteIn
VtrRetroOutVEctopic
AAVRetroIn AVVAnteIn
AVVConductor
RESULTSPacemaker parameters:• TLRI: (affects the) time the PM waits
before pacing atrium• TAVI: conduction time from atrium to
ventricle (affects the pacing of ventricle)
• Total electrical current during 10 sec HIL simulation
• 150 samples• Penalty (105 A) to parameters
yielding unsafe heart rates
RESULTSPacemaker parameters:• TLRI: (affects the) time the PM waits
before pacing atrium• TAVI: conduction time from atrium to
ventricle (affects the pacing of ventricle)
AV block: conduction defect in the AV node
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAV
I (s)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 105 AMean
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAV
I (s)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5x 104
A
Standard deviation
Best(- 1.64% default)
High uncertaintyDefault
• Total electrical current during 10 sec HIL simulation
• 150 samples• Penalty (105 A) to parameters
yielding unsafe heart rates
RESULTSPacemaker parameters:• TLRI: (affects the) time the PM waits
before pacing atrium• TAVI: conduction time from atrium to
ventricle (affects the pacing of ventricle)
AV block: conduction defect in the AV node
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAV
I (s)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 105 AMean
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAV
I (s)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5x 104
A
Standard deviation
Best Low uncertaintyClose to best Default
• Total electrical current during 10 sec HIL simulation
• 150 samples• Penalty (105 A) to parameters
yielding unsafe heart rates
RESULTSPacemaker parameters:• TLRI: (affects the) time the PM waits
before pacing atrium• TAVI: conduction time from atrium to
ventricle (affects the pacing of ventricle)
Bradycardia: slow heart rate
• Total electrical current during 10 sec HIL simulation
• 150 samples• Penalty (105 A) to parameters
yielding unsafe heart rates
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAV
I (s)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 105
A
Mean
0 0.5 1 1.5 2 2.50
0.5
1
1.5
2
2.5
TLRI (s)
TAV
I (s)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5x 104
A
Standard deviation
Best(-0.85% default) Low uncertaintyDefaultClose to best
Summary• Framework for energy optimisation of cardiac pacemakers• Hardware-in-the-loop: combines simulation on computer and execution on HW• Based on rigorous design methods (Stateflow/Hybrid Automata) • Safe and efficient parameters + Predictive energy model
Summary
Now…a quick demonstration!
• Framework for energy optimisation of cardiac pacemakers• Hardware-in-the-loop: combines simulation on computer and execution on HW• Based on rigorous design methods (Stateflow/Hybrid Automata) • Safe and efficient parameters + Predictive energy model