10
Computational requirements for Electrical Impedance Tomography of brain function David Holder Medical Physics, UCL Clinical Neurophysiology, UCH 15/06/22 1

Computational requirements for Electrical Impedance Tomography of brain function

  • Upload
    metea

  • View
    117

  • Download
    0

Embed Size (px)

DESCRIPTION

Computational requirements for Electrical Impedance Tomography of brain function. David Holder Medical Physics, UCL Clinical Neurophysiology, UCH. ResearchClinical. Clinical Neurophysiology, UCH. Medical Physics, UCL David Holder Kirill Aristovich James Avery Tugba Doru Hwan Koo - PowerPoint PPT Presentation

Citation preview

Page 1: Computational requirements for Electrical Impedance Tomography of brain function

Computational requirements for Electrical Impedance Tomography of brain function

David HolderMedical Physics, UCLClinical Neurophysiology, UCH

21/04/23 1

Page 3: Computational requirements for Electrical Impedance Tomography of brain function

Principal research :Electrical Impedance Tomography of brain function

• A new medical imaging method• Ring(s) of electrodes are placed around the subject• A tiny current is passed at ~50 kHz between various combinations of electrodes • Images of impedance are produced ~10 times per second

• Neuro applications :– Fast neural activity– Acute stroke– Epileptic seizures

321/04/23

Plane 3

Plane 2

Plane 1

Plane 4

xiphoid process

axilla

electrodes

1 2 3 4 5 6 7 80.5

1

1.5

2

2.5

3

3.5

Frequency in binary steps

OSE

COL

MIC

0 Frequency 1 MHz

Page 4: Computational requirements for Electrical Impedance Tomography of brain function

Current development

Perspex rod

Images :

actual EITEIT of brain function works in :

-Tanks

-Animal studies-Stroke-Epilepsy-Fast neural-Evoked responses

-Does not yet work in patients

Page 6: Computational requirements for Electrical Impedance Tomography of brain function

Recording protocol (illustrated for square wave)

+

=

Vol

tage

+

Vol

tage

Square wave

Evoked potential

Impedance change

I

Record V

500 ms

EP

Oh, T., Gilad, O., Ghosh, A., Schuettler, M., Holder, D. S. (2011). A novel method for recording neuronal depolarization with recording at 125-825 Hz: implications for imaging fast neural activity in the brain with electrical impedance tomography. Med Biol Eng Comput 49(5), 593-604

1

2

3

4

5

6

7

8

9

11

10

12

14

13

15

17

16

19

18

23

22

21

20

25

24

26

27

28

29

Page 7: Computational requirements for Electrical Impedance Tomography of brain function

Imaging

I

30x

~40 min

1min(120 av)

1min(120 av)

1min(120 av)

1M element rat brain FEM mesh Image ~300um, τ 8 ms reconstruction

1

2

3

4

5

6

7

8

9

11

10

12

14

13

15

17

16

19

18

23

22

21

20

25

24

26

28

29

0.2mm, 662,788 elements

Adaptive, 359,797 elements

0.8mm, 49,973 elements

Page 8: Computational requirements for Electrical Impedance Tomography of brain function

Forward Modelling• Forward solution:

• Computational complexity:

• O(N); O(NlogN);

• 2.106 - 100.106 FEM mesh→ ~ 10-100 Gb Memory ~ 1-30 Gigaflops ~ 10min - 4 hours

Computational considerations

Page 9: Computational requirements for Electrical Impedance Tomography of brain function

Image reconstruction

• Validation on saline tank phantom using 75k-element cylindrical mesh

• Total variation (TV) regularization:

Page 10: Computational requirements for Electrical Impedance Tomography of brain function

Projection for future

• Before optimization :– Practical

• ? 10.106 FEM mesh

→ ~ 100 Gb Memory~ 100 Gigaflops~ 2 days per image

– Ideal• ? 100.106 FEM mesh

→ ~ 2 Tb Memory~ 1 Teraflop ~ 20 days per image

• Optimized• ? 106 FEM mesh

→ ~ 10 Gb Memory~ 1 Gigaflop~ 10s per image

150ms

1mm

1.6m

m2.

2mm

2.8m

m3.

4mm

4mm

155ms 160ms 165ms 170ms

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04