6. brain necstwork: implementation

16
Brain NECSTwork Eleonora D'Arnese [email protected] limi.it Enrico Reggiani [email protected] limi.it Marco Gucciardi [email protected] imi.it image from http://i1-news.softpedia-static.com/images/news2/The-Brain-Super-Sized-Computer-Going-from-Internet-to-Fiber- Optics-2.jpg

Upload: brain-necstwork

Post on 16-Jan-2017

67 views

Category:

Engineering


2 download

TRANSCRIPT

Page 1: 6. Brain NECSTwork: Implementation

Brain NECSTworkEleonora D'Arnese [email protected] Enrico [email protected] Gucciardi [email protected]

image from http://i1-news.softpedia-static.com/images/news2/The-Brain-Super-Sized-Computer-Going-from-Internet-to-Fiber-Optics-2.jpg

Page 2: 6. Brain NECSTwork: Implementation

Images are taken from fMRI

Linearization of the images

Pearson’s Correlation Coefficient

Images with specific colored areas

Reconstruction of images

Brain NECSTwork

2

Page 3: 6. Brain NECSTwork: Implementation

3

Images Acquisition

O2

More active areas of the brain receive more oxygenated blood

Blood Oxygen-Level Dependent (BOLD) signal

Images taken from functional Magnetic Resonance Imaging

Page 4: 6. Brain NECSTwork: Implementation

Images are taken from fMRI

Linearization of the images

Pearson’s Correlation Coefficient

Images with specific colored areas

Reconstruction of images

Brain NECSTwork

4

Page 5: 6. Brain NECSTwork: Implementation

Image Linearization

We used Matlab to acquire the images:

Fast (14 sec on GPU/200 img)

Extracts the matrix from the images easily

5

Page 6: 6. Brain NECSTwork: Implementation

Image Linearization

We used Matlab to acquire the images:

Fast (14 sec on GPU/200 img)

Extracts the matrix from the images easily

…Images are now ready to be analysed!

5

Page 7: 6. Brain NECSTwork: Implementation

Images are taken from fMRI

Linearization of the images

Pearson’s Correlation Coefficient

Images with specific colored areas

Reconstruction of images

Brain NECSTwork

6

Page 8: 6. Brain NECSTwork: Implementation

7

Hardware Implementation

Hardware Design is created and FPGA can be programmed

PCC IP core is synthetized

Pearson’s Correlation Coefficient (PCC) C code implementation

Page 9: 6. Brain NECSTwork: Implementation

7

Hardware Implementation

Pearson’s Correlation Coefficient (PCC) C code implementation

𝑟=∑𝑖=1

𝑛

(𝑥 𝑖−𝑥)(𝑦 𝑖−𝑦 )

√∑𝑖=1𝑛

(𝑥𝑖−𝑥 )2∑𝑖=1

𝑛

( 𝑦 𝑖− 𝑦 )2

r is the index which represents the value of the Pearson’s Correlation:with x, y selected pixels and mean values

Page 10: 6. Brain NECSTwork: Implementation

Hardware Implementation

PCC IP core is synthetized

Pearson’s Correlation Coefficient (PCC) C code implementation

Obtainment of a high level code from the PCC C implementation by Vivado HLS

PCC IP core can be used for the hardware implementation

7

Page 11: 6. Brain NECSTwork: Implementation

Hardware Implementation

Hardware Design is created and FPGA can be programmed

PCC IP core is synthetized

Pearson’s Correlation Coefficient (PCC) C code implementation

7

Page 12: 6. Brain NECSTwork: Implementation

Images are taken from fMRI

Linearization of the images

Pearson’s Correlation Coefficient

Images with specific colored areas

Reconstruction of images

Brain NECSTwork

8

Page 13: 6. Brain NECSTwork: Implementation

Colored areas

Correlation shown by colored areas of the brain

C code will consider the average response to a stimulus

Activated areas will be highlighted depending on a time threshold

C code OpenCL on GPU

9

Page 14: 6. Brain NECSTwork: Implementation

Colored areas

Correlation shown by colored areas of the brain

C code will consider the average response to a stimulus

Activated areas will be highlighted depending on a time threshold

C code OpenCL on GPU

…But we are still working on this!

www.familyhappening.it

9

Page 15: 6. Brain NECSTwork: Implementation

Images are taken from fMRI

Linearization of the images

Pearson’s Correlation Coefficient

Images with specific colored areas

Reconstruction of images

Brain NECSTwork

10