search

firasm / CEST

Analysis of studies
2 stars 0 forks source link

Fitting Andrews Mouse Brain at high resolution, higher B1 #13

Open firasm opened 9 years ago

firasm commented 9 years ago

From Andrew's email:

This is from a single pixel in the cortex. At first glance, the data looks much better than before, but I am also using a low resolution comparable to what Kim was using. I repeated with a variety of powers. Looks like 1 uT gave us the best results on that day.

Tried to fit this with multiple lorentzians

firasm commented 9 years ago

First attempt at fit (native method - as opposed to MCMC):

3 peaks:

fit

---Peak 1--- A=-1.6 a.u w0 = -0.04 ppm lw = 0.7 ppm shift = 2.03 a.u

---Peak 2--- A=-0.4 a.u w0 = -0.42 ppm lw = 9.82 ppm shift = 2.03 a.u

---Peak 3--- A=-0.19 a.u w0 = -3.73 ppm lw = 2.92 ppm shift = 2.03 a.u

firasm commented 9 years ago

I think we're in business... here are the constituent lorentzians...

constituents

DrSAR commented 9 years ago

what's peak 2 (the ultra-broad one)?

firasm wrote:

I think we're in business... here are the constituent lorentzians...

firasm commented 9 years ago

I think it's pixie-dust - there to make the whole thing fit better :-P

but seriously, look at issue #6 for a picture of what KDs constituent lorentzians looked like for a tumour. Unfortunately there was nothing that close to the water peak

andrewcyung commented 9 years ago

Could this second peak around water be related to MT? I seem to remember firas sketching out what a MT peak would look like at high power (a central peak swallowed up by a high baseline)

firasm commented 9 years ago

hmm good idea. Although, I seem to remember the MT peak being much much wider.

Similar to this:

screen shot 2015-02-03 at 10 09 42 pm

Sheth, V. R., Li, Y., Chen, L. Q., Howison, C. M., Flask, C. A., & Pagel, M. D. (2011). Measuring in vivo tumor pHe with CEST-FISP MRI. Magnetic Resonance in Medicine, 67(3), 760–768. doi:10.1002/mrm.23038

andrewcyung commented 9 years ago

... But maybe the width depends on the pulse power...?

firasm commented 9 years ago

Okay, just updating this thread with results from pixel-by-pixel maps of the CEST fits.

I've checked the fits for a few pixels, and all the fits look pretty good. Just to be clear, the maps below are generated using the least-squares fitting method where initial parameters are given and both the peak heights, widths, and peak positions are allowed to roam.

z-spectrum fits

The maps below include only pixels around the brain (non white). I'll look at things in more detail tomorrow, to find the reason why some fits are failing.

peak amplitudes peak widths peak locations peak integrals

andrewcyung commented 9 years ago

thanks firas! it's a good first start!

On 2/8/2015 1:40 AM, firasm wrote:

Okay, just updating this thread with results from pixel-by-pixel maps of the CEST fits.

I've checked the fits for a few pixels, and all the fits look pretty good. Just to be clear, the maps below are generated using the least-squares fitting method where initial parameters are given and both the peak heights, widths, and peak positions are allowed to roam.

z-spectrum fits https://cloud.githubusercontent.com/assets/2507459/6095921/5ea159ce-af29-11e4-8e2f-06f683425576.png

The maps below include only pixels around the brain (non white). I'll look at things in more detail tomorrow, to find the reason why some fits are failing.

peak amplitudes https://cloud.githubusercontent.com/assets/2507459/6096063/65fc363a-af33-11e4-902d-112421ee63a8.png peak widths https://cloud.githubusercontent.com/assets/2507459/6096066/661d8268-af33-11e4-8915-e9436c74b04d.png peak locations https://cloud.githubusercontent.com/assets/2507459/6096064/66196fc0-af33-11e4-99ae-802f89d4bfb6.png peak integrals https://cloud.githubusercontent.com/assets/2507459/6096065/661ceef2-af33-11e4-93c5-c9031400f01f.png

— Reply to this email directly or view it on GitHub https://github.com/firasm/analysis/issues/13#issuecomment-73403774.

firasm commented 9 years ago

These are from the MCMC fitting method - they look a bit better, but still hard to see structure. All integral maps are line width x amplitude

p2 map water peak map water map mt map p3 map p1 map

DrSAR commented 9 years ago

yes - a bit of a mess. A normal brain might not be an ideal playing ground to see regional heterogeneity. Also, one has to look carefully at B0 heterogeneities and other artefacts to make sure that any subtle variations in the P1, P2, P3 peak areas are not due to changes in what is, comparatively speaking, a huge baseline. But a good start nevertheless.

firasm wrote:

These are from the MCMC fitting method - they look a bit better, but still hard to see structure. All integral maps are line width x amplitude

p2 map https://cloud.githubusercontent.com/assets/2507459/6112769/44eec74a-b047-11e4-9b84-db202812c491.png water peak map https://cloud.githubusercontent.com/assets/2507459/6112772/450d49f4-b047-11e4-81a9-059454f84841.png water map https://cloud.githubusercontent.com/assets/2507459/6112775/451009b4-b047-11e4-82e0-96b958aecc18.png mt map https://cloud.githubusercontent.com/assets/2507459/6112773/450d8d10-b047-11e4-8f81-b45a572cfd3a.png p3 map https://cloud.githubusercontent.com/assets/2507459/6112771/450d2b4a-b047-11e4-887a-8dac461a45ee.png p1 map https://cloud.githubusercontent.com/assets/2507459/6112774/450dc2e4-b047-11e4-892f-367f8aa2a9bc.png

— Reply to this email directly or view it on GitHub https://github.com/firasm/analysis/issues/13#issuecomment-73563058.