Sunday, July 17, 2011

Basics 1: The Doppler-Effect

SLDF measures retinal blood flow. But what are the physical effects behind the scenes? Most know at least that this method uses the Doppler-Effect (ok, its part of the acronym :-) : The light that is reflected or scattered by moving blood cells gets a frequency shift. The effect everyone knows from a car passing by (

Figure 1 shows the interfering high fequency light and the resulting beat at low frequency. This is what the photo diode of the HRF camera records and what is saved in the SER-files: 256*64 recordings, each of ~2 seconds duration. You can see the time measurement either if you load a SER file by the HRF software or load it into SDLF.

Figure 1: Two waves of slightly different wave length (incoming and reflected light) interfering; frequency shift is caused by the moving blood cell

Figure 2: Same scenario, but the cell moving faster resulting in a higher frequency shift

The algorithms in HRF or SLDF analysis software again take the time signals of each pixel recording and determine the different frequency shifts. The distribution and magnitude of the frequency shifts are equivalent to distibution of blood cells moving at different velocities.

Sunday, December 14, 2008

Simple Validation of non-Flow

In the early phase of HRF my predecessor at the hostpital and me validated the SLDF method with quite simple experiments.

With one of the first (and easiest) experiments we also tried to find the acceptable range of DC, which is the intensity of an image.

I did an examination of a drawing paper, mounted in front of the HRF camera. In the DC image below you can also see the grid.

Figure 1: DC image of a drawing paper SLDF image

As we expected, the resulting Flow image below shows no Flow (figure 2, histogram in figure 3)

Figure 2: Flow and averaged Flow image of a drawing paper SLDF image

Figure 3:Histogram of averaged Flow image of a drawing paper SLDF image

This is a simple proof that the method works properly for areas with no moving particles. But it is getting even more interesting if we look at a scatterplot of DC-Flow values. This plot shows that the Flow values do not depend on the DC values as we expect - except for DC values above 225[AU]. These Flow values are plotted as blue points.

The overexposure leads to wrong Flow values, not correlated to DC anymore. The same effect we also observe for underexposed pixels and images. That's why these DC values are excluded from all calculations. The valid DC range can be adjusted using the settings dialog. But be aware: If you extend the range you will get wrong perfusion parameters and the vessel detection algorithms will produce wrong results.

Figure 4 DC-Flow scatter plot for a drawing paper SLDF image

Monday, November 10, 2008

What the hell AFFPIA stands for?

In one of his first articles Dr. Michelson invented the acronym AFFPIA to distinguish the HRF image analysis software from the measurement technique: AFFPIA stands for "automated full field perfusion image analysis".

But for me and many others AFFPIA sounds a bit strange and therefore the initial name "SLDF" has been coined.

Saturday, November 8, 2008

Paradox 7.0 Installation on XP

SLDF uses Paradox as database backend. Although you can export the SLDF database to dBase for further data analysis, doing queries on the original database can be done quite easy with Paradox.
Unfortunately, installing Paradox 7.0 is not possible on Windows XP PCs (Error "Not enough disk space") without using a loophole: You just have to change the TEMP environment variable so that it points to a disk drive with less than 2GB. You can do that in a command window ("set TEMP=E:\"). Then run the setup from this command window - that's it.
The root cause for the installation problem is an error in the calculation algorithm the installation uses. The installer has been written in times where nobody thought that a normal hard disk would store 2 or more GB. Thus, the calculation fails with free disk space above 2 GB (=2^31-1) due to a simple integer number overflow.

Friday, November 7, 2008

Vessel Wall Analysis improvement

Currently I'm working on an improvement for the Vessel Wall Analysis (for Prof. Schmieder/Med 4, Erlangen).

The goal is to overcome the granularity of 10µm of the DC diameter calculation. The screen shots below show the difference between the old and new (first) algorithm. The DC and Flow cross sections are interpolated to achieve sub-pixel resolution (remember: the pixel resolution is 10µm * 10µm at a record angle of 10°).

Figure 1: Current algorithm for computation of DC and Flow diameter

Figure 2: New algorithm for computation of DC and Flow diameter
Reason for this improvement was the fact that the DC diameter varies in 10µm steps whereas the Flow diameter varies in 1µm steps. With that the Wall Thickness varies by leaps and bounds!

Tuesday, November 4, 2008

SLDF now exists more than 10 years and I thought it's time to offer a discussion platform for all users of this wonderful measuring instrument.

So, I invite you to ask your questions, report your findings or - what's the main purpose - discuss.

Some words about the past of SLDF: In 1995 I started the development of a software for measurement data made by the Heidelberg Engineering Retina Flowmeter HRF.

Making a new software was necessary because the original software by HE was only a prototype with the ability to make these great 2D images of the retinal perfusion. But the software did not provide a reasonable way to analyse the data. The 10x10 (or bigger) moving rectangle analysis method was error-prone and far away from being observer-independent or reliable.

So I started implementing algorithms that provide reliable results. A colleague, Istvan Pal, provided some basic parts of the vessel recognition which I adapted and improved. And my colleague Dr. Harazny - Diploma Biologist - was (and still is) perfect in making good HRF images and could (and still can) do also complex database queries plus statistics in the twinkling of an eye :-) She also helped to improve the user interface as she was "my beta tester" sitting at the next desk. By the way, that was my first experience in "extreme programming".

Together we could evolve the algorithms and add more and more ways to analyse the HRF images.

Today SLDF can

  • analyse flow/pulsatility/cumulative frequence of RIM, temporal and nasal regions (plus a freehand aread)

  • perform vessel cross-section analysis

  • create scatter plots

  • create time plots, histograms, co-occurence plots, cumulative frequency plots

  • perform vessel wall thicknes analysis (an extension I made for Prof. Dr. Schmieder/Erlangen)

  • finally also visualize the results on a printout

For installation CDs (standard and extended versions) please contact Prof. Dr. R.Schmieder (