Medical Imaging

[stealthpaladin]

This field relates pretty heavily to another I made in response to . Essentially medical imaging and generating scene graphs are quite the same thing. Usually.

When it comes to prototyping new medical devices, this board is an absolute Godsend. It has been prohibitively expensive to have work out the accuracy and consistency issues required for this field on commodity hardware. Good enough hardware hackers and engineers have had some availability for the last few years but as a developer, whew - this is really an open door.

I can now have real-time data streams process large amounts of data while spitting the buffer to disk, greatly pushing the latency thresh-hold. In a previous post I mentioned cheaper MRIs. New advancements in sensors are going crazy. Did you know there's a magnetic sensor stretch between two nanotubes that wiggles when it hits a tiny tiny field? This jiggles the nanotubes and provides you a reading. Fancy that.

The new breed of nano-sensors of all varieties will provide more data than we know what to do with. Without adaptable, powerful processing, most programmers have no entry point to harvest and cull this massive data. Admittedly I need to do more research on the Epiphany but I understand it's ability to let cores communicate and that's really all I needed to hear along with understanding the benefits of the new class of FPGAs in town.

Longer story short, amazing sensors need this companion to prototype on. The ability to provide a 3D video of organs, including the brain and heart, is certainly a task this board is capable of at higher than standard quality given a few new sensors which I can already buy online. Biorhythm monitors and similar devices can also incorporate deep analysis of more standard readings.

Perhaps I'll be typing from my brain cap soon =)

[Greg Zuro]

I am encouraged by your aggressive project ideas!

I see a bit of 'never need more than 640K' attitude on this forum. I'd like to have as many high-speed interconnects as possible including some SATA capability. I'd love to see it, but I think 10GBE is a stretch since the 1G is free from the Zynq, after all...

I have a near-term need to process 300fps 16bit video with latencies <20ms, so...

-Greg

[stealthpaladin]

Glad I can offer some encouragement! It's a really good time to be getting into these fields.

Looking at the cluster kit now up at http://shop.adapteva.com/collections/pa ... luster-kit I wonder if it would be possible to connect a Wishbone based bus to the "Epiphany Link" connection with 10GB/s.

At present I'm hoping to use the SD slot as a mid-buffer between RAM and local SAN. Given the current RAM restrictions, I do imagine that the machine will cap out after a while for real-time measurements. This is not a bad problem to have however! If you make any device using the available resources, you can customize the open hardware to your needs considerably afterward.

Here is a link to an article the likes of which are popping up all over: . There's alot of non-SQUID alternatives showing up in neuroscience, alot of this is driven by new material sciences and stuff found out in some of the particle accelerators and such.

Already it's known what could be done if we had cheap sensors and cheap processing of the brain's EMF.
and also These example are actually getting kinda old but they certainly illustrate potential of a coming software interface I'd personally like to be part of.


Neuroscience isn't the only place sensors are hitting. Really everywhere there's a sensor, better ones are coming. Heart EMFs are much easier to detect than brainwave EMFs. There is a push to leave behind ultrasound for more passive sensors as well. From the looks of it, hospitals really need some help on the cost margin of these devices and I think that time is just around the corner.

[aolofsson]

stealthpaladin,

Thanks for pulling together this forum! Medical imaging (ultrasound in particular) was one of the focus areas of the Adapteva when the company was founded in 2008, but it never worked out (combination of GPU/CPU technology momentum, design cycles, and risk averse OEM customers). It would be incredible if we could somehow disrupt the medical imaging industry with the Parallella device! The hardware is the easy part though..

One could envision different daughter cards plugging into the Parallella to create a very small foot print portable medical device (especially if the RJ45 ethernet connector is removed).

Andreas

[stealthpaladin]

Have thought through this some more and have come up with the idea to form standing-wave constructs and tessellates of them based on the size and use of each memory bank.

Then using some bitmasks and cascading through masking techniques similar to (and including) a bloom filter - I can operate on the differentials between real life signals and best-matched, normalized 'magnitudes' of memory-constructed waves.

Essentially this would boil down to binary division of the core grid, tiering themselves such that cores closest to the incoming stream will task inbound cores with the most complexity-bearing ranges of the signals. Additionally the signals propagate to their neighbors across a quadrature so that the epiphany can take advantage of North-South-East-West routing rules.

Does this seem like a viable approach? I'll be testing out as much as possible once the board comes out of pre-release but I've yet to test one out so any input would be greatly appreciated as I conceptualize the basic components of signal processing on the epiphany.

[brucewillym]

is there any relation between and medical imaging and pain imaging??