GPU Conference Presentations

I mentioned previously about a GPU conference that discussed the implications of the technology for doing scientific research. There are certain classes of problems that GPU's are especially suited for and they offer a speed up when compared to CPU's. As an example, researchers have recently developed a relatively inexpensive 13 gpu "supercomputer" with about 12 teraflops of computing power for scientific problems. GPU's have been rivaling the complexity of intel's most advanced technology and 3 billion transistor gpu chips will probably hit the market shortly. Nvidia also believes that they can reach 10 billion transistors easily. With this speed up of processing power coupled with machine learning, we will be able to learn more about the brain than ever before. While it's probably somewhat facile to make a blanket statement that computing power is increasing exponentially, there are still some interesting exponential trends in the field that will likely continue for at least the next 5 or ten years.

Nvidia has put media from that 2009 conference online and several of them are related to neuroscience. The company Evolved Machines is "pioneering the reverse engineering of brain circuitry to build intelligent machines". An audio talk can be found here (6.1 MB).

"Reconstructing the Brain: Extracting Neural Circuitry with CUDA and MPI" is a 37.6 MB video presentation (download here). The following is an excerpt about that video;

In this talk we will present our insights and lessons learned in using CUDA to reconstruct neural connections in high-resolution EM data. We will present technical details and non-trivial issues regarding the implementation of NeuroTrace, our system for semi-automatic segmentation and interactive visualization of terabytes of EM image data. The segmentation method is based on a sequence of 2D level set segmentations of cell membranes integrated with an image correspondence energy for robust transition between consecutive slices and a weighted path extrapolation method to trace a 3D centerline of a neural pathway along non-axis aligned slices.

Optimizing Ion Channel Kinetics Using A Massively Parallel Genetic Algorithm on the GPU (26.4 MB video presentation);

Voltage-gated ion channels effect the integration of information in many neurons. Some neurons express over 10 voltage-gated channels that turn information processing into a highly non-linear affair.

The currently popular analysis techniques suffer from various shortcomings that limit the ability of the researcher to rapidly produce physiologically relevant models of voltage-gated ion channels.

To solve this computational bottleneck we have been converting our optimization algorithm to work on a GPU using CUDA. We have succeeded to parallelize the process on a GTX 295 giving a speed increase of roughly X100 over that of the CPU.

Medical Image Registration with CUDA (37.6 MB video presentation);

Speedups of up to 750 times were obtained as compared to code in daily use at Addenbrookes Hospital and Bio-Medical Campus. Some very recent results are shown in the figures. This work is of direct application in both research and clinical practice. A particular application is voxel based MRI morphometry in humans and in animal brains.

High-Throughput Science (keynote speech);

How did the universe start? How is the brain wired? How does matter interact at the quantum level? These are some of the great scientific challenges of our times, and answering them requires bigger scientific instruments, increasingly precise imaging equipment and ever-more complex computer simulations.

The rest of the presentations can be found here. They cover a wide variety of topics.

Virtual Whole Body Simulations for Personalized Healthcare

The virtual physiological human (VPH) initiative is another project that is related to NeuGrid. Researchers are aiming to develop better computer simulations of the human body. This could potentially allow for personalized medicine, with tailored therapies for each individual.

There is a ton promising biotechnology in the pipeline that has really taken a long time to come to fruition. I think there has been considerable difficulty in translating the research of stem cells, gene therapy and rna interference into approvable therapies. With pharmaceutical drugs, much of the low hanging fruit has already been picked. New drug approvals for 2009 were flat compared to previous years, even though we probably know more about human biochemistry than ever before. Gene therapy has been around for a long time, but just hasn't yet panned out too well in the way of usable treatments. A main problem is that it can be tough to perform successful clinical trials. A lot of diseases are multi-faceted as well, which adds another layer of complexity. So in the short term I think it may become even more difficult to get an FDA nod for specific treatments. However, in the long term I am optimistic that computer simulations of the human body could speed up the approval process.

Below are a few of the goals of this extensive undertaking.

a) Development of patient-specific computer based models and simulation of the physiology of human organs and pathologies.

b) Development of ICT tools, services and specialized infrastructure for the biomedical researchers to support at least two of the following three activities: i) to share data and knowledge needed for a new integrative research approach in medicine (biomedical informatics), ii) to share or jointly develop multiscale models and simulators, iii) to create collaborative environments supporting this highly multidisciplinary field.

Under the umbrella of the VPH initiative there are many specific projects that scientists are working on.

The SurgAid project has been conceived to develop and apply new methods for diagnosis and support in mitral valve (MV) surgery repair procedures, based on the combined analysis and integration of the FEM approach with the advanced processing of real-time 3D echocardiographic images.

Cancer is another area that could witness improvements in outcomes.

HAMAM – Highly Accurate Breast Cancer Diagnosis through Integration of Biological Knowledge, Novel Imaging Modalities, and Modelling - is a three year project that started in September 2008.

The above information comes from a recent newsletter january (2010) (PDF). Another newsletter from last year discusses a little more about the project (see PDF file). Basically they mention that we need more computing power.

PCs has been advancing steadily since decades, but even the most recent central processing units (CPUs) are far from being able to follow the dynamics of an average protein, with atomic detail, for milli- or even microseconds of simulated time: this is the “scale gap” between the molecular and the biological macroscales. The mission of GPUGRID at IMIM, part of the VPH NoE Toolkit, is to provide members of the VPH with the tools to bridge this gap.

A roadmap of the VPH can be found here (PDF). Another article on it can be found here.

Grid computing and health (PDF)

NeuGrid: Brain Imaging Infrastructure for Defeating Neurodegenerative Diseases

There is a European project called NeuGrid. The main intent of this undertaking is to improve the treatment of neurodegenerative disorders such as Alzhiemer's. Neuroscientists are begining to collect a large amount of data from brain scans about specific diseases. Both Henry Markram and Ray Kurzweil have mentioned about certain trends in number crunching and how they can be applied to accelerating progress on these fronts. The information that we are gathering is far too great for any one person to learn even a very tiny fraction of. However new software and technology may be able to increase our understanding of these processes beyond what any single human intelligence could possibly comprehend.

Grid computing architectures should enable scientists to collaborate and share data faster than ever before. Computing resources can be spread across multiple research areas. Avenues that are being scaled up include;

1. On demand computing
2. Secure data sharing
3. dynamic data analysis

Other projects related to brain research could also be accelerated;

In silico drug discovery

Grid-enabled virtual screening

Impact of mutations on existing drugs

Homology modeling

See this PDF file for more information. This page has the ultimate goals of NeuGrid.

neuGRID aims to become the "Google for Brain Imaging", providing a centrally-managed, easy-to-use set of tools with which scientists can perform analyses and collaborate.

A new video about the NeuGrid project is below;


Older videos on NeuGrid are here and here.

I think the most interesting question pertaining to this is whether we can use this explosive growth in the understanding of our own biochemistry to actually defeat the aging process itself. Doing this is obviously considerably more difficult than merely slowing the progression of currently disabling disease states. However, I think we are moving closer to this seemingly improbable end point.

Brain Chip to Restore Functioning from Damage

The ReNaChip project is developing electronic biomimetic technology that could serve to replace damaged or missing brain tissue. This is basically neuromorphic engineering that seeks to mimic how neurons function. In the future this may be useful for people who have had injuries due to stroke or other illnesses. There are numerous obstacles to getting this tech off the ground. Having the microchip interface properly with the surrounding neural tissue is one issue that could be difficult to circumvent. It is also unclear if some of the models used actually represent specific regions of the mind accurately enough for this to work properly.

This page gives an overview of this undertaking;

The objective of this project is to develop a full biohybrid rehabilitation and substitution methodology; replacing the aged cerebellar brain circuit with a biomimetic chip bidirectionally interfaced to the inputs and outputs of the system. Information processing will interface with the cerebellum to actuate a normal, real-time functional behavioural recovery, providing a proof-of-concept test for the functional rehabilitation of more complex neuronal systems.

More information can be found at this page.

Experiments are carried out with two different types of stimuli; a tone which serves a conditioned stimulus producing no naïve response and an aversive puff to the eye (unconditioned stimulus) resulting in a naïve eyeblink response. The tone always precedes the airpuff in the course of the experiment.

A recent article about it is here.

The project aims include making a computer model of a well-defined brain pathway as proof of concept for the replacement of more complex brain circuits. Implementation of this model in a microchip will be used to create a biohybrid in which a lost behaviour is restored

This abstract discusses about interfacing with neuronal cells.

One of the major goals of the ReNaChip research program was to develop implantable electrodes with a very small size sensing pads. In this talk we will describe the process related issues of such electrodes and their principle of operation.

The researcher Ed Boyden has recently called for the creation of an "exocortex" to augment human abilities. It seems conceivable that a device implanted on top of the head could be used to increase intelligence or other traits. The exocortex would have to communicate with actual brain cells in some fashion. Perhaps this could be done using optogenetics or ultrasound pulses. Neither are approved for human disorders yet. Optogenetics would be more invasive but it is also much more selective in its ability to activate neuron subpopulations. I've been somewhat skeptical about whether it would actually get an FDA nod any time soon, however. Ultrasonic neuromodulation does not currently have the same targeting accuracy, but it would not require a person's skull to be breached. A sophisticated exocortex could potentially allow a two way communication between the external apparatus and the mind. The contraption could essentially scale up the amount of neurons in your brain by an artificial means. Most likely it would be used to improved the disabled first, with other applications being more speculative possibilities.

Page of a researcher who is working on the ReNaChip project.

An older article about Renachip is here.