INDUSTRY

IBM today announced that Electro-Optical Sciences will be using IBM’s deep computing on demand facility to access supercomputing power over the Internet, helping to free them from the fixed costs and management responsibility of owning a supercomputer. Electro-Optical Sciences (EOS) is the first example of a life sciences customer using deep computing on demand to help doctors make better diagnostic and treatment decisions in the emerging field of information-based medicine. EOS has created a computerized hand-held skin scanning device, MelaFind®, that can be used in a doctor’s office for early detection of malignant melanoma, the deadliest of the skin cancers. The only effective way to cure a patient from melanoma is to detect it early and to surgically excise it with the proper margins. MelaFind detects melanoma early, objectively, reliably, non-invasively, and automatically. A doctor or a nurse will use MelaFind to capture a series of multi-spectral digital images of a suspicious lesion on a patient’s skin in one second. These images are automatically processed with algorithms that reliably differentiate melanoma from benign look-alike lesions and then recommend whether the lesion should be biopsied or not. This entire process occurs while the patient is in the office and takes approximately one minute. “Development of these algorithms has been facilitated by the computational power offered by deep computing,” said Dr. Alex Bogdan, a principal developer of the algorithms at EOS. “We employed statistical pattern recognition methods to analyze our database of over 20,000 high resolution 1.5 MB lesion images and to select the best combination of image features for reliable detection of melanoma”. “The results of on-going clinical tests suggest that MelaFind’s computer vision can outperform even the most experienced dermatologists, who rely on human vision, in clinical detection of early melanoma,” said Dr. Marek Elbaum, EOS’ President and CEO. “Adding supercomputing power allows us to improve MelaFind’s performance by accelerating the process by which we train our algorithms on an ever-increasing database. We see the availability of supercomputing power as a facilitator for extending the applications of our photonic computer vision technology platform to the automatic and objective detection of other epithelial cancers (80% of all cancers).” Supercomputing on demand provides variable capacity, as they need it, from IBM's center located in Poughkeepsie, New York. "The on demand computing model brings supercomputing power directly into the doctor’s office, demonstrating the perfect synthesis of medicine and technology," said David Turek, vice president, IBM Deep Computing. "Deep computing allows doctors to focus their skills on what’s important: helping patients." With deep computing on demand customers can: * help avoid large IT capital outlays and long term depreciation or lease schedules --this is especially important for companies with short term projects * bring to bear, on a temporary basis, massive amounts of compute power that could be otherwise unaffordable and which helps provide strategic insight * help lower overall operating costs * help improve price/performance for compute-intensive applications and processing of massive amounts of data * off-load certain system maintenance to IBM * use the latest technologies to help maximize performance * meet the urgent computational needs of new business opportunities that could otherwise be cost-prohibitive The deep computing on demand facility is located in a highly secure section of IBM's Poughkeepsie, New York plant. The system consists of a cluster of IBM eServer xSeries Intel-based Linux systems with related disk storage, and is planned to include pSeries UNIX servers. Designed for scalability to meet increased demand, the deep computing on demand facility is also planned to incorporate a variety of blade technologies and AMD technologies over time. The service caters to a broad spectrum of companies that have peaks and valleys in their need for supercomputing power. These include Hollywood studios that use supercomputing power to create animated movies as well as petroleum companies for seismic imaging and research. Financial services organizations, government agencies and national research laboratories are also likely customers.