by November 5, 2002 0 comments



We all know that the image processing and rendering requirements of even elementary games are huge. That is why the graphics cards of today build in humungous amounts of their own RAM and even their own dedicated processors. But the volume of data that they have to process pale to insignificance when compared to the volume of data that is handled by a medical image-acquisition system.

Machines like CT or MRI scanners run multiple embedded systems

Unlike a gaming PC, where there is no separate image acquisition and where processing and display happen on the same unit, a medical image-acquisition system acquires the image at one point, and then transfers it to another point (for example, to a viewing station) for processing. And often, the acquisition and processing/viewing is real time, with the physician or technician adjusting the source, according to what is being viewed. Thus, a medical image-processing system should be capable of not only processing humungous amounts of image information, but also of transmitting the same amounts of information.

Let us take the example of a CT scanner. Lain Goddard, writing in the Medical Electronics Manufactring Magazine, points out
that early CT scanners of 1974 vintage produced an image of 6400 pixels every five to six minutes. Moving forward, by 1987, these machines could produce an image of a quarter million pixels every eight seconds. 

Computer inside

The computational platforms on some MRI (Magnetic Resonance Imaging) equipment is as below
Philips ACS-NT: DEC Alpha AXP 64 bit
Siemens Magnetom Impact SM 101: Sun SPARC 2 with 2.1 GB hard drive
Alkomed NAM P023A: Two computer systems connected over TCP/IP. The main one is a PIII 700 MHZ Windows NT workstation with 128 MB RAM

By 2000, multislice CT systems could do 4 slices simultaneously at 2 images per second, and a patient’s dataset could be in excess of over 150 MB. This translates to a processing requirement of over 8
GFLOPS. 

And today’s high-end scanners do 16 slices! 
Thus, the computing systems used in these machines have to match these processing and transfer requirements, which far exceed that of traditional PCs and workstations, even though they are classified as embedded systems, and many of them do not run anywhere near the gigahertz range of processor speeds that even entry-level PCs run.

Machines like a CT scanner run multiple embedded systems, with each doing a specific job. On top of this, there are specific microprocessors that control the movement of the gantry and the patient table.

Data that is acquired by systems like CT or MRI scanners is sent to processing equipment that are really workstations. Medical-equipment vendors have their own workstations. 

Storage options

The Siemens Somatom Sensation 16 is one of the most advanced CT scanners currently available. The data storage options on this machine include

Main storage: 73 GB (100,000 images)
Raw data capacity:
150 GB

Archiving options
1) CD-R: 1,100 images per CD
2) Optional Magneto-optical Drive: 5.2 GB (2.3/4.1 Gb cartridge, 4000/7500 images)
3) Optional Read Only Magneto-optic drive: 1.7 GB
Source: Product brochure

Plus, there are any number of third-party vendors, like SGI, who also offer workstations for medical-image processing.
Another area where IT equipment comes into play is in storage. Huge amounts of data is acquired by these imaging systems and it needs to be archived. Storage and archival systems start from hard disks and range through CD writers and different types of tape backup.

Many of these systems let service technicians access, diagnose and repair them remotely. Such remote access is through secure connections over the Internet, using ISDN modems and
the like.

Krishna Kumar

No Comments so far

Jump into a conversation

No Comments Yet!

You can be the one to start a conversation.

Your data will be safe!Your e-mail address will not be published. Also other data will not be shared with third person.