Raw data acquired by digital scanners can provide substantiallymore diagnostic information than is displayed on the film imagesthese machines produce. Magnetic resonance imaging systems, for instance, produce datawith a 12-bit dynamic range,
Raw data acquired by digital scanners can provide substantiallymore diagnostic information than is displayed on the film imagesthese machines produce.
Magnetic resonance imaging systems, for instance, produce datawith a 12-bit dynamic range, corresponding to about 4000 shadesof gray. Film cannot display as many intensity levels. Even iffilms could display that much information, the human eye couldnot distinguish the variations, said Ralph Bernstein, managerof the image science and applications department of the IBM ScientificCenter in Palo Alto, CA.
"What that means in practical terms is that there is information(in the digital image data) that may not be discerned," hesaid.
IBM's Palo Alto center has been engaged in applied researchin medical imaging for about five years. The group works closelyin this effort with both Stanford University and the MagneticResonance Science Center at the University of California at SanFrancisco, he said.
The MRSC, which opened about a year ago, processes image informationfrom a GE Signa high-field MRI scanner using IBM's 3090 supercomputer.IBM has supported this effort with research grants. Funding hasalso been received from Schering for the investigation of contrastimaging and spectroscopy, noted Dr. Alexander Margulis, centerdirector.
"We are a research lab that works with industry, but isnot owned by anybody," Margulis said.
While the IBM 3090 is a powerful mainframe computer with ahigh number-crunching capability, algorithms under developmentat the IBM Scientific Center could run on all types of computers,including PCs and the new IBM RS/6000 workstation. With modification,programs developed at the center could run other vendors' equipmentas well, Bernstein said.
Applications under investigation at the center include tissuecharacterization, three-dimensional presentations of data, andanimated sequence development to portray dynamic processes. Additionalinformation can be extracted from the raw data, which might, forexample, enhance the images of veins and arteries, he said.
The future of digital medical image processing appears bright,Bernstein said, particularly because many younger physicians havebeen raised on computers and are not intimidated by them. Theyrecognize that there is much diagnostic potential in the digitalimage information.
"The challenge is not to replace the radiologist, butto process the data so as to extract additional information orprovide a different visualization or characterization. (The radiologist)can then have both the conventional data and the computer-processeddata from which to make a judgment and diagnosis," Bernsteinsaid.
Mammography Study Compares False Positives Between AI and Radiologists in DBT Screening
May 8th 2025For DBT breast cancer screening, 47 percent of radiologist-only flagged false positives involved mass presentations whereas 40 percent of AI-only flagged false positive cases involved benign calcifications, according to research presented at the recent American Roentgen Ray Society (ARRS) conference.