An increasing number of radiology facilities are giving real-time speech recognition
technology a try, aided by newer, more accurate software and radiologys smaller,
specialized vocabulary.
Anyone who has tried dictating a letter straight into
a desktop computer using speech recognition software knows the frustration of fixing
garbled words. Some stick with it and achieve accuracy that makes speech recognition worth
the effort. For others, editing mistakes remains too time-consuming, and the software sits
idle.
Still, an increasing number of radiology facilities are giving real-time speech
recognition a try, aided by newer, more accurate software and radiologys smaller,
specialized vocabulary. Everyone seems to agree that dictating reports directly to text
speeds report turnaround time and saves money. While some radiologists are embracing the
change, others say speech recognition advantages do not outweigh the cost to radiologists.
The use of speech recognition technology will more than double in the next two years,
climbing from 20 percent to 46 percent, according to the 13th annual Health Information
and Management Systems Society (HIMSS of Chicago, Ill.) Leadership Survey of 355
information technology executives sponsored by Superior Consultant Co. Inc. (Southfield,
Mich.).
Manufacturers say the market is already growing quickly. Now that the benefits of
speech recognition are documented, people are becoming more interested and sales are
beginning to take off. Manufacturers estimate that 400 to 500 facilities in the United
States are using speech recognition, or about 5 percent of the market. Manufacturers also
say the global market, particularly in Europe and Asia, is growing quickly as well, and
products are available in most European languages.
Bringing speech recognition technology to radiology facilities is a melding of a number
of products. One component is the speech engine, which captures the sound of the voice and
translates it to text. It learns to adjust to individual pronunciation so a radiologist
from Boston and a radiologist from Dallas can both be understood. The engine uses a
specific vocabulary from which it recognizes words and analysis algorithms that process
the context of words so that it knows when a colon is the punctuation symbol and when it
is the organ.
Please refer to the November 2002
issue for the complete story.
For information on article reprints, contact
Martin St. Denis