PRESS RELEASE - STEREOMETRIX CORP.

goel at ug.eds.com goel at ug.eds.com
Tue Apr 25 18:40:38 EST 1995




STEREOMETRIX CORPORATION --------> Computer Aided Diagnosis for Medical Imaging
675 N. 1st Street
Suite 700
San Jose, CA 95112
Ph: (408)975-7660
FAX (408)975-7679

PRESS RELEASE					Media Contact: Bob Chapman



			  BREAKTHROUGH IN COMPUTER-AIDED
			BREAST CANCER DIAGNOSIS TECHNOLOGY

		Accuracy 95% With False Indications Reduced to Negligible Level


San Jose, April 26, 1995 - Stereometrix Corporation is in the final stage of
completing technology for early and accurate detection of breast cancer. A
recent breakthrough has resulted in a significant improvement in system
performance. False positive indications (like false alarms) have now been
reduced to near zero. False positives occur in both manual and computerized
x-ray reading from borderline cases and film artifacts. Dr. K.C. Saxena, Ph.D.,
Stereometrix's founder and developer of the technology, states that the updated
Mammogram Viewstation 5000 computer-aided-diagnosis system will be ready for
clinical trials in the near future. The company is presently Alpha testing
the new software which incorporates the breakthrough technology. This recent
development makes the system much more suitable for breast cancer mass
screening programs to combat this deadly disease.


Dr. Saxena, a well known photogrammetrist and mathematician, said that the
technology he developed over the past several years has been derived from a
combination of sciences including mathematics, statistics, image processing
and photogrammetry. The keys to success are the novel mathematical algorithms
and filters which are able to highlight, extract and classify small features
of clinical interest hidden or obliterated by other soft tissues, blood vessels
and densities present in mammogram x-rays. The classification algorithms are
derived from the criteria used by noted radiologists. They are able to
distinguish potential malignancies from benign features and reject film
artifacts and characteristics which resemble masses.


Some computerized mammographic interpretation techniques being investigated
have the possibility of losing elements of vital breast cancer information
on the x-ray. Dr. Saxena believes that such methods will not be able to reliably
detect or classify small size or early stage cancers. Stereometrix rejected
simplistic approaches early in its research and has concentrated on developing
a technology in which very obscure and small cancer-like features can be
detected. This has resulted in a system whose detection and location accuracy
significantly exceeds that achieved by other workers in this field. In addition,
the system and its user interface have been designed for efficient integration
into the clinical environment.


In preliminary testing, the ViewStation 5000's patented software technology
demonstrated a 95% accuracy rate in detecting pathologically confirmed
(biopsied) potentially malignant microcalcifications and masses on more than
400 mammograms. Detection accuracies for manual (visual) reading of the
x-rays typically range from 70% - 90%. The system is also capable of detecting
very small features - thus aiding earlier detection of breast cancer. The new
breakthrough discovery's reduction of false positives to an insignificant level
may also help reduce the number of unnecessary biopsies.


Breast cancer kills 46,000 American women every year and costs the U.S. health
care system $3.8 billion annually. Increased diagnostic accuracy is expected to
ultimately save lives and reduce medical costs.


Located in San Jose, California, Stereometrix Corporation develops advanced
image analysis technology software. Its present focus in medical applications
is to deliver increased diagnostic accuracy and improved patient outcomes.


					# # #

==============================================================================
Any e-mail enquiries resulting from this posting will be forwarded to
Stereometrix Corp. by this individual without any obligations to either
party.
==============================================================================




More information about the Acedb mailing list