NCTR Research Highlights

NCTR Employees' Invention Subject of Patent Application — Pattern Recognition Technology

The National Institutes of Health Office of Technology Transfer has contracted a Chicago firm to represent the intellectual property of a patent application led by two National Center for Toxicological Research (NCTR) employees, Drs. Dan Buzato and Jon Wilkes. The provisional patent (Serial No. 61/261.170) was filed with the U.S. Patent Office on November 13, 2009. This patent application represents the unique application of mathematical processes in a model to discriminate up to nine categories of brain tissue in Magnetic Resonance Spectroscopy (MRS) images with 96% accuracy. The model was designed as a tool to aid physicians with medical decisions based upon MRS images. Under current practice, the only option for a physician to substantiate a diagnosis, determine a prognosis, or follow the course of a disease is to perform a surgical biopsy. The nine categories of brain tissue conditions modeled were: necrosis, metastasized carcinoma, metastasized melanoma, demyelination, astrocytoma, oligodendroma, menigioma, glioblastoma multiforme, and ganglioma.

This work is the result of the NCTR researcher investigators' efforts to apply pattern-recognition mathematics to interpret medical-imaging technologies traditionally interpreted visually by physicians. The pattern-recognition technology initially was used by the investigators to develop rapid-detection assays in FDA counter-bioterrorism and food-safety efforts. The challenge for the imaging project was to develop a tool that could be used with the relatively low-resolution instruments commonly encountered in hospitals and clinics that operate with variances of hardware and/or other uncontrolled environmental factors.

The model uses distinct processing steps to transform variable spectroscopy into a set of defined conditions (biomarkers) for consistent interpretation:

Linear recalibration of spectra using two recognizable reference peaks
Enhancement of the most significant (reliable) peaks in spectra using a variable weighting process producing significant improvements of signal-to-noise ratios
Application of pattern recognition technology to the resulting processed spectrum to define signal sets that are characteristic of the different tissue conditions
The improved MRS spectrum processing steps hold promise to aid physicians in the practice of medicine. The process could reduce the need for surgical procedures, in itself improving the quality-of-life of the patient in all outcomes. Drs. Buzato and Wilkes also plan to study further application of pattern-recognition technology to different tissue tumors and disease types.

abrir aquí:
NCTR Research Highlights