Store 
User Stories
Filtering Tool Developed in MATLAB Improves the Reliability of Nuclear Medicine Imaging
Download this user story 256k
|
|
|
Thousands of physicians rely on nuclear images for diagnosing cancer and other serious diseases--according to the Society of Nuclear Medicine, more than 10 million nuclear medicine imaging and therapeutic procedures are performed each year in the US alone.
Unlike conventional radiology, which focuses on anatomy, nuclear medicine imaging documents both the organ structure and function. As a result, physicians can detect abnormalities in their very early stages, when there is the greatest chance of treating them successfully.
Nuclear images are often unreliable as diagnostic tools, however, because the quality of the image is degraded by quantum noise.
Dr. Ghada Jammal, of Darmstadt University of Technology in Germany, set out to increase diagnostic confidence in nuclear images by developing DeQuant, an image processing tool that significantly reduces quantum noise. Throughout this project, she relied on MATLAB and several MATLAB toolboxes.
Challenge
In nuclear medicine diagnostics, the patient receives small amounts of radioactive materials. As these materials decay, photons (particles of electromagnetic energy) are emitted from within the patient. A gamma camera captures these emissions, providing a map of the internal distribution of the administered radioactive material. The result is a nuclear image, or scintigram, that details both the organ's anatomy and function.
Practical limitations on imaging time and the amount of radioactivity that can be administered safely to patients result in nuclear medicine images exhibiting a low signal-to-noise ratio.
Jammal set out to reduce quantum noise in scintigrams by means of mathematical filtering methods. This involved developing an algorithm to decompose the image into its signal and noise components and to restore it once the noise components were eliminated.
She also decided to build an interface to enable physicians and radiologists to perform the image processing operations quickly and easily without having to go into the details of the underlying mathematics.
"When we began building DeQuant, I knew that I could rely on the algorithms in the Image Processing Toolbox. We were then able to concentrate on modifying and enhancing these algorithms rather than being bogged down in building image processing tools from scratch."Dr. Ghada Jammal
Technical University Darmstadt, Germany
Solution
Jammal had become familiar with MATLAB and the Signal Processing Toolbox through her engineering studies. She notes, "Although I was not specifically familiar with the Image Processing, Statistics, and Wavelet Toolboxes, I knew from my previous experience with MATLAB and other toolboxes that I would be able to immediately get up to speed with these products."
Beginning with the Image Processing Toolbox, her first goal was to obtain a representation of the image in a domain where salient information could be separated from noise. Using the Signal Processing and Wavelet Toolboxes, she developed an algorithm that transforms the information contained in the noisy scintigram into a small number of coefficients. She then used the Statistics Toolbox to analyze the statistical significance of the coefficients.
Jammal next used wavelet transform functions in the Wavelet Toolbox to reconstruct and enhance the image. The wavelet compression made the new images considerably smaller, enabling Jammal to easily archive the results.
For building DeQuant's graphical user interface, Jammal says that the GUIDE tool in MATLAB was "ideal." This interface enables the physician or radiologist to specify a region of interest in the image and apply the filtering algorithm to the region specified.
DeQuant has proven to be particularly effective in the detection of nodular thyroid disease. Statistical fluctuation introduced by quantum noise normally makes it extremely difficult for the physician to identify cancerous tissue from a thyroid scan. But DeQuant effectively reduced the quantum noise in thyroid scans while preserving the original data.
Results
- Streamlined algorithm development process. According to Jammal, "The extensible MATLAB platform and available functions in the form of M-files let us develop image processing solutions built on known and tested algorithms."
- Enhanced image quality. Accurate nuclear images are of particular importance in the detection of nodular thyroid disease. In one test, DeQuant increased the signal-to-noise ratio on a thyroid gland scintigram by 100%.
- Technology with broad applicability. DeQuant is primarily targeted towards gamma camera manufacturers and gamma camera software developers, but it can also be used in all applications where Poisson or quantum noise is present, including astronomy and communications.
For more information email: Technical University Darmstadt
