| |
AUTO-TUNING CT NOISE REDUCTION ENGINE
Reducing the radiation dose delivered to patients has become a key consideration in selecting CT scanning protocols. Lowering the dose increases image noise. While radiologists can often "see through the noise", the images appear inferior to referring physicians and confound automated post-processing algorithms used for bone masking, CAD, 3D visualization, tumor volume measurement and others. While the most recent generation of CT scanners now offer noise reduction algorithms, these algorithms are not available with older scanners and involve a considerable expense of money and processing time. Claron's CT denoiser provides an alternative to scanner-based noise removal which can be easily and economically deployed centrally in a PACS environment, improving the appearance of all low dose CT image data, regardless of its source scanner.
FEATURES
The CT Denoiser engine is a structure-preserving noise removal module designed to be "plugged in" a host PACS or advanced-visualization software system. It may be activated either in batch mode with the arrival of a low-dose CT series to produce and save an additional denoised version of that series, or interactively to denoise images on-the-fly as they are presented to the user. In case of doubt, the user has the option of toggling between the original and the denoised version of the image to ensure integrity was preserved.
Compared to other noise removal algorithms, Claron's denoiser has some unique qualities:
- Self-tuning: the algorithm performs an analysis of the image data to extract the noise characteristics and tune its noise suppression parameters. This further allows the parameters to be adjusted across different regions the volume that demonstrate different noise levels or patterns. The use/host controls the amount of noise removal using a single "aggressiveness" parameter with a range of 0 to 1. At a setting of 0.5, the denoiser would remove the maximum amount of noise it can while ensuring all 3D structural information in the image remains intact.
- Powerful: can improve signal/noise ratio by up to 600% in ultra-low-dose studies.
- Fast: ~ 100 slices/sec on 8-cores 3GHz CPU.
- Natural looking. There is no need to mix the original images with the denoised one to remove blotchy appearance.
- Flexible: The algorithm does not need to access the full volume. It can operate effectively incrementally as slices are loaded.
Investigational Use Only
|
|
