Oleg S. Pianykh, Presenter: Nothing to Disclose

CT radiation reduction, made possible by denoising filters, can greatly enhance diagnostic quality of the low-dose scans. However, applying the same filter to lungs, bones or soft tissues produces very suboptimal results, therefore several different organ-specific filters are commonly needed. We designed a single unified CT-denoising filter, applying different filtering strategies depending on the underlying CT densities.

CT denoising can be carried out in two principal ways: with scanner raw data, and with reconstructed DICOM images. The raw data approach is more complete but vendor-specific. Filtering in the reconstructed image domain is universal, and has another important advantage – we know the HU densities and locations of the structures in the image. Therefore we modified nonlinear bilateral CT denoising filter to incorporate different filtering settings depending on the underlying HU average: 1. Air and non-diagnostic areas were defined as below -500 HU, and were not processed at all. Some 50% of CT image belong to non-diagnostic background; excluding it from processing boosts the filtering speed 2. Soft tissues were identified as between -100 and 200 HU, to be denoised with standard filter parameters 3. Lung and boundary areas were identified as below -100 HU, to be processed with finer filtering, ensuring minimal aliasing. Same was applied to the areas above 200 HU (bones, contrast, metal) 4. Areas with local gradient above 80% image gradient quantile were treated as edges, to be processed with edge-preserving version of the filter The HU-specific filtering filter was implemented at our hospital to improve presentation of all low-dose CTs. 55 randomly selected studies were compared in a blind test, by 3 trained radiologists, to assess the filtering results in different image areas.

The hypothesis of improved filtering quality was confirmed in all image areas (p<0.001). The application of unified denoising filter greatly improved image interpretation speed, while reducing network traffic and PACS storage.

HU-specific image denoising provides a unified, vendor-independent approach to enhancing CT image quality, and CT dose reduction.

HU-specific image denoising provides a unified, vendor-independent approach to enhancing CT image quality, and CT dose reduction.

Pianykh, O, HU-specific CT Denoising for Dose Reduction.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11011894.html