This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Thorax, Computer applications, Radioprotection / Radiation dose, CT, Comparative studies, Computer Applications-Detection, diagnosis, Technology assessment, Image registration, Image verification
Authors:
M. Ohana1, A. Labani1, M.-Y. Jeung1, C. Ludes1, C. Roy1, C. Collet2, F. Heitz2, F. Rousseau2, V. Noblet2; 1Strasbourg/FR, 2Illkirch/FR
DOI:
10.1594/ecr2016/C-2126
Conclusion
We have demonstrated that NLM image-domain denoising is quantitatively and qualitatively efficient in improving ULD unenhanced chest CT,
with increase of both subjective as well as objective image quality when using a β coefficient of 1-2.
The difference in the optimal β coefficient between the qualitative and the quantitative evaluation (1-2 versus 4) could be explained by two points :
- the reference full dose acquisition is not rigorously noiseless,
and therefore the PSNR calculation is,
to some extent, biased.
- the registration is not strictly perfect,
and therefore more smoothing is necessary to compensate the small misregistration artifacts,
explaining that a higher β coefficient is needed with the quantitative evaluation.
This preliminary work,
even though promising,
raises many unanswered questions:
- Could other NLM parameters achieve better results? We choose ours based on empirical experience on non-medical images.
- Could other image-domain denoising algorithm achieve better results?
- Could machine-learning with an a priori knowledge of the image degradation be possible ? Based on the combination of the full dose and the ULD acquisition,
an atlas could be established and constitute a support for improved noise suppression.
- Could we better optimize the computational time?
- What would be the results on pathological examinations? Could denoising mask or "erase" subtle CT abnormalities?
We do tried to apply this method on a patient with a mild centrilobular emphysema,
and the results looks promising with easier and sharper delineation of the bullae.
Fig. 8: Additional testing on a patient with mild emphysema: original ULD-CT image
Fig. 9: Additional testing on a patient with mild emphysema: optimal denoised ULD-CT image (β1.5)
Additional studies are therefore under way to further improve these first results.