MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

A Tailored Deep Residual Neural Network for Ultra-Low Dose CT Denoising and Its Impact On Emphysema Scoring

T Zhao*, M McNitt-Gray , D Ruan , David Geffen School of Medicine at UCLA, Los Angeles, CA

Presentations

(Wednesday, 7/17/2019) 10:30 AM - 11:00 AM

Room: Exhibit Hall | Forum 2

Purpose: Effective enhancement is crucial in compensating for degradation caused by dose reduction in low-dose screening. This study tailors a neural network model for (ultra-)low-dose CT denoising, and assesses its performance in enhancing CT image quality and a quantitative imaging (QIN) task for emphysema scoring.

Methods: To count for the unique characteristics in low-dose CT, we �rst simulate the paired ultra-low-dose and targeted high-quality image of reference, with a well-validated pipeline. These paired images are used to train a denoising convolutional neural network (DnCNN) with residual mapping. The performance of the tailored DnCNN is assessed over various dose reduction levels, with respect to both image quality and emphysema scoring quanti�cation. The possible over-smoothing behavior of DnCNN and its impact on different subcohort of patients are also investigated.

Results: The tailored DnCNN provided signi�cant image quality enhancement, especially for very low dose levels, improving the peak signal-to-noise ratio by 8dB and the structure similarity index by 0.15 approximately for 3% dose CT images. It outperformed the original DnCNN and the state-of-the-art nonlocal-mean-type denoising schemes. Using emphysema mask derived from full-dose images as reference, the proposed DnCNN signi�cantly reduced false positive rate on 3%-dose images while preserving sensitivity, compared to the detection based alternative approaches. DnCNN denoising resulted in minor oversmoothing, which contributed to slight underestimation of emphysema score compared to the reference, but did not affect clinical conclusions. The proposed method provided effective detection for cases with appreciable emphysema and reasonable corrections for normal cases without emphysema.

Conclusion: The proposed tailored DnCNN achieves (ultra-)low-dose CT denoising, with signi�cant improvement on both the image quality and clinical emphysema quanti�cation accuracy over various dose levels. The clinical conclusion of emphysema obtained from the denoised low-dose images agrees well with that from the full-dose ones.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email