Takeshi Nakaura MD, Presenter: Nothing to Disclose

Yuji Iyama MD, Abstract Co-Author: Nothing to Disclose

Masafumi Kidoh, Abstract Co-Author: Nothing to Disclose

Shinichi Tokuyasu RT, Abstract Co-Author: Employee, Koninklijke Philips NV

Kazunori Harada, Abstract Co-Author: Nothing to Disclose

Yasuyuki Yamashita MD, Abstract Co-Author: Consultant, DAIICHI SANKYO Group

Shouzaburou Uemura, Abstract Co-Author: Nothing to Disclose

Toshinori Hirai MD, Abstract Co-Author: Nothing to Disclose

Seitaro Oda MD, Abstract Co-Author: Nothing to Disclose

Image noise is a serious problem in brain CT because of the requirements for contrast resolution. Previous report suggested that the recent introduced knowledge-based iterative model reconstruction (IMR) is able to reduce image noise, offer accurate CT attenuation, and enable improvement in low-contrast detectability. The purpose of this study was to evaluate the usefulness of IMR in brain CT especially with thin slice images.

This prospective study received institutional review board approval; prior informed consent to participate was obtained from all patients. This study enrolled 34 patients who underwent brain CT. We reconstructed axial images with filtered back projection (FBP), hybrid-iterative reconstruction (HIR) and knowledge-based-IMR with 1 and 5 mm slice thickness. We compared the CT number, image noise, contrast, and contrast noise ratio (CNR) between the thalamus and the internal capsule and the rate of increase of image noise in 1 mm thickness images from 5 mm thickness images between the reconstruction methods with the Holm test. Two independent readers assessed image contrast, image noise, image sharpness and the overall image quality of the 1mm thickness images with each reconstruction technique on a 4-point scale.

There were significant differences in the CT numbers between IMR and the other reconstruction techniques (p<0.01). The image noise was significantly lower with knowledge-based-IMR (2.4 HU ± 0.3) compared to FBP (4.9 HU ± 0.5) and HIR (4.1 HU ± 0.4) (p<0.01). The contrast and the CNR between the thalamus and the internal capsule were significantly higher with IMR (5.1 HU ±1.6; 2.2±0.8) relative to FBP (4.8 HU ±1.7; 1.0±0.4) and HIR (4.8 HU ±1.7; 1.2±0.4) (p<0.01). The rate of increase in noise in 1mm thickness images was significantly lower with IMR (70.7% ± 32.6) compared to FBP (130.1% ± 31.9) and HIR (129.6% ± 35.2) (p<0.01). The visual scores in image contrast, image noise and overall image quality with knowledge-based-IMR were significantly higher than that with other reconstruction images (p<0.05).

IMR offers significant noise reduction, higher contrast and CNR in brain CT especially with thin slice images compared to FBP and HIR.

IMR might offer higher image quality compared to FBP and hybrid-IR in brain CT especially for thin slice imaging.

Nakaura, T, Iyama, Y, Kidoh, M, Tokuyasu, S, Harada, K, Yamashita, Y, Uemura, S, Hirai, T, Oda, S, Comparison of Knowledge-based Iterative Model, Hybrid Iterative and Filtered Back Projection Reconstruction Techniques in Brain CT: Impact of Thin Slice Imaging.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14018754.html