Program Information
Detectability of Multiple Slice Breast Cone Beam Computed Tomography Images
M Han*, B Kim , J Baek
Presentations
TU-L-GePD-I-3 (Tuesday, August 1, 2017) 1:15 PM - 1:45 PM Room: Imaging ePoster Lounge
Purpose: To investigate the detectability of multiple slice breast cone beam computed tomography (CBCT) images.
Methods: To model breast anatomy volume, we filtered 3D Gaussian noise with a square root of 1/f³ kernel, and assigned the attenuation coefficient of glandular tissue (0.8cm⁻¹) for the top 30% voxel values and adipose tissue (0.46cm⁻¹) for the bottom 70% voxel values. Projection data of a breast volume were generated by the forward projection, and quantum noise was added to the projection data. Then, analytic projections of a spherical signal with a 5mm diameter and 0.03cm⁻¹ attenuation coefficient were added. The signal location was set near the center of the breast volume. The FDK algorithm with Hanning weighted ramp filter was used to reconstruct projection data. To evaluate detectability of multiple slice breast CBCT images, we used a multi-slice channelized Hotelling observer (CHO) model with Laguerre-Gauss (LG) and dense difference-of-Gaussian (D-DOG) channels. The CHO was applied to the transverse and longitudinal directions of the breast CBCT images.
Results: Detectability increases as the number of slices increases, and saturates when the slice thickness is larger than the signal size. With LG CHO, longitudinal plane yields higher detectability despite of its higher β value. When D-DOG CHO is used for a single slice, the longitudinal plane has higher β value, but yields higher detectability. However, with multiple slices, the transverse plane produces smaller β value with higher detectability.
Conclusion: In this work, we investigated detectability of multiple slice breast CBCT images. Using multiple slices was more beneficial than using a single slice for lesion detection, and the LG CHO (D-DOG CHO) yielded higher detectability in longitudinal (transverse) plane with multiple slice breast CBCT images.
Funding Support, Disclosures, and Conflict of Interest: MSIP (Ministry of Science, ICT and Future Planning), Korea, under the IT Consilience Creative Programs (IITP-R0346-16-1008) supervised by the IITP (Institute for Information & Communications Technology Promotion) and Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by MSIP (2015R1C1A1A01052268)
Contact Email: