With the advent of flat-panel detectors capable of high-speed data communication, conventional geometric tomography is once again attracting attention in the form of tomosynthesis. In this study, the use of radial tom...
详细信息
ISBN:
(纸本)9783642038785;9783642038792
With the advent of flat-panel detectors capable of high-speed data communication, conventional geometric tomography is once again attracting attention in the form of tomosynthesis. In this study, the use of radial tomosynthesis for tomographic image reconstruction of breast cancer specimens from different directions using the shift-and-add algorithm, the simplest image reconstruction method used in tomosynthesis was validated. To simplify image processing, synchrotron X-rays that provide a parallel X-ray beam were used. The X-ray energy was 17.5 keV. The range of rotation of the object corresponding to the swing angle of the X-ray tube was +/- 40 degrees, with the object rotated in 1 degrees increments to scan a total of 81 projection images. From these projection images, 41 images were extracted with a tomographic angle of +/- 20 degrees with respect to an arbitrary fulcrum plane, and the tomographic planes were rotated radially to reconstruct varying tomographic images. Unlike multiplanar reconstruction, which has a high degree of freedom of choice of tomographic planes, in radial tomosynthesis, the variation in orientation of the tomographic planes is limited to a single axis. However, this study demonstrated that this method not only possesses sufficiently high spatial resolution to meet the requirements of breast cancer diagnosis for high-resolution images, but also provides tomographic images that can be observed from different directions.
Digital tomosynthesis (DTS) was evaluated as an alternative to cone-beam computed tomography (CBCT) for patient setup. DTS is preferable when there are constraints with setup time, gantry-couch clearance, and imaging ...
详细信息
Digital tomosynthesis (DTS) was evaluated as an alternative to cone-beam computed tomography (CBCT) for patient setup. DTS is preferable when there are constraints with setup time, gantry-couch clearance, and imaging dose using CBCT. This study characterizes DTS data acquisition and registration parameters for the setup of breast cancer patients using nonclinical Varian DTS software. DTS images were reconstructed from CBCT projections acquired on phantoms and patients with surgical clips in the target volume. A shift-and-add algorithm was used for DTS volume reconstructions, while automated cross-correlation matches were performed within Varian DTS software. Triangulation on two short DTS arcs separated by various angular spread was done to improve 3D registration accuracy. Software performance was evaluated on two phantoms and ten breast cancer patients using the registration result as an accuracy measure;investigated parameters included arc lengths, arc orientations, angular separation between two arcs, reconstruction slice spacing, and number of arcs. The shifts determined from DTS-to-CT registration were compared to the shifts based on CBCT-to-CT registration. The difference between these shifts was used to evaluate the software accuracy. After findings were quantified, optimal parameters for the clinical use of DTS technique were determined. It was determined that at least two arcs were necessary for accurate 3D registration for patient setup. Registration accuracy of 2 mm was achieved when the reconstruction arc length was > 5 degrees for clips with HU >= 1000;larger arc length (>= 8 degrees) was required for very low HU clips. An optimal arc separation was found to be >= 20 degrees and optimal arc length was 10. Registration accuracy did not depend on DTS slice spacing. DTS image reconstruction took 10-30 degrees seconds and registration took less than 20 seconds. The performance of Varian DTS software was found suitable for the accurate setup of breast cance
An efficient FPGA or ASIC based hardware implementation of deep neural networks face the challenge of limited chip area, and therefore an area efficient architecture is required to fully harness the capacity of parall...
详细信息
ISBN:
(纸本)9789813297661;9789813297678
An efficient FPGA or ASIC based hardware implementation of deep neural networks face the challenge of limited chip area, and therefore an area efficient architecture is required to fully harness the capacity of parallel processing of FPGA and ASIC in contrast to general purpose processors. In literature, the challenges are to investigate a generalized mathematical model and architecture for neuron block in an ANN implementation. We have proposed a generalized architecture for neuron implementation based on the shift-and-add algorithm, collectively known as Coordinate Rotation Digital Computer (CORDIC) algorithm, having a wide range of application. The look-up-table (LUT) based approach with a shift-and-add algorithm is an alternative technique for polynomial approximation and implementation. Paper explains how the CORDIC algorithm works and investigates the power and area efficient versatile computational unit for ANN application. The derived model proves that for the hyperbolic tangent function required a double pseudo-rotation and additional subtraction compares to the sigmoid function. In this reference versatile approach based optimized sigmoid activation function is implemented. The function is synthesized and validate on Xilinx zynq XC7Z010clg400 SoC and result reveals the minimum resources utilization.
Linear Computed Laminography(LCL) is often preferred in industrial inspection due to its high efficiency and low ***,the existing shift-and-add(SAA) algorithm introduces superimposed artifacts and blurring in the reco...
详细信息
Linear Computed Laminography(LCL) is often preferred in industrial inspection due to its high efficiency and low ***,the existing shift-and-add(SAA) algorithm introduces superimposed artifacts and blurring in the reconstructed *** this work,the weighted iterative deblurring(WID) algorithm is proposed to reconstruct the plate-like samples in *** subtracts the non-target layers from the original projection to reduce the redundant *** addition,the weighting matrix based on structural similarity is derived to eliminate the inter-layer *** subtraction and weighting operations are subsequently incorporated into an iterative framework through the recursive difference between the input and output layers within each *** experiments on a simulated four-layer printed circuit board show that WID reconstructs the target layers with only 7 projections,and effectively reduces the superimposed artifacts and ***,WID has great potential in online inspection scenarios such as assembly line and security check.
暂无评论