Dominic J. Heuscher

Exact helical reconstruction using native cone-beam geometries.

This paper is about helical cone-beam reconstruction using the exact filtered backprojection formula recently suggested by Katsevich (2002a Phys. Med. Biol. 47 2583-97). We investigate how to efficiently and accurately implement Katsevichs formula for direct reconstruction from helical cone-beam data measured in two native geometries. The first geometry is the curved detector geometry of third-generation multi-slice CT scanners, and the second geometry is the flat detector geometry of C-arms systems and of most industrial cone-beam CT scanners. For each of these two geometries, we determine processing steps to be applied to the measured data such that the final outcome is an implementation of the Katsevich formula. These steps are first described using continuous-form equations, disregarding the finite detector resolution and the source position sampling. Next, techniques are presented for implementation of these steps with finite data sampling. The performance of these techniques is illustrated for the curved detector geometry of third-generation CT scanners, with 32, 64 and 128 detector rows. In each case, resolution and noise measurements are given along with reconstructions of the FORBILD thorax phantom.

A knowledge‐based cone‐beam x‐ray CT algorithm for dynamic volumetric cardiac imaging

With the introduction of spiral/helical multislice CT, medical x-ray CT began a transition into cone-beam geometry. The higher speed, thinner slice, and wider coverage with multislice/cone-beam CT indicate a great potential for dynamic volumetric imaging, with cardiac CT studies being the primary example. Existing ECG-gated cardiac CT algorithms have achieved encouraging results, but they do not utilize any time-varying anatomical information of the heart, and need major improvements to meet critical clinical needs. In this paper, we develop a knowledge-based spiral/helical multislice/cone-beam CT approach for dynamic volumetric cardiac imaging. This approach assumes the relationship between the cardiac status and the ECG signal, such as the volume of the left ventricle as a function of the cardiac phase. Our knowledge-based cardiac CT algorithm is evaluated in numerical simulation and patient studies. In the patient studies, the cardiac status is estimated initially from ECG data and subsequently refined with reconstructed images. Our results demonstrate significant image quality improvements in cardiac CT studies, giving clearly better clarity of the chamber boundaries and vascular structures. In conclusion, this approach seems promising for practical cardiac CT screening and diagnosis.

Redundant data and exact helical cone-beam reconstruction

This paper is about helical cone-beam reconstruction and the use of redundant data in the framework of two reconstruction methods. The first method is the approximate wedge reconstruction formula introduced by Tuy at the 3D meeting in 1999. The second method is a (exact) hybrid implementation of the exact filtered backprojection formula of Katsevich (2004 Adv. Appl. Math. at press) that combines filtering in the native cone-beam geometry with backprojection in the wedge geometry. The similarity of the two methods is explored and their image quality performance is compared for geometries with up to 112 detector rows. Furthermore, the concept of aperture weighting is introduced to allow the handling of variable amounts of redundant data. A reduction of motion artefacts using redundant data is demonstrated for geometries with 16, 32 and 112 detector rows using a pitch factor of 1.25. For scans with up to 100 rows, utilizing 50% of the redundant data provided excellent results without any introduction of cone-beam artefacts. For larger cone angles, an alternative approach that utilizes all available redundant data, even at reduced pitch factors, is suggested.

Evaluation of helical cone-beam CT reconstruction algorithms

In this work, three different reconstruction algorithms for short scan helical cone-beam CT are compared: Two approximate algorithms, PI-SLANT and WVEDGE-PI, with the recently published exact algorithm by Katsevich. It is shown that WEDGE-PI performs as well as the exact method for a 64 row scanner and almost as well for a 128 row scanner. PI-SLANT produces significantly more artifacts, in particular for the 128 row scanner.

An analytic model of the cardiac cycle

This paper analyzes the volume curves of the cardiac chambers in the cardiac cycle. Volume curves were measured from the cardiac CT test case reconstructed at 6 different phases. The ventricular and atrial volume curves were then interpolated and applied to an anthropomorphic analytical model human heart. The volume curves modify the analytical model at any given point of the cardiac phase creating a highly detailed model of a beating ventricular cycle. This model can be used in analytic projections simulating acquisition of a CT scanner and help study the temporal resolution and motion artifacts of image reconstruction techniques.

Prospectively gated cardiac CT

Future cardiac CT protocols will utilize large area detectors with whole heart scans performed within one heartbeat. For such scans, accurate prospective ECG gating is essential to capture the heart at the correct phase. This report addresses one of the main factors affecting the prospective gating accuracy: the ability to predict the cardiac phase from the ECG signal. Two different scanning methods with equal average dose utilization along with 6 different prediction algorithms are compared and evaluated on 245 patient ECG signals. In general, all the algorithms performed comparably while, perhaps surprisingly, the best scan method was to rescan the patient as necessary as opposed to elongating a single scan.

Measuring temporal resolution of cardiac CT reconstructions

Multi-slice CT today is capable of imaging the heart with excellent temporal resolution. Algorithms have been developed to perform reconstructions combining data from multiple cardiac cycles. This paper presents a simulation phantom that enables a direct measurement of the actual temporal resolution achieved by these algorithms. This is not only useful for assessing the temporal resolution but also for validating the algorithms themselves. A simulation phantom was developed that consists of a 20 cm. diameter water phantom containing an array of cylinders whose intensities are pulsed for various durations ranging from 10 msec. to 250 msec. The intensity varied between the background value of water (0 HU) and 800 HU. By measuring the nominal attenuation value at the center of each cylinder, a curve can be derived representing the response over the given temporal range. A temporal resolution representing the FWHM value is determined based on the half-max value of this curve. Reconstructions were performed using a multi-cycle cardiac algorithm described previously in the literature. The measured FWHM values agree quite well to the temporal resolution predicted by the cardiac algorithm itself. Even the variation along the longitudinal axis can be accounted for by the predicted values. A simulated phantom can be used to accurately assess the temporal resolution of cardiac reconstruction algorithms. Excellent agreement was achieved between the predicted and measured temporal resolution values for the multi-cycle algorithm used in this study.