Practical approach to misalignment correction in a single-circle orbit cone-beam tomography

This contribution addresses the reduction of artifacts in cone-beam tomography (CT and SPECT/PET) that are caused by imperfect scanner mechanical alignment (i.e. misalignment). Such artifacts may show up as double contours, split edges, smearing, loss of resolution, change in magnification etc. Misalignment is a problem commonly recognized in SPECT/PET (see for example [12]) whereas in CT its importance has only recently been emphasized with the advent of cone beam geometry [5][9][11], although residual misalignment can be responsible for stair-step artifacts even in single-row helical scanners [7]. In order to improve the quality of reconstructed images it is therefore necessary to take care of misalignment errors, either by proper scanner alignment or by measuring the errors in advance and forwarding them to the reconstruction algorithm. As high precision mechanical components are very expensive it may be more promising to take the latter approach but the accurate measurement of misalignment parameters still remains a great problem, in particular in the case of high resolution microCT (μCT). Here we investigated a practical approach: Is it possible to achieve satisfactory artifact reduction by correcting only a few vital misalignment parameters that can be measured easily. We used an modified Feldkamp-based circular scan reconstruction algorithm that we have developed for our μCT cone-beam scanner [1]. This algorithm incorporates corrections for all possible misalignment errors. Misalignment compensation is embedded in the backprojection. Cone-beam geometry and scanner misalignment The key components of a cone beam CT scanner are the Xray source, a two-dimensional X-ray detector and a sample positioner located between them. In μCT typically the tube and detector remain fixed and the object is rotated while in medical CT the object is at rest. However cone beam CT with rotating source and detector has not been realized yet. In SPECT the detector (gamma camera) is equipped with a converging collimator, which focuses at some point behind the object. The detector rotates about the patient along a circular path. For convenience in this contribution we will refer to a CT scanner, but our analysis remains valid for SPECT as well. In an ideal case the scanner is perfectly aligned, i.e.: • the straight line between the X-ray focal spot and the center of the detector is normal to the detector surface; this is called the central ray that together with central row of the detector defines the midplane • the axis-of-rotation (AOR) is parallel to the detector columns and is projected onto the central column. There are several reasons why in practice residual misalignments are unavoidable: • fine adjustment of the scanner requires high precision positioning mechanics, which might be often too expensive to be worth building into the scanner. • misalignment may result from an unstable X-ray focal spot position, which is usually the case in X-ray tubes with a very small focus size. There are several degrees of freedom for deviation from the ideal geometry. If we arbitrarily take the central ray and the midplane as a reference then the misalignment errors can be defined as follows (Fig. 1): • deviation of the AOR from ideal orientation and position tilt (inclination) towards the X-ray tube skew (rotation) around the central ray horizontal transversal off-center shift, i.e. along detector rows horizontal longitudinal shift, i.e. deviation from the ideal position between the X-ray tube and the detector • deviation of the X-ray source location from the ideal position vertical shift from the midplane horizontal transversal shift from the central ray horizontal longitudinal shift, i.e. deviation of the X-ray tube ↔ detector distance from the assumed value. Other possible errors (AOR wobble etc.) are assumed to be negligible or not present at all. For spiral/helical and other more complex acquisition paths more errors have to be appended to the list. For example the direction of object translation (table feed in medical CT scanners) may not be parallel to the AOR. The scanned object is then incrementally shifted off-center while being advanced to the next projection position on the spiral path. AOR skew and transaxial shift Ideal AOR