TY  - GEN
N2  - Improvements for Medical Imaging
Robotic computer tomography systems offer a wide range possibilities for non-circular and patient specific trajectories. These could reduce metal artifacts,
improve the image quality or increase the available space for the practitioners.
But every trajectory needs to be calibrated to reconstruct the CT image. There
are two main approaches, an offline calibration uses a special phantom that has
to be imaged before the patient and the online calibration where the trajectory
is calibrated with the images acquired from using a prior CT image.
In chapter 5, a method for calibrating arbitrary Cone-Beam Computed
Tomography (CBCT) trajectories (FORCASTER) is developed. It detects
and matches feature points in simulated projections and acquired projections.
These paired feature points are then used to correct the projection parameters
by directly calculating the correction factor or with a minimization algorithm,
which is also presented in this part. To simulate projections a prior image of
the object is needed, but it is also shown that this calibration functions if the
prior image has some differences compared to the object that is currently imaged. The FORCASTER algorithm achieves similar calibration results when
compared to state-of-the-art algorithms.
The following chapter 6 introduces FORCAST-EST, an extension for the
FORCASTER algorithm, which estimates the starting parameters for the calibration. This allows the calibration of trajectories that come without positional
and rotational data for the individual projections. It simulates projections in a
grid and then compares every acquired projection to the grid, first in a coarse
search and then in the surrounding of the best matches. The estimates are
sufficient to calibrate the trajectory with a similar accuracy as FORCASTER
using the regular starting parameters recorded by the CBCT system.
In chapter 3 of this thesis, an algorithm for detecting arteries in dynamic
contrast enhanced MRI images is presented. Through thresholding, flood fill
and morphological operations, a mask for the artery at the time point where
the initial wave of contrast agent arrives is created. The process is deterministic
and independent of the user experience. Since the annotation of the artery has
a strong impact on the calculated perfusion values, helps this algorithm with
reproducibility and comparability of perfusion imaging.
The chapter 4 presents a web-based MRI image generator for teaching students. It allows the students to choose different sequences and set the relevant
parameters, and they can then see how it changes the resulting MRI image.
The software runs completely inside the browser and is open source1
.
The presented work show algorithms that help in the imaging process by
improving reproducability and comparability or in reconstructing CBCT volumes.
ID  - heidok34473
KW  - Kalibrierung
CY  - Heidelberg
UR  - https://archiv.ub.uni-heidelberg.de/volltextserver/34473/
AV  - public
A1  - Tönnes, Christian
Y1  - 2024///
TI  - Automatic interventional Imaging for the Parametrisation of Vascular Structures
ER  -