Current Openings for Students
Here you can find our currently available student projects (published via SiROP).
Melanie Brtan (email@example.com) did an internship in the TOMCAT group and worked with AI algorithms for image denoising on grating-interferometry breast computed tomography (GI-BCT) images. She experimented with different methods to improve the neural network called INSIDEnet, in order to outperform nowadays denoising neworks. Her goal was to implement a more general denoising algorithm as a proximal mapping within an interative phase constrast CT reconstruction pipeline.
Ketan Gupta (firstname.lastname@example.org) did his semester project on the topic of Talbot-Lau interferometry. He worked on implementing continuous phase stepping to reduce the acquisition time. He developed a framework to move and track the grating position over the image acquisition in a predefined pattern. This allowed him to implement a generalised Computed Tomography (CT) acquisition and analysis pipeline for a continuously moving source grating. He successfully demonstrated the viability of his work by showing similar Contrast-to-Noise ratios of stepped and continuous measurements.
Sandra Haltmeier (email@example.com) did her semester thesis on wavelet-based regularizers for iterative tomographic reconstruction in grating interferometry breast computed tomography (GI-BCT). She implemented a reconstruction algorithm called fast iterative shrinkage-thresholding (FISTA), which uses a wavelet transform to distinguish between relevant high-frequency image content and noise to eliminate the latter by using soft-thresholding. Ultimately, combining total variation (TV) regularization, commonly used in absorption-based CT image reconstruction, and FISTA increased the reconstruction quality on simulated and actual data compared to standard non-machine-learning-based reconstruction methods.
Javier García Baroja (firstname.lastname@example.org) did his semester thesis on the quantitative material decomposition of grating-interferometer-based computed tomography images (GI-BCT) to evaluate the protein, lipid and water content of the scanned tissue based on the complementary information from the attenuation and phase-contrast reconstructions. The application of this analysis is two-fold. On the one hand it can facilitate tumor localization as they have a different composition from their surroundings. Additionally, this tool has the potential of reducing the work load of radiologists as they progressively receive larger amounts of data.
Eliot Jermann (email@example.com) did his semester thesis on system optimisation and software design in Grating-Interferometry (GI) X-Ray Imaging laboratories. Sample stage of the micro-GI setup was successfully made independent from PSI controlling system and has been used for cone beam X-ray-CT scans of cancerous and non-cancerous breast samples.
Alexander Pereira (firstname.lastname@example.org) did his master thesis on the topic of differential phase contrast (DPC) mammogram denoising and integration. To obtain the clinically interesting phase contrast image, the collected DPC image from the mammogram has to be integrated. Unfortunately, this integration step has demonstrated to be very challenging to perform, due to the intrinsic noise arising in the image acquisition. To overcome this problem, he developed deep learning models for image denoising. Using simulated mammogram images that he created himself, he was able to train the networks both supervised and unsupervised, which eventually led to promising denoising results even for real captured images and outperformed conventional denoising techniques.
Dominik Etter (email@example.com) did his master thesis on topic of dark field imaging of porcine lungs to boost the clinical diagnostics of emphysema-diseased lungs with non-absorption types of contrast in X-ray imaging. An artificial thorax designed by him was used to inflate porcine lungs in a passive manner. This allowed to image lungs on three different imaging setups – two conventional Talbot-Lau grating interferometer and a so-called dual phase interferometer. He was actively involved in handling and optimising the imaging process as well as the data evaluation afterwards. His work allowed to characterise lung tissue in terms of their small angle scattering properties in comparison to well defined, standardised samples to allow for a more targeted developments of future (pre-)clinical imaging setups for dark-field imaging of lungs.
Lionel Peer (firstname.lastname@example.org) did his bachelor's thesis on the optimisation of the biomedical table-top system for high-resolution X-ray imaging and tomography. A Software environment, that was written by him, helped to increase the acquisition speed and overall performance of setup. Apart from that, he was actively involved in the HW design & tests which helped to come up with better parameters of the aforementioned system and resulted in a scientific publication in 2021. In addition, he gained valuable practical work experience in the lab and mastered his team-work skills.
Lars Lenherr, student of the Electrical Engineering Department of ETH, realised his bachelor's thesis "A Measurement Control Package for Grating-Interferometry X-Ray Imaging Laboratories" in our group in Spring 2011. He spent most of his time in our laboratory in the Paul Scherrer Institute, where he first familiarised himself with the control aspects of x-ray imaging with grating interferometry and later moved to design and implement a Python-based framework for equipment control.
Leonie Sommer (email@example.com) did her bachelor's thesis on the integration of a direct piezo drive into an X-ray Talbot-Lau interferometer to enable new phase acquisition methods. She commissioned the Piezo-drive and wrote the software to integrate its controls into a pre-existing software package. She then evaluated the visibility profile of multiple source gratings to optimise the data quality of the setup and developed adapted acquisition and analysis scripts suited for continuous source grating motions. In her thesis she was able to show the feasibility of continuous source grating motion during image acquisition for radiography with different motion cycles.
Simon Spindler from the department of physics did his master thesis on "Optimisation of a Quantitative Wide-Field-of-View Phase-Contrast Computed-Tomography Setup". He worked on a prototype Grating-Interferometry-based Breast-CT system, which he learned to operate and whose images he learned to analyse. With the help of simulation Simon has implemented he identified causes of artefacts present in the images, proposed changes to the imaging protocol and showed that they greatly improve the image quality.