Research Gallery
This is a selection of images and papers from various research topics I have explored. They are presented in no particular order, and the images were chosen for their visual and didactic value rather than their research impact.
Benchmarking Learned CT reconstruction with real data
Benchmarking learned algorithms for computed tomography image reconstruction tasks Maximilian B. Kiss, Ander Biguri2, Zakhar Shumaylov, Ferdia Sherry, K. Joost Batenburg, Carola-Bibiane Schönlieb, and Felix Lucka, Applied Mathematics for Modern Challenges, Vol 3, 2025
Modern data-driven reconstruction methods outperform all traditional reconstruction in toy simulated examples, however it is not clear how they perform with real data. Using the 2DeteCT dataset we test 12 learned methods from 4 mathematical categories under 9 different CT problems.
Coffee Bean
Arbitrarily large tomography with iterative algorithms on multiple GPUs using the TIGRE toolbox Ander Biguri, Reuben Lindroos, Robert Bryll, Hossein Towsyfyan, Hans Deyhle, Ibrahim El khalil Harrane, Richard Boardman, Mark Mavrogordato, Manjit Dosanjh, Steven Hancock, Thomas Blumensath, Journal of Parallel and Distributed Computing, Volume 146, 2020, Pages 52-63
This image shows a coffee bean of size 4000x4000x1000 reconstructed on a machine with a single GPU with less memory than the image size. The article proposes a novel computational pipeline for fast and arbitrarily large operators for tomography, allowing iterative reconstruction when it wasn’t possible before.
Commonly used Image Quality Assesment metrics are not appropiate for medical images
A Study of Why We Need to Reassess Full Reference Image Quality Assessment with Medical Images Anna Breger, Ander Biguri, Malena Sabaté Landman, Ian Selby, Nicole Amberg, Elisabeth Brunner, Janek Gröhl, Sepideh Hatamikia, Clemens Karner, Lipeng Ning, Sören Dittmer, Michael Roberts, AIX-COVNET Collaboration & Carola-Bibiane Schönlieb , Journal of Imaging Informatics in Medicine, 2025
Dr. Anna Breger leads this work investigating why the commonly used SSIM and PSNR metrics for image quality are not aligned with qualitative evaluations by experts. This particular study collects a variety of examples from different imaging modalities to highlight this issue. I recommend checking out other works by Anna for much deeper analysis.
Head Phantom on a clinical Varian-CBCT
TIGRE v3: Efficient and easy to use iterative computed tomographic reconstruction toolbox for real datasets Ander Biguri, Tomoyuki Sadakane, Reuben Lindroos, Yi Liu, Malena Sabaté Landman, Yi Du, Manasavee Lohvithee, Stefanie Kaser, Sepideh Hatamikia, Robert Bryll, Emilien Valat, Sarinrat Wonglee, Thomas Blumensath and Carola-Bibiane Schönlieb, Engineering Research Express 7 015011, 2025
The latest version of TIGRE does not just provide over 25 methods for iterative reconstruction, it also comes with many tools to load and handle real datasets from many clinical and industrial scanners. This image shows two reconstructions algorithms on the same data, acquired in a Varian-CBCT scanner at Beijing Cancer Hospital by Yi Liu. See the article for many more great pictures.
Image guided lung cancer radiation therapy
A general method for motion compensation in x-ray computed tomography Ander Biguri, Manjit Dosanjh, Steven Hancock and Manuchehr Soleimani, Physics in Medicine & Biology, Volume 62, Number 16, 2017
My PhD was on lung radiation therapy, a topic I still care about. These images show how regularized iterative reconstruction can significantly reduce the noise in images at low dose CBCT. Radiation therapy requies repeated imaging, particularly 4D-CBCT. Algorithms to reduce the dose significantly while maitaning high image quality can significantly improve patient prognosis. Nowadays I explore this methods with ML-driven models.
Tomography on tetrahedra meshes
Numerically robust tetrahedron-based tomographic forward and backward projectors on parallel architectures Ander Biguri, Hossein Towsyfyan, Richard Boardman, Thomas Blumensath, Ultramicroscopy, Vol 214 113016, 2020
This works explores using tetrahedra, instead of voxels, to represent images in Computed Tomography, and proposes tomographic operators that natively work in this image discretization. This discretization allows for arbitrary resolution with minimal memory footprint, and can incorporate prior information from e.g. CAD models as a discretization.
Trayectory optimization in CBCT
Source-detector trajectory optimization in cone-beam computed tomography: a comprehensive review on today’s state-of-the-art S Hatamikia, A Biguri, G Herl, G Kronreif, T Reynolds, J Kettenbach, T Russ, A Tersol, A Maier, M Figl, J H Siewerdsen and W Birkfellner, Physics in Medicine & Biology, Vol 67 16TR03, 2022
Together with Dr Sepideh Hatamikia and her group, we have explored wasy to optimize the scanning strategy to mazimize image quality, given kinematic and dose constraints. Refer to my articles with Sepideh for all the different advances in the field.