NEWS Researchers at Umeå University have developed methods for imaging with positron emission tomography, PET, despite fewer and less accurate data. The methods can enable faster, cheaper and safer diagnosis of cancer patients.
Klara Leffler and her supervisor, Professor Yun Yu, at the Department of Mathematics and Mathematical Statistics.
ImageJohan GunséusPositron emission tomography, abbreviated PET, is a medical imaging technique that can visualise how different substances move in the body by following the decay of radioactively labelled molecules. Its main clinical application area is oncology, where PET plays an important part in cancer detection, staging, and treatment assessment.
Klara Leffler, PhD student at the Department of Mathematics and Mathematical Statistics, shows in her thesis how improving and streamlining data collection in PET examinations is possible. By reducing the need for large amounts of measurement data, the examinations can be carried out faster and with less radioactive dose, which increases the patient's safety and comfort.
Images from a PET examination.
The challenge is to deliver sharp images even with incomplete or "cloudy" data. Klara Leffler has used what, in simple terms, can be described as "mathematical fine-tuning" to selectively manage and improve data (sparse signal processing). The result is methods that can improve or completely reproduce parts of the measurement data.
"It's like solving a puzzle with missing pieces," says Klara Leffler.
Two main challenges have been in focus during the research work. One deals with noisy data at lower radioactive doses or shorter scan times. The second is to supplement incomplete information from PET scanners with gaps in their measurement patterns (sparse detector configurations).
The results show how enhancing and recovering valuable data can be improved using mathematical computational methods and deep learning.
In addition to safer and more efficient examinations, the results can reduce costs and facilitate access to advanced imaging techniques. This research is not only relevant to PET imaging but may also inspire new methods in other areas of biomedical imaging.
"This work has been a journey of discovery, not only scientifically but also in how different disciplines can work together to bridge the gap between theory and practical application. Seeing how mathematical models can be adapted to medical needs has been both challenging and enlightening," says Klara Leffler.
On Thursday, 29 February, Klara Leffler, Department of Mathematics and Mathematical Statistics, will defend her thesis entitled The PET Sampling Puzzle: Intelligent data sampling methods for positron emission tomography. The defence takes place at 9:00 a.m. in room BIO.E.203 – Aula Biologica, Umeå University.
