Compensating for Limited Spatial Resolution of Positron Emission Tomography

While positron emission tomography (PET) has the unique ability to probe living cells in a non-invasive and highly sensitive manner, it is still plagued with relatively low spatial resolution compared to anatomy-oriented imaging devices such as MRI or CT.

This webinar will present an overview of the techniques used to compensate for the limited spatial resolution of emission tomography imaging systems such as PET, as applied to the human brain. Starting with the fundamental principles of PET and the factors affecting spatial resolution, the speaker will then dive into a detailed description of the fundamental aspects of the object-image relationship and the general problem of image restoration.

A review of the various methods developed over the past 30 years to compensate for the so-called partial volume (PV) effects will follow. Emphasis will be given to the most widely used methods in current use today, concentrating on post-reconstruction PV correction (PVC) methods, and discussing their fundamental differences.

Details on the practical implementation of a PVC scheme on large patient cohorts and in a mostly unsupervised fashion will be presented through the description of image processing pipelines. Characterization of the interaction between the radioactivity distribution and the PET scanner will be discussed in terms of brain segmentation, and the scanner’s effective spatial resolution assessment through physical phantom experiments.

The presentation will then convey applications of PVC in clinical and research applications, notably neurodegenerative disorders such as Alzheimer’s and Parkinson’s Disease, and their potential implication in a clinical trial setting.

Finally, a section will be devoted to an extension of one particular method: GTM-PVC, that allows for head movement correction, another factor potentially participating to the degradation of PET images, and that is becoming a major contributing factor to image blur, being typically of the same order of magnitude as intrinsic spatial resolution of modern PET systems.

Speaker

Olivier Rousset, PhD, PET Scientist, Neuroscience, Bioclinica

Academic Training:
PhD in BioMedical Engineering, Claude Bernard University, Lyon 1 (France)
BSc in Physics -Science and Structure of Matter, Claude Bernard University, Lyon 1

Imaging Experience and Specialization:
Experience in Emission Tomography techniques with an emphasis on PET. Experience with structural MRI for multimodality correlation (registration, segmentation, atlasing).
Formalized an entirely new method to address partial volume effects, the major source of quantification inaccuracies in PET. Extensively validated with original phantom experiments and computer simulations. Notably reported its accuracy and sensitivity to statistical noise, and proposed method to compute maximum expected inherent data variance increase -with experimental data closely following my theoretical predictions. Also developed methods to assess true image resolution of PET scanners as well as head movement correction solutions. My method (usually referred to as the GTM, RSF, or Rousset method) has been adopted by groups such as Columbia University, Univ of Berkeley, Karolinska Institute (Sweden), and UCLA, as well as being supported by both open-source (e.g., SPM, PETpvc) and commercial (e.g., PMOD) software.

Therapeutic Area Experience:
Neuroscience: Alzheimer’s disease, Rett syndrome, Tourette, Cocaine addiction, PD

Message Presenter