Skip Navigation
Search

Doctoral Defense

Improving resolution recovery and non-invasive quantification in Positron Emission Tomography

Nandita Joshi

April 11, 2019
12:00 PM
Light Engineering room 250
Advisor: Prof.  Christine DeLorenzo

In applications such as psychiatry, diagnoses are primarily subjective and qualitative based on patient report. Therefore, it is increasingly important to develop valid, unbiased scientific markers that reduce diagnostic bias. Positron Emission Tomography (PET) is an in vivo nuclear imaging technique used to study physiological, metabolic, and molecular processes in the human body through three-dimensional functional maps. Full quantification in PET imaging requires the collection of dynamic PET data along with invasive arterial blood sampling, which is expensive and deterrent to research participant enrollment.

The main objectives of this dissertation are: (1) to improve diagnostic accuracy of Major Depressive Disorder (MDD) by enhancing image resolution in PET (2) to investigate the possibility of non-invasive arterial blood sampling.

In order to improve resolution recovery, a novel Maximum Likelihood Expectation Maximization (MLEM) based image reconstruction technique with a gradually decreasing point-spread-function scheme was studied. The qualitative and quantitative performance of this technique was evaluated against existing MLEM techniques using simulated, cylindrical, and brain phantoms. In order to investigate the clinical utility of this technique, its performance was evaluated in a cohort of healthy individuals and individuals with MDD. The technique demonstrated improved resolution recovery in small structures, which could potentially be a useful objective tool in improving objective diagnosis of MDD by achieving a classification accuracy of 91%.

The VersaPET, which is a high-resolution auxiliary PET system developed by Dr. Paul Vaska, was used in order to investigate the possibility of non-invasive quantification. The image-derived radiotracer activity concentration in the lower leg was simultaneously measured using the VersaPET in an ongoing FDG PET study, while also collecting invasive arterial samples. Appropriate corrections were derived for decay, deadtime, normalization, and attenuation. Since the arteries in the leg are small, the images were also corrected for partial volume error. Additional preliminary experiments were performed in order to evaluate device stability to determine an optimized equivalence between invasive and VersaPET derived blood sample measurements.

These studies demonstrated that improving image resolution in PET could be a valuable objective tool for diagnostic classification of MDD and the potential for VersaPET to be utilized for non-invasive quantification.