Uncertainty characterization and propagation for emerging Nuclear Medicine targeted therapy.
This project is with the Division of Nuclear Medicine of the Dept. of Radiology at the NIH Clinical Center.

This project can also be virtual.

Project Description: 

The resurgence of the use of radiotracers for cancer therapy has resulted from two recent innovations: 1) the improvements in PET and SPECT scanners (total body PET imaging and faster and more quantitative SPECT tomography) and 2), the development of new radiolabeled agents with high diagnostic and therapeutic potential that can be paired by virtue of their co-localization on the same specific target.  Such pairs are referred to as “theranostic”:  having a diagnostic arm with positron or single-gamma emitting labels suitable for imaging—allowing more accurate dosimetry and follow up— and a therapeutic arm that uses the same ligand but attached instead to alpha or beta particle emitters. This allows quantifiable monitoring and ablation of the diseased cells in cancers such as prostate, neuroendocrine tumors and others.  These innovations, have also made it possible to better take advantage of the quantitative nature of nuclear imaging to better manage and potentially cure certain cancers that, so far, have eluded other therapies. As with any therapeutic intervention, a most important attribute of these procedures is their dose-response predictability.  For targeted therapy, the challenge in this respect, is the design of practical methods that ensure optimal accuracy of the actionable and medical relevant parameters (AMR) such as the therapeutic dose to be administered or the degree of response to therapy. These values are derived from measurements extracted from diagnostic images, from other devices, or from operator interventions—e.g. manually drawing regions of interest. All these sources have uncertainties of their own that propagate in the estimation of the AMR.  It is essential, therefore, to evaluate the magnitude and propagation rate of these contributing errors.  Then, based on a proper error-sensitivity analysis, it should be possible to properly design algorithms or workflows that minimize the most impactful errors or that mitigate their effect on the accuracy of these key values used for clinical decisions.  This Summer’s internship will focus on one or more aspects of this challenge.