In the last 5 months, Dr. Warren
has focused on the task of computing a lobe-based model for the lungs and
developing a method for computing 3D warps between successive images in a
This paper describes a
semi-automatic method for computing such a lobe-based model and constructing
the desired 3D deformations between successive 3D CT scans of the lungs. Note
here that the paper focuses only on semi-automatic methods that require some manual
intervention by the radiologist. In our last progress report, we had hoped to
construct fully automatic method for these problems. However, based on our
efforts in December 2006-February 2007, we conclude that building a fully
automated method that is robust in the presence of the type of data variant
present in our test data was simply too ambitious in the short term.
Since generating plausible
preliminary work is crucial for a successful NIH grant, we instead decided to
focus on building semi-automatic methods in the short term. The goal of these
methods would to construct the high-quality deformations necessary to assess
whether ventilation and perfusion can be computed from 3D deformations between
CT images. To this end, we have successfully constructed such deformation and
developed a new, simplified mathematical model for computing ventilation and
perfusion from this deformation.
Dr. Guerrero is enthusiastic
about the quality of our results and feels that the paper has a chance at being
accepted in a high-quality journal. We are currently deciding whether to submit
the paper to a low-quality journal that has short turnaround or submit the
paper to a higher-quality journal with longer turnaround. Our current thought
is to probably submit to fast-turnaround journal and look to submit a proposal
to NIH in June with the paper already accepted.
Dr. Zhang has also worked on
developing an improved deformation method based on optical flow. In his latest
paper, Compressible
image registration for thoracic computed tomography images,
Dr. Zhang has developed a 2D version of optical flow that is capable of
modeling compressible flows (such as air flow in the lungs). Dr. Zhang’s
graduate student have implemented this method and tested the method on pairs of
2D images of lung cross-sections. Based on favorable results from the 2D
experiments, Dr. Zhang and his students are currently working on a 3D version
of the method.
Our decision to delay submission
of an NIH grant from February to June was also motivated by one more factor.
Dr. Guerrero has asked Dr. Warren to participate as a co-PI in an NIH proposal
to study assess the effectiveness of various drugs in mitigating the effects of
radiation exposure. In particular, the aim of the proposal is to
identify drugs that someone who has been exposed to a large dose of radiation
(say as the result of the explosion of a “dirty” bomb) may take to mitigate the
sometimes fatal damage to their small intestines from this radiation.
Although this topic does not
directly address GC4R’s main topic of cancer research, Dr. Warren agreed to
participate because he (and Dr. Guerrero) felt that this proposal had a high
chance of success due to its relevance to Homeland Security. Based on advice
from others who have been successful in pursuing NIH funding, we felt that
having an established record of successful NIH funding would enhance our
chances of eventually securing NIH funding for our lung project.