Research Interests
My research over the last 10 years has focused upon computerized analysis of medical images, with particular emphasis upon the extraction of blood vessels and brain tumors from medical image data. Our group is currently a world leader in both blood vessel and brain tumor egmentation from in vivo medical images, and we have been NIH funded since 1997 toward this task under the R01 "3 D cerebral vessel ocation for surgical planning".

The intent was initially to define the relationships between blood vessels and brain tumors for surgical planning purposes. However, about two years ago we made the unexpected and serendipitous observation that fmall blood vessels in the vicinity of malignant tumors were abnormally tortuous, whereas blood vessels in the vicinity of benign tumors were not. This observation led to new work on methods of mathematically defining the shapes of vessels extracted from medical images, and then a pilot study aimed at defining those vessel shape measures that seem distinguish "vessels of malignancy" from healthy vessels.

Most recently, Dr. Matt Ewend and I completed a blinded study of 21 brain tumors in patients whose images were analyzed prior to total gross resection of each lesion. The goal was to determine how well the program could distinguish between malignant and benign tumors, with the program's declaration compared to the final histological diagnosis. All 21 tumors, including 9 benign and 12 malignant lesions, were correctly classified.

Of importance, an unexpected finding was that vessel shape
abnormalities are present not only within tumor margins but also in the surrounding tissue, thus making it possible to classify even pinpoint or hemorrhagic lesions correctly. The approach therefore would appear to hold high promise of diagnosing cancer noninvasively. Moreover, the Iiterature suggests that abnormal vessel tortuosity is a feature of many kinds of malignant tumors and is not restricted to brain. The approach thus has the potential of providing a generic method of recognizing cancer that is applicable to any anatomical location.

Another interesting finding of the blinded study is that patients successfully cured of metastatic disease and now presenting with pure radiation necrosis without histologic evidence of recurrent tumor exhibit normalization of blood vessel shape. The program thus correctly declared all such tumors benign. Histological studies have shown that successful treatment of tumors in animal models produces rapid normalization of vessel shape. The analysis of vessel shape from in vivo patient images thus may have the potential of providing a rapid means of determining whether a patient may respond to a particular therapy or not.

Indeed, the Dana Farber Cancer Institute, in collaboration with UNC, Vanderbilt, and Georgetown, has received a grant from NSF support to evaluate the response of breast tumors metastatic to brain to a new chemotherapeutic agent. One of the major difficulties in treating such patients is determining early whether or not a patient will respond to therapy. Dr. Lisa Carey is leading the UNC effort. Dana Farber has adopted development of the vessel analysis method for early evaluation of treatment response as one of their specific aims, and will supply us with patient images obtained over time tram their own and other institutions.

An obvious extension of the current work is to analyze the vasculature of animals with brain tumors. We have entered into collaboration with Drs. Terry Van Dyke and Weili Lin, and are currently examining the vasculature of animals with gliomas and choroid plexus carcinomas. Preliminary results are exciting, and so far suggest that the vessel shape changes seen in mice are identical to those we find in humans. Future work will involve evaluation of vessel shape normalization during treatment. Weare also conducting pilot studies in collaboration with Dr. Kim Rathmell to analyze vessel shape and density in different types of teratomas implanted in the neck.

In terms of relevant planned and pending grant submissions, I currently have submitted a competing renewal of the "vessel location" RO I. This competing renewal focuses upon the diagnosis of cancer in human beings. I plan to submit a new NIH ROI proposal on animal work once we have additional preliminary data. As previously noted, the Dana Farber Cancer Insitute has submitted an NSF proposal that includes the vessel analysis work in patients with metastatic breast cancer.

We include the following five publications. The first three discuss measures of vessel shape and the diagnosis of malignancy. The last two address measures of defining brain tumors automatically from medical image data.

Publications
Bullitt E, Gerig G, Pizer S, Aylward SR (2003) Measuring tortuosity of the intracerebral vasculature from MRA images. IEEE-TMI 22:1163-1171.
Bullitt E, Gerig G, Aylward S, Joshi S, Smith K, Ewend M, Lin W (2003) Vascular Attributes and Malignant Brain Tumors. Lecture Notes in Computer Science 2878:671-679.
Bullitt E, Jung I, Muller K, Gerig G, Aylward S, Joshi S, Smith K, Ewend M (2004) Determining malignancy of brain tumors by analysis of vessel shape. Submitted LNCS 2004.
Prastawa M, Bullitt E, Moon N, Van Leemput K, Gerig G (2003) Automatic brain tumor segmentation by subject specific modification of atlas priors. Acad. Radiol. 10: 1341-1348
Prastawa M, Bullitt E, Ho S, Gerig G (2004) A brain tumor segmentation framework based on outlier detection. Accepted MedIA.

E-mail: bullitt@med.unc.edu
Telephone: 919-843-3101
FAX: 919-843-1500
Address: CASILab, 349 Wing C Chapel Hill, NC
URL: http://casilab.med.unc.edu

Click here to update this profile