PET-guided tumor volume delineation for radiation therapy Positron-emission tomography (PET), especially 18F-FDG, is increasingly being used in radiation therapy to assist the radiation oncologist in delineating the three-dimensional tumor shape for treatment planning. While PET imaging provides access to very sensitive molecular probes, the PET imaging process suffers from relatively poor spatial resolution. This, along with the PET image acquisition and reconstruction process, causes the apparent PET radioisotope uptake distribution to spread well beyond the physical radioisotope uptake distribution. Existing methods to segment the tumor boundary using PET images have focused on intensity-based thresholding, which is limited by ambiguity with respect to the threshold selection. The reason that intensity-based thresholding methods do not work well with PET images is because the intensity histograms are not bimodal in shape, so there is no stable threshold that can be automatically and objectively detected. We propose a novel two-stage image analysis process that overcomes the difficulties with the existing segmentation methods. The first stage, adaptive region growing (ARG), analyzes the PET image such that a reliable and reproducible landmark is produced that defines a """"""""preliminary"""""""" tumor volume, which is larger than the actual tumor volume. The second stage employs a dual-front active contour (DFAC) model to conform the preliminary tumor surface to the PET radioisotope defined tumor volume. We will develop and optimize this two-stage model on phantom experiments, Monte Carlo simulated data, and finally on head-and-neck cancer patient studies, culminating in a prospective pilot study with spatially co-registered PET/CT images and pathology-based tumor volume correlation. Head- and neck cancer is ideal for developing and validating the segmentation process because it is anatomically complex, curable, it is possible to correlate anatomic and metabolic imaging findings with pathology, the tumors readily metabolize 18F-FDG, and accurate tumor segmentation is highly desirable due to the numerous surrounding critical structures. This proposal responds to the Program Announcement PA-06-371 titled """"""""In vivo Cancer Imaging Exploratory/Developmental Grants (R21)"""""""", particularly, the two topics """"""""in vivo cancer image displays and analyses"""""""" and """"""""in vivo image-guided cancer interventions"""""""". We will use the R21 mechanism to develop this promising technique and provide preliminary data for a subsequent R01 application. If this process works, it will provide to the radiation oncologist, for the first time, an objective, reliable, and accurate PET segmentation tool to aid them in delineating the tumor, improving the accuracy and consistency of radiation therapy treatment plans and ultimately improving the quality of radiation therapy care.
We propose to develop and validate reliable methods for delineating tumors in Positron Emission Tomography (PET) for patients with head and neck cancer. Improved tumor delineation may lead to improved quality and accuracy of radiation dose delivery. This may improve the local control, and eventually the survival of these cancer patients. ? ? ?
