Any tumor, whether malignant or benign, within the rigid confines of the skull is potentially fatal if it continues to enlarge at the expense of the volumetric requirement of the brain. Thus, it is important to be able to determine the lesions location, size, growth rate, and architecture for diagnosis, treatment, and most importantly, for determining the prognosis for the patient. Magnetic resonance imaging (MRI) has proven to be useful in demonstrating the three-dimensional relationship of a tumor to surrounding anatomical structure. The use of MRI for segmentation and quantitative analysis is of current interest in delineation of brain tumors, however its clinical feasibility has not been established. An image processing technique (Eigenimage Filtering) has been developed and reported. This technique has the ability to use multiparameter information present in a sequence of MR images of the same anatomical site, to segment an object of interest from other objects. Thus, improving its visualization, the ease of obtaining morphological measurements of the object, and the reproducibility and accuracy of these measurements. The overall objective of this proposal is to establish the clinical feasibility of this technique to the segmentation of brain tumors from normal brain tissue for the purpose of delineation of the lesions and its sub regions (architecture). The objective of this proposal can be partitioned into; 1.) validation of the segmentation techniques: and 2.) application of the technique to tumor volume measurements and guiding biopsy sampling. The study will be conducted utilizing an interactive semi-automatic version of the Eigenimage Filtering software. The validation of the technique will be through the comparison of the segmented lesion images (eigenimages) with histology images of an animal (rat) tumor model and with comparison of the eigenimages with biopsy results from selected clinical cases. The eigenimage technique will be used to segment lesions, sub-regions of the lesions and regions of partial volume averaging between the lesion and normal tissue. These eigenimages will be used to determine the volume of the lesion, its sub regions, and regions of partial volume averaging, and to determine the spatial coordinates for taking biopsy samples. The advantage of the Eigenimage Filter technique is that it is based upon solid mathematical justification and is the only linear filter that provides partial volume information. The ability to segment objects of interest and apply image analysis techniques to them can significantly improve the diagnostic, treatment planning, and treatment evaluation utilizing medical imaging.
