NMR Characterization of Pathological Disease States: Monitoring Response to Single-Dose Radiotherapy in a RIF-1 Tumor Model and the Role of Spreading Depression in the Evolution of Ischemic Stroke: a Dissertation

Publication Date

May 2005

Document Type

Doctoral Dissertation


Graduate School of Biomedical Sciences


Magnetic Resonance Imaging; Neoplasms; Cerebrovascular Accident; Spreading Cortical Depression; Academic Dissertations


Part 1: Monitoring Response to Single-Dose (1000cGy) Radiotherapy in a RIF-1 Tumor Model

The current standard of measure for monitoring chemotherapeutic and radiotherapeutic treatment response is tumor volume. Unfortunately, changes in tumor volume are generally slow and tumor volume does not necessarily indicate the degree of tumor viability. The development of marker(s) with the ability to detect an early therapeutic response would greatly aid in patient management, opening the possibility for both rapid dose optimization and replacement of ineffective therapies with alternative treatment. Previous studies have shown that diffusion measurements using magnetic resonance imaging (MRI) techniques are sensitive to therapy-induced changes in cellular structure, allowing demarcation between regions of necrosis and viable tumor tissue. This sensitivity, based on the correlation between water apparent diffusion coefficient (ADC) values and tumor cellular density, may allow diffusion measurements to be employed in non-invasive monitoring of treatment response. Therapy-induced increases in tumor ADC preceding tumor regression have been reported in a variety of experimental tumor models and several human brain tumors. Despite the demonstrated diffusion sensitivity to therapeutic response in these particular studies, shortcomings still remain that hinder the efficacy of clinical application in oncology. Earlier studies have concentrated on the mean ADC present within the tumor, either within the entire tumor volume or a region of-interest (ROI) defined by the user, and their evolution pre-treatment and post-treatment. Because of inter- and intra-tumor heterogeneity, volume-averaged ADC measurements suffer from poor correlation with treatment efficacy. In addition, most studies make little or no attempt to characterize the entire tumor volume (necrotic, viable, edema). The identification of regions of differing tissue viability should aid in the staging of treatment, therefore making accurate and reproducible tissue segmentation an important goal.

The results of earlier, single-parameter studies indicate that a multi-parametric approach in which several MR parameters are monitored (ADC, T2, M0) may provide greater power than that of the single parameter approach. A multi-parametric or multi-spectral (MS) analysis uses pattern-recognition techniques, such as clustering, for image segmentation. Clustering algorithms use characteristics of the multiple MR-parameter dataset to group tissue of similar type, e.g., fat, muscle, viable tumor, necrosis. Specifically, k-means (KM) is an unsupervised segmentation algorithm that groups together similar tissue based on the difference in MR parameter space between the image voxel of interest and the mean parameter values of the voxels in that cluster. In the first step of the classification algorithm, it is applied to separate the data into two clusters (k = 2), tissue and background noise voxels. All voxels classified as background noise are set to zero and removed from further processing. In the second step, KM is applied to the remaining tissue voxels to segment the data into multiple tissue types. In the case of tumors, it is not clear in advance how many different types of tissue exist. The number of clusters, k, should be varied to ensure that all relationships between tissues are found. In the final step, the resulting KM maps may be compared to histological slices taken from the same tissue as the imaging slices in order to identify the tissue type of each cluster.

In line with the studies and analyses described above, quantitative MRI was performed to investigate the spatial correlation between ADC, spin-spin (T2) relaxation times, and proton density (M0) in murine radiation-induced fibrosarcoma (RIF-1) tumors following single-dose (l000cGy) radiotherapy using the KM algorithm (Chapters 3 and 4) and different combinations of features and/or clusters. For all cluster/feature combinations, an in-depth comparison between KM-derived volume estimates and conventional histology via the hematoxylin-eosin (H&E) staining procedure (for identification of viable tumor versus necrosis), as well as via hypoxic-inducible factor-lα. (HIF-1α) immunohistochemistry (for identification of regions of hypoxia versus well-oxygenated tissue) was performed (Chapter 3). The optimal cluster/feature combination was determined by minimizing the sum-of-squared-differences (SSD) between the actual datapoints and the ideal one-to-one correlation that should exist between KM-derived volume estimates and histology-derived volume estimates. The optimal cluster/feature combination was determined to be a 2-feature (ADC, T2) and 4-c1uster (2 regions each of viable tissue and necrosis) segmentation. This KM method was then applied in analysis of the radiotherapeutic response: first, to gain insight into the various processes whose combination yield the total ADC response over time; second, to identify the contribution of tissue heterogeneity to the treatment response and changes in tumor growth kinetics and cell kill (Chapter 4). Comparisons between control and various time-points out to 14 days post-radiotherapy permitted more accurate tissue characterization and prediction of therapeutic outcome over analysis using ADC alone.

The results based on histological validation demonstrated: (1) MS analysis provides an improved tissue segmentation method over results obtained from conventional methods employing ADC alone; (2) MS analysis permits subdivision based on the degree of necrosis, as well as delineation between well-oxygenated and hypoxic viable tissue; and (3) Individual KM volumes corresponded well with both H&E volumes and regions with increased HIF-1α expression. The results based on the radiotherapeutic response demonstrated: (1) MS analysis provides a method for monitoring the range of tissue viability as a function of time post-treatment; (2) MS analysis permits assessment of the various contributions to the total ADC response post-treatment; (3) The relative fractions of well-oxygenated (i.e., radiosensitive) versus hypoxic (i.e., radioresistant) tissue pretreatment may be predictive of treatment response; and (4) The early ADC increase did not seem to be a result of radiation-induced vasogenic edema, but instead was most likely due to a slight reduction in cellular density following therapy. These studies provide a non-invasive method of tissue characterization that may be used in monitoring treatment response and optimizing drug dose-timing schemes, with the potential for predicting treatment efficacy.

Part 2: Role of Spreading Depression in Ischemic Stroke

Stroke is a prevalent disease that ranks as the 3rd leading cause of death and disability in the United States, according to NIH statistics, costing millions of dollars in medical costs and lost wages. At present, the mechanism by which focal ischemia evolves into infarction remains poorly understood. By determining the patho-physiological mechanisms involved in the evolution of focal brain ischemia, therapeutic strategies may be designed for instances of acute ischemic stroke. In the late 1980s, researchers discovered MRI techniques that allow the detection of stroke very early after onset. Such techniques as diffusion-weighted imaging and perfusion-weighted imaging (DWI and PWI) have been applied both clinically and experimentally. Previous studies employing these techniques suggest that cortical spreading depression plays a detrimental role in the evolution of focal brain ischemia. Spreading depression (SD) is characterized by a spontaneous and reversible depression of cortical electrical activity that spreads from the site of onset as a wave with a speed of 2-5 mm/min. It is accompanied by an ionic redistribution, with efflux of potassium ions (K+) and influx of sodium, chloride, and calcium (Na+, Cl-, Ca2+) ions, as well as water. This results in cellular swelling and a decreased extracellular space (ES), yielding a decline in ADC. A positive correlation between the number of both spontaneous and induced SDs and infarct volume has well been documented, supporting the idea that SD inhibition might be neuroprotective if initiated early after ischemic onset.

Even though these studies show promise in their ability to track SD using diffusion mapping, changes in ADCs reflect cytotoxic edema and do not necessarily correspond to SD or SD-like depolarizations or calcium (Ca2+) influx, leading to cell death. Recent studies have reported the use of manganese ions (Mn2+) as a depolarization-dependent contrast agent in monitoring brain activation through the application of glutamate, as well as in the study focal ischemia. Since extracellular accumulation of potassium (K+) ions or glutamate in ischemic tissue is believed to play a central role in the initiation and propagation of SDs, and knowing that Mn2+, having an ionic radius similar to that of Ca2+, is handled in a manner similar to Ca2+, these studies suggest the possible use of manganese ions (Mn2+) in tracking SD or SD-like depolarizations in the evolution of focal brain ischemia.

In order to determine the utility of Mn2+ as a marker for SD, two sets of T1-weighted MRI experiments were performed before applying Mn2+ in an experimental stroke model (Chapter 6). First, for verification purposes, a glutamate administration group was evaluated to validate our use of the manganese-enhanced MRI (MEMRI) method previously developed by Aoki et al, a modification of the original by Lin and Koretsky. When satisfied that the contrast enhancement was specific to glutamate only, a second set of experiments was performed. Here, experimental SD was elicited by chemical stimulation (direct application of concentrated potassium chloride [KC1] on the exposed cortical surface) and compared with control conditions (perfusion of sodium chloride [NaC1] on exposed brain cortex). This study demonstrated: (1) Mn2+, specific to Ca2+ channel activity, is a more accurate marker for SD than DWI or T2 methods; (2) Cortical restriction of MEMRI enhancement supports the contention that apical dendrites are necessary for SD propagation; (3) Subcortical enhancement is a result of corticalsubcortical neuronal connectivity; and (4) Because of the relatively slow clearance of Mn2+, MEMRI permits higher spatial resolution and signal-to-noise ratios (SNRs) via increased signal averaging. Based on these results, preliminary experiments involving the study of SD in focal ischemia using Mn2+ were performed (Chapter 7). Initial results indicate: (1) MEMRI of ischemia, when compared with standard DWI/PWI methods, may provide a method for estimating the likelihood of progression to infarct at acute time points post onset of stroke. These studies provide a foundation for further investigation into the role of SD in stroke, and the application of Mn2+ towards the design of therapeutic strategies targeting SD inhibition.

Conclusions and Medical Significance

The research within this dissertation employed magnetic resonance imaging techniques for monitoring the temporal evolution of pathological disease states such as focal ischemia and cancer, with and without therapeutic intervention. Optimization of these techniques in experimental models will open the possibility for future application in a clinical setting. Clinical availability of these non-invasive methods, with the ability to detect an early therapeutic response or to provide staging and prediction of tissue fate, would greatly aid in patient management of both cancer and stroke.


In the process of seeking author's permission to provide full text.

Rights and Permissions

Copyright is held by the author, with all rights reserved.

This document is currently not available here.