Pre-Treatment Prediction of Brain Tumors’ Response to Radiation Therapy Using Tumor Complexity Analysis in MRI

Several magnetic resonance (MR) methods have been suggested recently as having potential for prediction of tumor response to treatment.

Several magnetic resonance (MR) methods have been suggested recently as having potential for prediction of tumor response to treatment. Contrast-enhanced MRI has been shown to be able to reveal distinct tumor patterns that can serve as a predictor of response to chemotherapy in human breast cancer (Esserman et al, 2001). Dynamic contrast MRI has been shown to be useful in characterizing the microvasculature of tumors and has shown potential in predicting response to antiangiogenic treatments (Neeman et al, 2003).


The problem
Studies has shown that there is a large correlation between the spatial structure complexity of the tumor and it’s response to treatment. Linguistic complexity introduced by the researchers a decade ago (Trifonov, 1990), is a highly intuitive notion for this problem. In our project we introduce a Morphological Image analysis of this problem.


Project goals & The solution
Our goals where: First, analysis of T1 imaging slices for identification of the tumor borderline. Second, analysis of T2 imaging slices for analysing spatial structure complexity in 3D using Morphological Image analysis. From this we evaluated a Hi factor indicating the tumor comlexity.
Here we demonstrate the stages of the algorithm that identifies of the tumor borderline:



We proposed 3 algoritms for analysing the tumor in 3D using Morphological Image analysis:

1. “Euler Algorithm” – the main idea of this algorithm is that the more uniform the tumor it has less complexity. The outcome is determined by the number of Morphological delations and Euler number between sequential delated images.

2. “Sources Algorithm” – the main idea is that a comlelx tumor is an outcome of a large number of sources relative to it’s Value size (in 3D). Here we used on the tumor Morphological Ultimate Erosions in 3D to allocate the sources of the tumer and Euler number for counting them.

3. “Substances Algorithm” – this was a more biological approach, the idea is that cancer cells near necrotic regions may experience hypoxic conditions and therefore are less sensitive to treatment. Necrosis spread over several regions in the tumor increases its heterogeneity and will have a larger surface area than a single necrotic core. The larger surface area will consist of a larger number of slow metabolizing cells.
We defined 4 measurements for this notion:
a. Face Area.
b. Number of objects.
c. Number of Different sizes of objects.
d. Value.
Then we characterised the Necrosis as a bands is the 3D histogram of the tumer gray scale levels. This is done by the user as so:

Finally we used pirson correlation on those measurements with a pre-defined Homogeny and Heterogeny tumors to determine it’s complexity. Here we can see figures of the 4 measurements vs. it’s substances:



This project was implemented entirely in Matlab, version 6.5 and 7.1. The T1 & T2 images where taken from Shiba institute.


In this project we see that morphological image Analysis on 3D tumors can be effective when detemining response to treatments and by this helping the doctor decision.


We are grateful to our project supervisor Ehud Orian for his help and guidance throughout this work.
We are also grateful to the Ollendorf Minerva Center Fund for supporting this project.