Project Title: Identification of malignancy grade of
glial brain tumors using computer assisted image processing and analysis methods
Treatment of brain tumors originated from glial cells depends on the location of
the tumors and their malignancy grade. Location of the tumors within the brain can
be identified by neurodiagnostic scanners, while their malignancy grading, which
is a system defined by the World Health Organization, can only be identified by
histopathological analysis. Histopathology is based on the inspection of tumor’s
microscopic appearance and requires a sample of the tumor.
Sampling is achieved through a process known as biopsy that employs stereotactic
equipment or surgical resection and may damage the brain resulting in an irreversible
state in patient’s health. The pathologies in which surgery is not possible, such
as in the case of tumors are located in deeper areas of the brain, histopathology
of the tissues can not be analyzed and therefore tumor properties necessary for
treatment planning can not be understood.
The need for developing tools to support diagnosis processes of the brain tumors
is subjected to the considerable amount of research efforts in the image processing
and analysis literature. Existing approaches can be classified as Neural Network
based supervised and unsupervised learning oriented, spatial registration oriented,
and spatial prior probability oriented methods. Regardless of the problem solving
approach employed and neurodiagnostic data used, these existing methods are focused
either on the segmentation of the brain tissues or on the identification of the
pathologies from normal structures in the brain. There is not any method available
in the literature that has focused on the identification of malignancy grade of
brain tumors using computer assisted image processing and analysis methods.
Within the context of this work, MR images, taken from brain tumor patients where
pathological analysis results are also available are going to be digitized. The
modal and volumetric properties of the tumors, necrosis, contrast and non-contrast
areas and their ratio to the tumor area, existence of malign edema with its finger-like
shape, the ratio of edema to tumor, and the existence of ring and patchy enhanced
patterns of the tumor will be examined using linear and non-linear approaches of
computer assisted image processing and analysis methodology, and a digital medical
diagnostic tool will be developed in order to identify malignancy grade of tumors.
This tool is planned as a supplementary method in diagnosis processes intending
to identify and emphasize radiological reflections of histopathological tissue differentiations
digitally and mathematically. Numerical and quantitative values developed will be
repeatable, questionable, and measurable through basic statistical methods and will
be independent of human subjectivity.
The conservation of patient health and its sustainability will be the major contribution
of this work. Erroneous diagnosis, treatment, and irreversible deformation in patient’s
health that can be originated from human imperfection will be prevented and decision
making processes of medical doctors will be facilitated by providing human-independent,
objective, numerical results. Additionally, this work will also contribute in reduction
of surgery risks, in saving time, work load, and financial resources assigned by
the government for treatment, and in reduction of treatment costs.
This work will cover the identification of malignancy grade of the most frequent
glial brain tumors, classified by World Health Organization.