Date of Award

Fall 2017

Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical/Computer Engineering

Committee Director

Khan M. Iftekharuddin

Committee Member

Dean Krusienski

Committee Member

Jiang Li

Committee Member

Yaohang Li


This dissertation proposes novel texture feature-based computational models for quantitative analysis of abnormal tissues in two neurological disorders: brain tumor and stroke. Brain tumors are the cells with uncontrolled growth in the brain tissues and one of the major causes of death due to cancer. On the other hand, brain strokes occur due to the sudden interruption of the blood supply which damages the normal brain tissues and frequently causes death or persistent disability. Clinical management of these brain tumors and stroke lesions critically depends on robust quantitative analysis using different imaging modalities including Magnetic Resonance (MR) and Digital Pathology (DP) images. Due to uncontrolled growth and infiltration into the surrounding tissues, the tumor regions appear with a significant texture variation in the static MRI volume and also in the longitudinal imaging study. Consequently, this study developed computational models using novel texture features to segment abnormal brain tissues (tumor, and stroke lesions), tracking the change of tumor volume in longitudinal images, and tumor grading in MR images. Manual delineation and analysis of these abnormal tissues in large scale is tedious, error-prone, and often suffers from inter-observer variability. Therefore, efficient computational models for robust segmentation of different abnormal tissues is required to support the diagnosis and analysis processes. In this study, brain tissues are characterized with novel computational modeling of multi-fractal texture features for multi-class brain tumor tissue segmentation (BTS) and extend the method for ischemic stroke lesions in MRI. The robustness of the proposed segmentation methods is evaluated using a huge amount of private and public domain clinical data that offers competitive performance when compared with that of the state-of-the-art methods. Further, I analyze the dynamic texture behavior of tumor volume in longitudinal imaging and develop post-processing frame-work using three-dimensional (3D) texture features. These post-processing methods are shown to reduce the false positives in the BTS results and improve the overall segmentation result in longitudinal imaging. Furthermore, using this improved segmentation results the change of tumor volume has been quantified in three types such as stable, progress, and shrinkage as observed by the volumetric changes of different tumor tissues in longitudinal images. This study also investigates a novel non-invasive glioma grading, for the first time in literature, that uses structural MRI only. Such non-invasive glioma grading may be useful before an invasive biopsy is recommended. This study further developed an automatic glioma grading scheme using the invasive cell nuclei morphology in DP images for cross-validation with the same patients. In summary, the texture-based computational models proposed in this study are expected to facilitate the clinical management of patients with the brain tumors and strokes by automating large scale imaging data analysis, reducing human error, inter-observer variability, and producing repeatable brain tumor quantitation and grading.



Available for download on Thursday, October 18, 2018