Electronic Thesis and Dissertation Repository

Thesis Format



Master of Engineering Science


Biomedical Engineering


Khan, Ali


Focal cortical dysplasia (FCD) are localized regions of malformed cerebral cortex that are frequently associated with drug-resistant epilepsy. Currently, there is a lack of research towards providing quantitative methods for characterizing minor abnormalities in cortical architecture, hindering efforts to determine whether removal affects surgical outcome, and define potential imaging correlates. In our work, we have developed a tool to extract relevant features associated with cortical architectural abnormalities that can deal with artifacts including cortical layer distortions and morphological differences caused by cortical folding effects, and processing artifacts due to improper sectioning. This framework was applied to detect abnormalities across multiple subjects and slides using an unsupervised anomaly detection algorithm. Our results suggest that the technique is able to identify anomalies that correspond to visually-identifiable histological abnormalities. The frequency of abnormalities was found to differ among patients; however, the clinical significance of these findings is yet to be investigated.

Summary for Lay Audience

Drug-resistant epilepsy occurs in over 30% of epilepsy patients, with many patients undergoing surgical treatment to alleviate their seizure activity. Focal cortical dysplasia (FCD) is a common pathology associated with drug-resistant epilepsy and is mainly characterized by the presence of abnormal brain layering. However, FCD is challenging to diagnose and treat since lesions are not often seen with clinical imaging protocols. Furthermore, minor cortical architectural abnormalities are seen in many temporal lobe epilepsy cases, however, their significance is not well understood because of the challenges in objectively quantifying these abnormalities. The goal of this research is to develop analysis tools for histopathology of resected brain tissue to objectively and quantitatively evaluate cortical abnormalities in epilepsy patients. Quantitative evaluation of cortical architecture in histology samples would complement specialists in the detection of subtle epileptogenic lesions, reducing inter-rater variability and ultimately providing a more accurate reference for in-vivo diagnostic techniques.

A major challenge in analyzing cortical histology to quantitatively describe cortical architecture is that the layering of the brain can be distorted in different parts of the sample. This is caused by mainly two sources of variability: displacement of the brain layering due to the complex geometry of the brain, and out-of-plane sampling problems when the section is not perpendicular to the brain surface. These alterations can cause issues for computer-aided analysis, as many algorithms have difficulty in handling these artifacts.

In this study, we summarize our digitized tissue samples into relevant features that describe the cortical architecture. The result of the summarized features showed sensitivity to changes commonly associated with FCD pathology. Additionally, we implemented a method to extract relevant layering information computationally and applied computational techniques to align the displaced data across multiple slides and subjects. Furthermore, we apply a machine learning technique to identify and eliminate the processing artifacts presented in the data due to out-of-plane sampling problems. We then applied an outlier detection algorithm to test whether outliers in the data represent abnormal tissue in the brain. Our results suggest that anomalies in the summarized image features resemble abnormalities in the tissue samples.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.