Presentation
Time:
Ph.D. Dissertation Defense: Mohammed Alali
Date:
11:30 am –
12:30 pm
Avery Hall
Room: 347
Target Audiences:
1144 T St
Lincoln NE 68508
Lincoln NE 68508
Additional Info: AVH
Ph.D. Dissertation Defense: Mohammed Alali
Wednesday, May 3, 2023
11:30 AM
347 Avery Hall
“Histopathology Image Classification using Machine Learning”
Pathology focuses on studying the causes and effects of diseases using different kinds of images, such as histopathology, cytopathology, and dermatopathology images. Histopathology images are used to study diseases from the underlying tissue architecture. Whole-Slide Images (WSIs) are very large histopathology images that are used in digital pathology. Slide images or WSIs are digital slides produced from the conventional glass slides used by pathologists. These large images can be of size 100,000 x 100,000 pixels and are mostly referred to as gigapixel images. The fact that each single slide image is of gigapixel size makes a dataset of even a few WSIs considered large-scale. Pathologists examine histopathology WSIs for cancer diagnosis. However, these diagnoses may vary across pathologists and also across time by the same pathologist. This introduces variability in the diagnosis process alongside the inherent variability in the appearance of tissue patterns. WSIs are increasingly being produced, but their pixel-level ground truth annotation, which is locating the cancerous pattern, is time consuming and laborious. On the other hand, pathologists’ availability may be limited and expensive. This poses a need for automatic processing of WSIs in order to achieve better and faster diagnosis. Applying machine learning in digital pathology provides a second opinion to pathologists that would reduce their workload. Additionally, it serves as a quality measure after pathologists’ recommendation and also can be used to train new pathologists. Incorporating diagnosis decision from machine learning alongside pathologists would produce more reliable results, since errors in both ways of diagnosis are not usually the same type. Deep learning has been applied successfully in computational pathology and shows immense improvement. In this PhD dissertation, we are interested in designing machine learning pipelines for classification of histopathology images.
Committee Members:
Dr. Jitender Deogun (Chair)
Dr. Lisong Xu
Dr. Juan Cui
Dr. Etsuko Moriyama
Wednesday, May 3, 2023
11:30 AM
347 Avery Hall
“Histopathology Image Classification using Machine Learning”
Pathology focuses on studying the causes and effects of diseases using different kinds of images, such as histopathology, cytopathology, and dermatopathology images. Histopathology images are used to study diseases from the underlying tissue architecture. Whole-Slide Images (WSIs) are very large histopathology images that are used in digital pathology. Slide images or WSIs are digital slides produced from the conventional glass slides used by pathologists. These large images can be of size 100,000 x 100,000 pixels and are mostly referred to as gigapixel images. The fact that each single slide image is of gigapixel size makes a dataset of even a few WSIs considered large-scale. Pathologists examine histopathology WSIs for cancer diagnosis. However, these diagnoses may vary across pathologists and also across time by the same pathologist. This introduces variability in the diagnosis process alongside the inherent variability in the appearance of tissue patterns. WSIs are increasingly being produced, but their pixel-level ground truth annotation, which is locating the cancerous pattern, is time consuming and laborious. On the other hand, pathologists’ availability may be limited and expensive. This poses a need for automatic processing of WSIs in order to achieve better and faster diagnosis. Applying machine learning in digital pathology provides a second opinion to pathologists that would reduce their workload. Additionally, it serves as a quality measure after pathologists’ recommendation and also can be used to train new pathologists. Incorporating diagnosis decision from machine learning alongside pathologists would produce more reliable results, since errors in both ways of diagnosis are not usually the same type. Deep learning has been applied successfully in computational pathology and shows immense improvement. In this PhD dissertation, we are interested in designing machine learning pipelines for classification of histopathology images.
Committee Members:
Dr. Jitender Deogun (Chair)
Dr. Lisong Xu
Dr. Juan Cui
Dr. Etsuko Moriyama