Spatially Constrained Compound Magnification Framework for Histopathology Whole Slide Images
MetadataShow full item record
Histopathology whole slide images (WSIs) contain a considerable amount of valuable information required for diagnosis, prognosis, and treatment planning. These scans are the product of a contemporary technology of high-precision digitizers that seemed to aid in the clinical procedure. The availability of internet tools might improve treatment process in a variety of ways, like remotely communicating opinions amongst pathologists. Due to the significance of details in glass tissue slides, these images can be as big as 100,000 x 200,000 pixels and are often scanned in often three like red, green, and blue or luma component, blue-difference, and red-difference. Furthermore, WSIs are often kept in pyramidal magnification structures that include images of varying resolutions as well as clinical information. At higher magnification levels of the pyramid, a much greater quantity of information is stored. Although lower level magnification does not incorporate all information held at higher levels, it may considerably aid the clinical evaluation by producing high-density pixel information from lower resolution. The purpose of this research is to create a mechanism capable of recovering information at high-resolution such as 20X magnification, from lower-level magnifications such as 5X. This strategy reduces the amount of storage space and bandwidth needed for image transfer. Having models to reproduce the 20X magnification slides out of 5X magnification with high accuracy can considerably facilitate the adoption of digital pathology. one of the barriers to going digital is the high cost of storing many gigapixel WSIs. Transferring histopathology slides to a remote location can be facilitated by this method, while smaller size slides are easier for data sharing. Finally, in digitizing glass slides, scanners may not always correctly focus on the tissue, resulting in blurry portions negatively affecting diagnosis in certain cases. The same techniques may overcome the optical hurdles of obtaining crisp images at high magnifications. The suggested solutions would allow for the storing of digital slides in much smaller-scale versions while maintaining the reproducibility of high-resolution images. On the other hand, storing merely a lower magnification rather than the actual high magnification image and depending only on the software may not result in an adequate results. To improve the capabilities of the suggested approach even further, it is advisable to save some portions of the original high quality picture to guarantee a more accurate outcome. The second challenge that this research is focusing on is identifying the areas that may need further attention in order to store a part of them. The anomalous region of a slide is the piece that substantially proposes the diagnosis when defining the concern of an essential disease. A strategy for detecting this region is also explained here. To detect this area, a technique that can do it unsupervised is appropriate since it reduces pathologists' effort by not needing them to provide accurate annotations for a certain kind of cancer before using the approach. This thesis further relies on the unsupervised localization of an abnormal WSI region.
Cite this version of the work
Mehdi Afshari (2022). Spatially Constrained Compound Magnification Framework for Histopathology Whole Slide Images. UWSpace. http://hdl.handle.net/10012/18742