Abstract:
This paper examines how magnification affects content-based image search in digital pathology archives and proposes multi-magnification image representation. Researchers and medical professionals can match patient records in large digital pathology slide archives and learn from evidently diagnosed and treated cases. Pathologists use microscopes to find and evaluate morphological features by switching magnification levels. To bridge the gap between AI-enabled image search methods and clinical settings, we investigated digital pathology magnification levels and combinations. The proposed searching framework can search millions of raw whole slide images without regional annotation. This paper compares two magnification level combination methods. The first approach uses a single-vector deep feature representation for a digital slide, while the second uses a multi-vector representation. A subset of The Cancer Genome Atlas (TCGA) repository was searched using 20×, 10×, and 5× magnifications. The experiments show that diagnostics require cell-level information at the highest magnification. Low-magnification information may improve this assessment depending on tumor type. Our multi-magnification image search improved F1-scores by up to 11% for urinary tract and brain tumor subtypes compared to single-magnification.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
