Medical image similarity perception for content-based image retrieval
Abstract/Contents
- Abstract
- Content-based image retrieval (CBIR) of perceptually similar images for medical decision support may improve the accuracy and efficiency of radiological diagnosis. A robust reference standard for perceptual similarity in medical images is vital to CBIR applications, but is challenging to create due to the large amount of observer data necessary as well as high inter-observer variability. The aims of this thesis are (1) to develop techniques for creating reference standards based on perceptual similarity that are scalable for large databases, and (2) to model inter-reader variability in observer assessments of image similarity. This thesis first discusses a novel technique for predicting visual similarity for pairs of images in a database from perceptual data obtained by viewing each image individually. In an observer study using 19 CT liver lesions in the portal venous phase, three radiologists provided point-wise ratings for visual attributes of images containing liver lesions displayed individually and also for perceptual similarity in all pair-wise combinations of these images. Using the data generated from viewing images individually, this work develops a scheme to generate ratings of pair-wise similarity using linear fitting, and demonstrates that this is a scalable technique for developing a reference standard. This work also develops a statistical model from these data that may be used to predict inter-reader variability and accuracy of similarity estimates in larger sets of data. Finally, this thesis discusses leveraging compressive sensing techniques, which may be used to complete highly sparse matrices, to impute similarity matrices from a small subset of observer ratings. By being scalable for large databases, this technique may be a reliable and efficient method for creating similarity reference standards. The results of this work thus provide a better understanding of perception of medical image similarity, which may in turn be used to train and validate medical decision support systems.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2014 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Faruque, Jessica Suzanne |
|---|---|
| Associated with | Stanford University, Department of Electrical Engineering. |
| Primary advisor | Napel, Sandy |
| Primary advisor | Van Roy, Benjamin |
| Thesis advisor | Napel, Sandy |
| Thesis advisor | Van Roy, Benjamin |
| Thesis advisor | Rubin, Daniel (Daniel L.) |
| Advisor | Rubin, Daniel (Daniel L.) |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Jessica Suzanne Faruque. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2014. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2014 by Jessica Suzanne Faruque
- License
- This work is licensed under a Creative Commons Attribution Non Commercial No Derivatives 3.0 Unported license (CC BY-NC-ND).
Also listed in
Loading usage metrics...