The interpretation of mammograms can be difficult. Some breast cancers might be misclassified as benign lesions, and many benign lesions are sent for biopsy. The presentation of images with known pathology similar to an unknown case can be useful. The objective of this study was to investigate similarity measures for the selection of similar images in the distinction between benign and malignant lesions on mammograms. The images of masses and clustered microcalcifications were obtained from the Digital Database for Screening Mammography. In order to select similar images that are really similar and useful for radiologists, subjective similarity ratings for 300 pairs of masses and 300 pairs of clustered microcalcifications were obtained from ten breast radiologists, and the average ratings were employed as a "gold standard" for this study. The result indicated that the similarity measures determined based on pixel-value correlation of images and likelihood of malignancy of lesions were not very useful. The similarity measures based on the distances in the selected image feature space provided moderate correlations with the gold standard. The correlations were improved when similarity measures were determined by use of the ANN trained with the image features and subjective similarity ratings. The usefulness of similar images was evaluated in an observer study. Sixty cases were selected as unknown cases, and a set of benign and malignant images similar to each unknown image was selected based on the sizes and similarity measures. Eleven radiologists provided their confidence level regarding the malignancy of the lesions without and with the similar images. The result indicated that the observers' performances without and with similar images were comparable in terms of the area under the curve of the receiver operating characteristic analysis. However, in terms of the change in confidence level of malignancy, there were many cases in which the similar images had a beneficial effect to the observers. The presentation of similar images has a potential to increase radiologists' confidence and improve their diagnostic performance. For similar images to be useful to radiologists, the similar-image database and the selection scheme must be further improved and reevaluated in the future.
|School:||The University of Chicago|
|School Location:||United States -- Illinois|
|Source:||DAI-B 69/01, Dissertation Abstracts International|
|Keywords:||Breast cancer, Breast lesions, Computer-aided diagnosis, Mammograms|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be