Determination of similarity measures for pairs of mass lesions on mammograms by use of BI-RADS lesion descriptors and image features.
ABSTRACT To determine similarity measures for selection of pathology-known similar images that would be useful for radiologists as a reference guide in the diagnosis of new breast lesions on mammograms.
The images were obtained from the Digital Database for Screening Mammography developed by the University of South Florida. For determination and evaluation of similarity measures, the "gold standard" of similarities for 300 pairs of masses was determined by 10 breast radiologists. For determining similarity measures that would agree with radiologists' similarity determination, an artificial neural network (ANN) was trained with the radiologists' subjective similarity ratings and the image features. The image features were determined subjectively using the Breast Imaging Reporting and Data System (BI-RADS) lesion descriptors and objectively by computerized image analysis. The similarity measures determined by the ANN were compared to the gold standard and evaluated in terms of the correlation coefficient.
The similarity measures determined using the BI-RADS descriptors only were not as useful as those determined by use of the image features only. When the BI-RADS margin ratings were combined with the image features, the correlation coefficient between the subjective ratings and the objective measures improved slightly (r = 0.76) compared to those based on the image features alone (r = 0.74).
The inclusion of the BI-RADS margin descriptors may be useful for determination of similarity measures, especially when it is difficult to obtain the manual outlines of the masses and if the BI-RADS descriptors were provided consistently by radiologists.
- SourceAvailable from: Daniela Stan Raicu[Show abstract] [Hide abstract]
ABSTRACT: There is considerable research in the field of content-based medical image retrieval; however, few of the current systems investigate the relationship between the radiologists' visual impression of image similarity and the computer calculated content-based similarity. Furthermore, those research studies that investigate these relationships analyze the visual similarity with respect to degree of malignancy without including specific characteristics that are important in the diagnosis process. The creation of the NIH/NCI Lung Image Database Consortium (LIDC) dataset offers the opportunity to perform the proposed research. Each nodule out of the 932 distinct nodules (larger than 3mm in diameter) was delineated and annotated by up to four radiologists using nine semantic characteristics that are important in the lung nodule interpretation process. Using the LIDC images, we propose to encode the radiologists' characteristic-based similarity and further discover if there is any relationship between this conceptual/characteristic-based similarity and the content-based similarity for lung nodule interpretation. Our preliminary results show that it is a challenging problem to model the characteristic-based and content-based relationships for a broad category of lung nodules. A correlation of only 0.1 was obtained between the characteristic-based similarity and the predicted characteristic-based similarity using an artificial neural networked trained on four types of low-level image features (size, intensity, shape, and texture) calculated for 640 random pairs of nodules. Future research is necessary to investigate the appropriateness of the considered image features to model both the variation in the human interpretation of the lung nodules and the perceived characteristic-based similarity.Proceedings of the 11th ACM SIGMM International Conference on Multimedia Information Retrieval, MIR 2010, Philadelphia, Pennsylvania, USA, March 29-31, 2010; 01/2010
- [Show abstract] [Hide abstract]
ABSTRACT: We have been developing a computerized scheme for selecting visually similar images that would be useful to radiologists in the diagnosis of masses on mammograms. Based on the results of the observer performance study, the presentation of similar images was useful, especially for less experienced observers. The test cases included 50 benign and 50 malignant masses. Ten observers, including five breast radiologists and five residents, were asked to provide the confidence level of the lesions being malignant before and after the presentation of similar images. By use of multireader, multi-case receiver operating characteristic analysis, the average areas under the curves for the five residents were 0.880 and 0.896 without and with similar images, respectively (p=0.040). There were four malignant cases in which the initial ratings were relatively low, but the similar images alerted the residents to increase their confidence levels of malignancy close to those by the breast radiologists. The presentation of similar images may cause some observers falsely to increase their suspicion for some benign cases; however, if similar images can alert radiologists to recognize the signs of malignancy and also help them to decrease their suspicion correctly for some benign cases, they can be useful in the diagnosis on mammograms.Proceedings of SPIE - The International Society for Optical Engineering 02/2009; DOI:10.1117/12.811447 · 0.20 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: There is considerable research in the field of content-based image retrieval (CBIR); however, few of the current systems incorporate radiologists' visual impression of image similarity. Our objective is to bridge the semantic gap between radiologists' ratings and image features. We have been developing a conceptual-based similarity model derived from content-based similarity to improve CBIR. Previous work in our lab reduced the Lung Image Database Consortium (LIDC) data set into a selection of 149 images of unique nodules, each containing nine semantic ratings by four radiologists and 64 computed image features. After evaluating the similarity measures for both content-based and semantic-based features, we selected 116 nodule pairs with a high correlation between both similarities. These pairs were used to generate a linear regression model that predicts semantic similarity with content similarity input with an R 2 value of 0.871. The characteristics and features of nodules that were used for the model were also investigated.Proceedings of SPIE - The International Society for Optical Engineering 03/2010; DOI:10.1117/12.844507 · 0.20 Impact Factor