Name
Title | ||
|---|---|---|
A method for lesion visibility prediction in mammograms by local analysis of spectral anatomical noise |
Abstract | ||
|---|---|---|
Detection of mass lesions in mammograms via visual readings is a challenging task, and the radiographic density of the breast tissue or its strong anatomical structure may render lesions completely invisible. In order to assess visibility of lesions of a certain size in a given mammogram, we propose a measure for prediction of lesion visibility that complements established approaches for breast density assessment by taking also local structure into account. This measure is based on the analysis of spectral anatomical noise in terms of local standard deviation values for several frequency bands of the mammogram. The resulting values are used to generate two dimensional visibility maps for different lesion sizes. Phantoms of structured tissue equivalent materials were imaged using a full-field digital mammography (FFDM) system, and spherical lesions of different sizes were artificially added to the images. In an observer study with ten observers visibility thresholds were determined from a total of 290 simulated lesions. The resulting nonlinear threshold curve was verified in a second observer study, where 66 lesions were artificially added in clinical mammograms of varying breast density according to BI-RADS classification. A prediction accuracy of 92% was obtained, suffering mostly from different image characteristics in the breast tissue regions near the skinline or the pectoral muscle. |
Year | DOI | Venue |
|---|---|---|
2012 | 10.1007/978-3-642-31271-7_71 | Digital Mammography / IWDM |
Keywords | Field | DocType |
observers visibility threshold,breast tissue,different lesion size,dimensional visibility map,observer study,lesion visibility,local analysis,spectral anatomical noise,breast density assessment,lesion visibility prediction,varying breast density,breast tissue region,different image characteristic | Digital mammography,Computer vision,Visibility,Lesion,Local structure,Radiography,Artificial intelligence,Observer (quantum physics),Local analysis,Standard deviation,Medicine | Conference |
Citations | PageRank | References |
0 | 0.34 | 2 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Stephanie Simbt | 1 | 0 | 0.68 |
| Hanns-Ingo Maack | 2 | 0 | 0.68 |
| Harald S. Heese | 3 | 22 | 4.77 |