Automatic Dual-View Mass Detection in Full-Field Digital Mammograms
Mammography is the first-line modality for screening and diagnosis of breast cancer. Following the common practice of radiologists to examine two mammography views, we propose a fully automated dual-view analysis framework for breast mass detection in mam
- PDF / 1,864,692 Bytes
- 9 Pages / 439.363 x 666.131 pts Page_size
- 29 Downloads / 205 Views
Abstract. Mammography is the first-line modality for screening and diagnosis of breast cancer. Following the common practice of radiologists to examine two mammography views, we propose a fully automated dual-view analysis framework for breast mass detection in mammograms. The framework combines unsupervised segmentation and random-forest classification to detect and rank candidate masses in cranial-caudal (CC) and mediolateral-oblique (MLO) views. Subsequently, it estimates correspondences between pairs of candidates in the two views. The performance of the method was evaluated using a publicly available full-field digital mammography database (INbreast). Dual-view analysis provided area under the ROC curve of 0.94, with detection sensitivity of 87% at specificity of 90%, which significantly improved single-view performance (72% sensitivity at 90% specificity, 78% specificity at 87% sensitivity, P 0.5 in 96% of the images. Candidate mass classification with achieved an AUROC of 0.71, with sensitivity of 36% at specificity of 90% (Fig 3). Following learning-based candidate ranking, the 10 highest-ranked candidates included a true detection (Dice > 0.5) in 89% of the images. In 70% of the images the first-ranked candidate was a true detection, which corresponds to an average FPR of 0.3 per image. With 2.2 false-positives per image, the detection sensitivity was 83%. The AUROC of candidate classification by the single-view score was 0.92 (Fig. 3), with detection sensitivity of 72% at a specificity of 90%. Classification of the correspondence descriptors into true- and false pairwise matches provided an AUROC of 0.96, with optimal sensitivity and specificity of 89%
Automatic Dual-V View Mass Detection in Full-Field Digital Mammograms
49
and 96%, respectively. Following the one-to-one candidate assignment, correct ppairs of true CC-MLO mass can ndidates were found in 67% of the pairs, and the corrrect breast quadrant was estimatted in 77% of the cases. The combined dual-vieew score + improved the candidate classificaation compared to the single-view w score (Fig 3), with AUROC of 0.94 and detection seensitivity of 87% vs. 72%, at sp pecificity of 90% (P
Data Loading...
