eduzhai > Applied Sciences > Engineering >

On segmentation of pectoralis muscle in digital mammograms by means of deep learning

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: Computer-aided diagnosis (CAD) has long become an integral part ofradiological management of breast disease, facilitating a number of importantclinical applications, including quantitative assessment of breast density andearly detection of malignancies based on X-ray mammography. Common to suchapplications is the need to automatically discriminate between breast tissueand adjacent anatomy, with the latter being predominantly represented bypectoralis major (or pectoral muscle). Especially in the case of mammogramsacquired in the mediolateral oblique (MLO) view, the muscle is easilyconfusable with some elements of breast anatomy due to their morphological andphotometric similarity. As a result, the problem of automatic detection andsegmentation of pectoral muscle in MLO mammograms remains a challenging task,innovative approaches to which are still required and constantly searched for.To address this problem, the present paper introduces a two-step segmentationstrategy based on a combined use of data-driven prediction (deep learning) andgraph-based image processing. In particular, the proposed method employs aconvolutional neural network (CNN) which is designed to predict the location ofbreast-pectoral boundary at different levels of spatial resolution.Subsequently, the predictions are used by the second stage of the algorithm, inwhich the desired boundary is recovered as a solution to the shortest pathproblem on a specially designed graph. The proposed algorithm has been testedon three different datasets (i.e., MIAS, CBIS-DDSm and InBreast) using a rangeof quantitative metrics. The results of comparative analysis show considerableimprovement over state-of-the-art, while offering the possibility of model-freeand fully automatic processing.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...