A deep learning approach for the analysis of masses in mammograms with minimal user intervention

Dhungel, Neeraj, Carneiro, Gustavo and Bradley, Andrew P. (2017) A deep learning approach for the analysis of masses in mammograms with minimal user intervention. Medical Image Analysis, 37 114-128. doi:10.1016/j.media.2017.01.009

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
UQ482579_OA.pdf Full text (open access) application/pdf 11.23MB 0

Author Dhungel, Neeraj
Carneiro, Gustavo
Bradley, Andrew P.
Title A deep learning approach for the analysis of masses in mammograms with minimal user intervention
Journal name Medical Image Analysis   Check publisher's open access policy
ISSN 1361-8423
Publication date 2017-04-01
Year available 2017
Sub-type Article (original research)
DOI 10.1016/j.media.2017.01.009
Open Access Status File (Author Post-print)
Volume 37
Start page 114
End page 128
Total pages 15
Place of publication Amsterdam, Netherlands
Publisher Elsevier BV
Language eng
Subject 3614 Radiological and Ultrasound Technology
2741 Radiology Nuclear Medicine and imaging
1707 Computer Vision and Pattern Recognition
2718 Health Informatics
1704 Computer Graphics and Computer-Aided Design
Abstract We present an integrated methodology for detecting, segmenting and classifying breast masses from mammograms with minimal user intervention. This is a long standing problem due to low signal-to-noise ratio in the visualisation of breast masses, combined with their large variability in terms of shape, size, appearance and location. We break the problem down into three stages: mass detection, mass segmentation, and mass classification. For the detection, we propose a cascade of deep learning methods to select hypotheses that are refined based on Bayesian optimisation. For the segmentation, we propose the use of deep structured output learning that is subsequently refined by a level set method. Finally, for the classification, we propose the use of a deep learning classifier, which is pre-trained with a regression to hand-crafted feature values and fine-tuned based on the annotations of the breast mass classification dataset. We test our proposed system on the publicly available INbreast dataset and compare the results with the current state-of-the-art methodologies. This evaluation shows that our system detects 90% of masses at 1 false positive per image, has a segmentation accuracy of around 0.85 (Dice index) on the correctly detected masses, and overall classifies masses as malignant or benign with sensitivity (Se) of 0.98 and specificity (Sp) of 0.7.
Keyword Bayesian optimisation
Deep learning
Structured output learning
Transfer learning
Q-Index Code C1
Q-Index Status Provisional Code
Grant ID DP140102794
Institutional Status UQ

Document type: Journal Article
Sub-type: Article (original research)
Collections: HERDC Pre-Audit
School of Information Technology and Electrical Engineering Publications
Version Filter Type
Citation counts: TR Web of Science Citation Count  Cited 6 times in Thomson Reuters Web of Science Article | Citations
Scopus Citation Count Cited 8 times in Scopus Article | Citations
Google Scholar Search Google Scholar
Created: Tue, 28 Feb 2017, 00:20:24 EST by Web Cron on behalf of Learning and Research Services (UQ Library)