br Published by Elsevier B V on behalf of Nalecz
© 2019 Published by Elsevier B.V. on behalf of Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences.
* Corresponding author at: SGGS Institute of Engineering & Technology, Nanded, Maharashtra 431606, India.
E-mail address: [email protected] (S. Sapate).
0208-5216/© 2019 Published by Elsevier B.V. on behalf of Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences.
Please cite this article in press as: Sapate S, et al. Breast cancer diagnosis using abnormalities on ipsilateral views of digital mammograms. Biocybern Biomed Eng (2019), https://doi.org/10.1016/j.bbe.2019.04.008
mance of CAD with proper balance of sensitivity and
73
1.
Introduction
specificity is crucial for accurate diagnosis and treatment
74
[15]. This becomes easy with CC and MLO views together. The
75
16
Q2 Early detection is viewed as the best hope to reduce the breast
recent studies have shown that the two view systems can
76
cancer mortality by allowing timely treatment at an earlier
improve the detection and diagnosis of breast abnormalities
77
cancer progression stage. The imaging techniques such as X-
and can become the effective alternative as a second reader to
78
ray, MRI, ultrasound, etc. are widely used for detection,
radiologists [16].
79
diagnosis, and treatment of abnormalities that can be
The CAD systems combining the advantages of single
80
cancerous [1]. Mammography is the most reliable cost-
view and two view systems can be useful as a simulta-
81
effective imaging modality used by choice for early detection
neous mammograms analysis tool for the radiologists [17].
82
of breast cancer [2]. Full-field digital mammography (FFDM)
Accordingly, we propose a different fusion based CAD
83
provides a high GW-311616 high-resolution images which are
system that leverages advantages of existing approaches
84
found useful in improving an efficacy and performance of
with improved case-based lesion detection and diagnosis
85
mammography [3]. Particularly, the FFDM images are more
at a comparable false positive (FP) rate. The dual view
86
significant in the mammographic screening of dense breasts of
analysis framework presented in this article is different
87
younger women [4]. The majority of breast cancers detected in
than other conventional ipsilateral mammograms analysis
88
symptomatic patients during screening mammography ap-
techniques as follows.
89
pear as masses within a range of 5–20 mm in diameter.
It selects the width of annular region dynamically based on
90
Visibility of masses smaller than 5 mm is rare and sometimes
921
the nonpalpable masses larger than 20 mm are not visible on
size of the lesion and the breast to match the corresponding
932
mammogram [5]. During mammography, two projection views
pairs of suspicious lesions from CC and MLO view.
943
namely CC and MLO views, known as ipsilateral views, are
Idea of template matching using selected set of similarity
9645
routinely obtained for each breast [6]. As ipsilateral views
features towards improving lesion pairing on ipsilateral
975
encompass the entire breast parenchyma from two different
views.
986
sides, the suspicious lesions, if are there, are visible and can be
It utilizes radiomics features and location-based features to
100997
detected on either or both the views [7]. Medical practitioners
improve film-based single view performance of cancer
10198
usually analyze ipsilateral views to describe the breast
detection.
10299
composition, morphological descriptors such as bilateral
It combines single view features and two view correspon-
1043
asymmetry, abnormal lesions, architectural distortion, micro-
dence score to improve the case-based two views perfor-
1051
calcifications, etc. Then they make their clinical decisions
mance of cancer diagnosis.
1062
regarding grading the abnormality or performing further
diagnostic tests. The mammogram interpretation may go
The remaining part of this article is structured as follows.
1084
wrong or inaccurate due to fatigue or more workload [8]. The
Section 2 gives an overview of existing CAD approaches using
1095
CAD system using ipsilateral views can be a second pair of eyes
ipsilateral views. Section 3 covers the clinical data used and
110
to diagnose the breast anomalies and improve the positive
the methodology is explained in Section 4. The experimental
111
predictions while reducing false predictions during diagnosis
results are analyzed in Section 5. The discussion on the various
112
key issues is presented in Section 6. Section 7 concludes the