Hybrid local descriptor for improved detection of masses in mammographic CAD systems
by Devi Vijayan; R. Lavanya
International Journal of Advanced Intelligence Paradigms (IJAIP), Vol. 25, No. 1/2, 2023

Abstract: Early detection of breast cancer increases the chances of survival considerably. Computer aided detection (CAD) systems are assistive tools which can render cost-effective, quick and objective decision making. The proposed work addresses the development of a CAD system to automatically classify suspicious regions in mammograms as normal and abnormal, aiming to improve the detection of masses. To this end, we propose to fuse two local descriptors, namely SIFT based bag of words (BoW) and uniform local binary pattern (ULBP), using principal component analysis (PCA). The effect of issues including obscured masses, typical of dense breasts, is alleviated by the use of segmentation-devoid local descriptors. Experiments on the benchmark DDSM database with multilayer perceptron (MLP) classification, on features extracted from 4,316 suspicious regions demonstrate the efficiency of proposed system. The obtained result significantly reduces false positives and false negatives, achieving an accuracy of 92.81% with an F1 score of 0.91.

Online publication date: Thu, 11-May-2023

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Advanced Intelligence Paradigms (IJAIP):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com