Home // ICSEA 2023, The Eighteenth International Conference on Software Engineering Advances // View article


Towards Improving Accurate Breast Cancer Diagnosis: Leveraging Pre-trained Convolutional Neural Network for Mammogram Analysis

Authors:
Marwa Ben Ammar
Dorra Zaibi
Faten Labbene Ayachi
Riadh Ksantini
Halima Mahjoubi

Keywords: Breast cancer; mammography; deep learning; YOLO; early diagnosis; one-class classification; three stages methodology; data imbalance

Abstract:
Breast cancer poses a significant global health challenge, emphasizing the need for improved diagnostic approaches for early diagnosis and intervention. Mammography, a widely used screening method, provides valuable insights into breast tissue anomalies. Nevertheless, its effectiveness is marred by error-prone interpretations and time-consuming analyses. To address this, our study introduces an innovative strategy to enhance breast cancer diagnosis by employing a Three-Stage One-Class You Only Look Once (YOLO) classification framework, harnessing the power of Deep Learning (DL). By incorporating the YOLO-v8 network, cutting-edge convolutional neural network (CNN) architecture, our proposed methodology aims to mitigate the shortcomings of conventional mammography interpretation. To assess the model's effectiveness, we utilize the Mammography Image Analysis Society (MIAS) dataset, which encompasses inherent data imbalances and intricacies. The framework we present is divided into three stages, each contributing to the refinement of the diagnostic process. Through the application of a one-class classification technique, our model effectively distinguishes between normal and abnormal mammograms. Furthermore, it offers a higher level of granularity by categorizing abnormalities into masses or calcifications. Additionally, the model can differentiate between benign and malignant cases, thereby facilitating precise clinical decision-making.

Pages: 116 to 121

Copyright: Copyright (c) IARIA, 2023

Publication date: November 13, 2023

Published in: conference

ISSN: 2308-4235

ISBN: 978-1-68558-098-8

Location: Valencia, Spain

Dates: from November 13, 2023 to November 17, 2023