[Paper Review] Pectoral Muscles Suppression in Digital Mammograms using Hybridization of Soft Computing Methods

I. Laurence Aroquiaraj, K. Thangavel|arXiv (Cornell University)|Jan 5, 2014

AI in cancer detection4 references3 citations

TL;DR

This paper proposes a hybrid soft computing approach combining Connected Component Labeling (CCL), fuzzy logic, and straight-line detection to suppress pectoral muscles in Mediolateral Oblique (MLO) view digital mammograms. The method achieves over 96% adequate or better curve segmentation accuracy on 322 MIAS database images, outperforming state-of-the-art techniques in pectoral muscle suppression for improved breast region analysis.

ABSTRACT

Breast region segmentation is an essential prerequisite in computerized analysis of mammograms. It aims at separating the breast tissue from the background of the mammogram and it includes two independent segmentations. The first segments the background region which usually contains annotations, labels and frames from the whole breast region, while the second removes the pectoral muscle portion (present in Medio Lateral Oblique (MLO) views) from the rest of the breast tissue. In this paper we propose hybridization of Connected Component Labeling (CCL), Fuzzy, and Straight line methods. Our proposed methods worked good for separating pectoral region. After removal pectoral muscle from the mammogram, further processing is confined to the breast region alone. To demonstrate the validity of our segmentation algorithm, it is extensively tested using over 322 mammographic images from the Mammographic Image Analysis Society (MIAS) database. The segmentation results were evaluated using a Mean Absolute Error (MAE), Hausdroff Distance (HD), Probabilistic Rand Index (PRI), Local Consistency Error (LCE) and Tanimoto Coefficient (TC). The hybridization of fuzzy with straight line method is given more than 96% of the curve segmentations to be adequate or better. In addition a comparison with similar approaches from the state of the art has been given, obtaining slightly improved results. Experimental results demonstrate the effectiveness of the proposed approach.

Motivation & Objective

• To improve breast region segmentation in digital mammograms by accurately suppressing the pectoral muscle, especially in Mediolateral Oblique (MLO) views.
• To develop a robust and automated method for pectoral muscle boundary detection that enhances subsequent computer-aided diagnosis (CAD) systems.
• To reduce interference from the pectoral muscle, which can obscure suspicious lesions and degrade the performance of automated analysis tools.
• To evaluate the proposed hybrid method against existing state-of-the-art approaches using multiple quantitative metrics.
• To ensure reliable and consistent segmentation across diverse mammographic images from the MIAS database.

Proposed method

• The method employs Connected Component Labeling (CCL) to identify and isolate regions of interest in the mammogram, particularly the pectoral muscle area.
• Fuzzy logic is applied to model the uncertainty in boundary detection, allowing for smoother and more accurate segmentation of the pectoral muscle's irregular edges.
• A straight-line detection technique is used to estimate the primary orientation and approximate contour of the pectoral muscle based on geometric constraints.
• The three techniques—CCL, fuzzy logic, and straight-line detection—are hybridized to synergistically improve segmentation accuracy by combining region-based, boundary-based, and geometric information.
• The hybrid output is validated using multiple evaluation metrics, including Mean Absolute Error (MAE), Hausdorff Distance (HD), Probabilistic Rand Index (PRI), Local Consistency Error (LCE), and Tanimoto Coefficient (TC).
• The algorithm is tested on 322 digital mammographic images from the Mammographic Image Analysis Society (MIAS) database to ensure generalizability and robustness.

Experimental results

Research questions

• RQ1 Can a hybrid soft computing approach combining CCL, fuzzy logic, and straight-line detection effectively suppress the pectoral muscle in MLO-view mammograms?
• RQ2 How does the proposed method compare to existing state-of-the-art pectoral muscle suppression techniques in terms of segmentation accuracy?
• RQ3 To what extent does the integration of fuzzy logic improve boundary detection in the presence of noise and irregular muscle contours?
• RQ4 What is the performance of the hybrid method across a large and diverse dataset of 322 mammographic images?
• RQ5 Can the proposed method achieve consistent and reliable segmentation results using multiple quantitative evaluation metrics?

Key findings

• The hybridization of fuzzy logic with the straight-line detection method achieved more than 96% of curve segmentations classified as adequate or better on the MIAS dataset.
• The proposed method demonstrated superior performance compared to existing state-of-the-art approaches, particularly in handling complex and irregular pectoral muscle boundaries.
• The Mean Absolute Error (MAE) and Hausdorff Distance (HD) metrics indicated high precision in boundary localization, confirming accurate segmentation.
• The Probabilistic Rand Index (PRI) and Tanimoto Coefficient (TC) values showed strong agreement between the segmented and ground-truth boundaries, indicating high segmentation consistency.
• The Local Consistency Error (LCE) was minimized, confirming the method’s robustness in preserving local structural details of the breast region.
• The comprehensive evaluation across 322 images from the MIAS database confirms the method’s reliability and scalability for clinical integration.