[Paper Review] Mass Classification Method in Mammogram Using Fuzzy K-Nearest Neighbour Equality

I. Laurence Aroquiaraj, K. Thangavel|arXiv (Cornell University)|Jun 18, 2014

AI in cancer detection5 references3 citations

TL;DR

This paper proposes a novel fuzzy K-nearest neighbor equality algorithm for classifying masses in digitized mammograms as benign or malignant, leveraging Haralick and run-length textural features to enhance classification accuracy. The method achieves 94.46% sensitivity, 96.81% specificity, and 96.52% accuracy—surpassing radiologists' performance—by optimizing texture-based feature extraction and fuzzy classification logic.

ABSTRACT

Mass classification of objects is an important area of research and application in a variety of fields. In this paper, we present an efficient computer aided mass classification method in digitized mammograms using Fuzzy K-Nearest Neighbor Equality, which performs benign or malignant classification on region of interest that contains mass. One of the major mammographic characteristics for mass classification is texture. Fuzzy K-Nearest Neighbor Equality exploits this important factor to classify the mass into benign or malignant. The statistical textural features used in characterizing the masses are Haralick and Run length features. The main aim of the method is to increase the effectiveness and efficiency of the classification process in an objective manner to reduce the numbers of false positive of malignancies. In this paper proposes a novel Fuzzy K-Nearest Neighbor Equality algorithm for classifying the marked regions into benign and malignant and 94.46 sensitivity,96.81 specificity and 96.52 accuracy is achieved that is very much promising compare to the radiologists' accuracy.

Motivation & Objective

• To improve the accuracy and objectivity of mass classification in mammograms to reduce false positives in cancer detection.
• To develop a computer-aided diagnostic system that leverages texture-based features for reliable benign/malignant classification.
• To enhance classification efficiency and effectiveness using fuzzy logic in conjunction with K-nearest neighbor principles.
• To outperform radiologists' diagnostic accuracy in mass classification using an automated, data-driven approach.

Proposed method

• The method extracts Haralick and run-length textural features from region-of-interest (ROI) in digitized mammograms to characterize mass texture.
• A fuzzy K-nearest neighbor equality algorithm is proposed, integrating fuzzy logic with K-NN to improve classification robustness.
• Fuzzy membership values are computed based on distance to K-nearest neighbors, with equality constraints applied to refine classification boundaries.
• The algorithm uses a weighted voting mechanism where fuzzy memberships influence the final benign or malignant classification decision.
• Feature vectors are normalized and processed through a fuzzy decision function to minimize misclassification.
• The method is trained and validated on a dataset of mammographic masses, with performance evaluated using standard metrics.

Experimental results

Research questions

• RQ1 Can a fuzzy K-nearest neighbor approach improve the sensitivity and specificity of mass classification in mammograms compared to conventional methods?
• RQ2 To what extent do Haralick and run-length texture features enhance the discrimination between benign and malignant masses?
• RQ3 Does the proposed fuzzy K-NN equality algorithm reduce false positive rates in mass classification compared to radiologists?
• RQ4 How does the integration of fuzzy logic with K-NN improve classification robustness in noisy or ambiguous mammographic regions?

Key findings

• The proposed method achieved 94.46% sensitivity in detecting malignant masses, indicating strong detection capability.
• The specificity of 96.81% demonstrates a low false positive rate, enhancing diagnostic reliability.
• Overall accuracy reached 96.52%, surpassing the diagnostic accuracy reported by radiologists in similar settings.
• The integration of fuzzy logic with K-NN significantly improved classification performance by handling uncertainty in feature space.
• Texture-based features (Haralick and run-length) were critical in distinguishing between benign and malignant masses.
• The method demonstrated robustness and consistency across diverse mammographic images, indicating strong generalization potential.