Binary tunicate swarm algorithm based novel feature selection framework for mammographic mass classification

The efficacy of a CAD system for mammographic mass classification depends on high-quality features for maximum throughput. Feature selection is crucial for accurate breast cancer diagnosis in CAD solutions. Leveraging the success of metaheuristics, a novel binary tunicate swarm algorithm is proposed...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Measurement : journal of the International Measurement Confederation Jg. 235; S. 114928
Hauptverfasser: Laishram, Romesh, Rabidas, Rinku
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.08.2024
Schlagworte:
Binary tunicate swarm algorithm
CAD system
Feature selection
Mammographic mass categorization
Feature selection
Binary tunicate swarm algorithm
Mammographic mass categorization
CAD system
ISSN:0263-2241
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The efficacy of a CAD system for mammographic mass classification depends on high-quality features for maximum throughput. Feature selection is crucial for accurate breast cancer diagnosis in CAD solutions. Leveraging the success of metaheuristics, a novel binary tunicate swarm algorithm is proposed for feature selection. The framework incorporates pre-trained deep learning architectures as feature extractors, alongside the proposed feature selection and classifier, for mass categorization. The proposed feature selection strategy also emphasizes the design of new fitness functions tailored to the specific characterization of suspicious mass lesions. Experiments on benchmark databases (mini-MIAS and DDSM) yield sensitivities of 95.31% and 92.18%, false positives of 0.18 and 0.26 per image, and diagnosis accuracies of 97.86% and 93.86%, and in three-class categorization, F1 scores for malignancy reach 0.93 and 0.84, respectively. Overall, the results demonstrate promising progress compared to competing schemes in metaheuristics-based feature selection within the state-of-the-art. •Feature selection is crucial for accurate breast cancer diagnosis in CAD solutions.•Proposing novel binary tunicate swarm algorithm for feature selection.•The framework introduces new fitness functions for mass characterization.•Propose strategy outperform some state-of-the-art in mini-MIAS and DDSM datasets.
ISSN:0263-2241
DOI:10.1016/j.measurement.2024.114928