Features Selection Study for Breast Cancer Diagnosis Using Thermographic Images, Genetic Algorithms, and Particle Swarm Optimization

• Published in: International journal of artificial intelligence and machine learning Vol. 11; no. 2; pp. 1 - 18
• Main Authors: Rodrigues da Silva, Amanda Lays, Araújo de Santana, Maíra, Lins de Lima, Clarisse, Silva de Andrade, José Filipe, Silva de Souza, Thifany Ketuli, Jacinto de Almeida, Maria Beatriz, Azevedo da Silva, Washington Wagner, Fernandes de Lima, Rita de Cássia, Pinheiro dos Santos, Wellington
• Format: Journal Article
• Language: English
• Published: Hershey IGI Global 01.07.2021
• Subjects: Artificial intelligence; Breast cancer; Cancer therapies; Classification; Diagnosis; Digital imaging; Disease; Genetic algorithms; Image analysis; Infrared imaging; Machine learning; Mammography; Medical diagnosis; Medical imaging; Medical screening; Open access publishing; Optimization; Particle swarm optimization; Performance evaluation; Professionals; Radiation; Thermography; Womens health
• ISSN: 2642-1577; 2642-1585; 2642-1585
• DOI: 10.4018/IJAIML.20210701.oa1

Early detection of breast cancer is critical to improve treatment efficiency and chance of cure. Mammography is the main method for breast cancer screening; however, it has some limitations. Infrared thermography is a technique that is being studied for its benefits. The existing tumor classification systems are detailed, complex, and have low usability. Therefore, combining specialized professionals with methods of digital image analysis using thermography can help improve the diagnosis. Considering this, some computational areas are working on studies and creating methods to assess these data. The features selection plays a key role in this process, as it is a way to help solving data multidimensionality problems. This study aims to reduce the amount of features from thermographic images with mammary lesions. The authors used genetic algorithm and particle swarm optimization for features selection and compared the performance of each method to the performance using the entire set of features.