1. Skin Lesion Classification by Ensembles of Deep Convolutional Networks and Regularly Spaced Shifting
- Author
-
Thurnhofer-Hemsi, Karl, López-Rubio, Ezequiel, Domínguez, Enrique, Elizondo, David A., [Thurnhofer-Hemsi,K, López-Rubio,E, Domínguez,E] Department of Computer Languages and Computer Science, Universidad de Málaga, Málaga, Spain. [Thurnhofer-Hemsi,K, Domínguez,E] Biomedic Research Institute of Málaga (IBIMA), Málaga, Spain. [Elizondo,DA] School of Computer Science and Informatics, De Montfort University, Leicester, U.K., and This work was supported in part by the Ministry of Science, Innovation, and Universities of Spain, through European Regional Development Fund (ERDF), project name ‘‘Automated Detection with Low-Cost Hardware of Unusual Activities in Video Sequences,’’ under Grant RTI2018-094645-B-I00, in part by the Autonomous Government of Andalusia, Spain, through ERDF, project name ‘‘Detection of Anomalous Behavior Agents by Deep Learning in Low-Cost Video Surveillance Intelligent Systems,’’ under Project UMA18-FEDERJA-084, in part by the University of Malaga, Spain, project name ‘‘Anomaly Detection on Roads by Moving Cameras,’’ under Grant B1-2019_01, in part by the University of Malaga, project name ‘‘Self-Organizing Neural Systems for Non-Stationary Environments,’’ under Grant B1-2019_02, in part by the Universidad de Málaga, and in part by the Instituto de Investigación Biomédica de Málaga (IBIMA).
- Subjects
Lesiones por desenguantamiento ,Information Science::Information Science::Computing Methodologies::Image Processing, Computer-Assisted [Medical Subject Headings] ,Aprendizaje profundo ,Red nerviosa ,Neoplasias cutáneas ,Diseases::Neoplasms::Neoplasms by Site::Skin Neoplasms [Medical Subject Headings] ,Diseases::Skin and Connective Tissue Diseases::Skin Diseases [Medical Subject Headings] ,Information Science::Information Science::Classification [Medical Subject Headings] ,Diseases::Immune System Diseases::Hypersensitivity [Medical Subject Headings] ,Deep learning ,Clasificación ,Classification ,Analytical, Diagnostic and Therapeutic Techniques and Equipment::Investigative Techniques::Epidemiologic Methods::Statistics as Topic::Probability [Medical Subject Headings] ,Organisms::Eukaryota::Animals::Chordata::Vertebrates::Mammals::Primates::Haplorhini::Catarrhini::Hominidae::Humans [Medical Subject Headings] ,Skin lesion ,Image processing ,Anatomy::Integumentary System::Skin::Epidermis [Medical Subject Headings] ,Phenomena and Processes::Mathematical Concepts::Neural Networks (Computer) [Medical Subject Headings] ,Skin cancer ,Convolutional neural networks ,Procesamiento de imagen asistido por computador ,Anatomy::Cells::Epithelial Cells::Melanocytes [Medical Subject Headings] ,Melanoma ,Diseases::Neoplasms::Neoplasms by Histologic Type::Nevi and Melanomas::Melanoma [Medical Subject Headings] - Abstract
Skin lesions are caused due to multiple factors, like allergies, infections, exposition to the sun, etc. These skin diseases have become a challenge in medical diagnosis due to visual similarities, where image classification is an essential task to achieve an adequate diagnostic of different lesions. Melanoma is one of the best-known types of skin lesions due to the vast majority of skin cancer deaths. In this work, we propose an ensemble of improved convolutional neural networks combined with a test-time regularly spaced shifting technique for skin lesion classification. The shifting technique builds several versions of the test input image, which are shifted by displacement vectors that lie on a regular lattice in the plane of possible shifts. These shifted versions of the test image are subsequently passed on to each of the classifiers of an ensemble. Finally, all the outputs from the classifiers are combined to yield the final result. Experiment results show a significant improvement on the well-known HAM10000 dataset in terms of accuracy and F-score. In particular, it is demonstrated that our combination of ensembles with test-time regularly spaced shifting yields better performance than any of the two methods when applied alone. Yes
- Published
- 2021