Application of DenseNets for Classification of Breast Cancer Mammograms
| Autor: | Anita Rybiałek, Łukasz Jeleń |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
medicine.diagnostic_test
business.industry Computer science Deep learning Pattern recognition 02 engineering and technology medicine.disease Class (biology) Convolutional neural network 030218 nuclear medicine & medical imaging 03 medical and health sciences 0302 clinical medicine Breast cancer Computer-aided diagnosis 0202 electrical engineering electronic engineering information engineering medicine Mammography 020201 artificial intelligence & image processing Artificial intelligence business Breast cancer classification Focus (optics) |
| Zdroj: | Computer Information Systems and Industrial Management ISBN: 9783030476786 CISIM |
| DOI: | 10.1007/978-3-030-47679-3_23 |
| Popis: | In this study, we focus on the problem of a breast cancer diagnosis using mammography images by classifying them as belonging either to a negative or to a malignant mass class. We explore the potential of densely connected convolutional neural network (DenseNet) architectures by comparing its three different variants that were trained to classify the abnormalities in breast tissue. The models have been tested in a series of systematic experiments. With a limited dataset (2247 images per class), it was necessary to perform tests to verify whether the amount of data used in this work is sufficient to allow for the conclusion that the experimental results are not dependent on the subset of the data. The training was conducted using stratified 10-fold cross-validation to obtain statistically reliable metrics estimates. DenseNet-201 was found to be the best model achieving: 0.96 value for area under the curve (AUC), 0.92 for precision, 0.90 for recall, and 91% for accuracy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|