CNN in for cancer detection using biomedical images

Dra. Aura Conci (Universidade Federal Fluminense)
Resumen: Cancer has high prevalence and their early identification is crucial for a more effective treatment. This work presents a methodology for determining thyroid and breast nodules by image analysis and CNN to classify regions of interest and identifying which ones refer to malignant nodules. CNNs are applied for the classification of anomaly areas and therefore for the identification of malignant nodules. Two CNNs models are tested based on GoogLeNet and ResNet. These architectures are selected based in previous works. Three learning rates values and five datasets are tested, in order to maximize the performance of the algorithms. The good results of CNN in the classification (with 96% accuracy), show that the viability of the proposed methodology.