Anzeige
Erschienen in:
2018 | OriginalPaper | Buchkapitel
Classification of Breast Cancer Histology Image using Ensemble of Pre-trained Neural Networks
verfasst von : Sai Saketh Chennamsetty, Mohammed Safwan, Varghese Alex
Erschienen in: Image Analysis and Recognition
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
Breast cancer is one of the most commonly occurring types of cancer and the treatment administered to a subject is dependent on the grade or type of the lesion. In this manuscript, we make use of an ensemble of convolutional neural networks (CNN) to classify histology images as Normal, In-situ, Benign or Invasive. The performance of CNN is dependent on the network architecture, number of training instances and also on the data normalization scheme. However, there exists neither a single architecture nor a pre-processing regime that promises best performance. For the reason stated above, we use 3 CNNs trained on different pre-processing regimes to form an ensemble. On the held out test data (n = 40), the proposed scheme achieved an accuracy of 97.5%. On the challenge data (n = 100) provided by the organizers, the proposed technique achieved an accuracy of 87% and was jointly adjudged as the top performing algorithm for the task of classification of breast cancer from histology images.