NovaSearch at ImageCLEFmed 2016 Subfigure Classification Task

This paper describes the NovaSearch team participation in
the ImageCLEF 2016 Medical Task in the subfigure classification subtask.
Deep learning techniques have proved to be very effective in automatic
representation learning and classification tasks with general data. More
specifically, convolutional neural networks (CNNs) have surpassed humanlevel
performance in the ImageNET classification task, making them a
promising model for the task of medical modality classification. We assess
how each model behave when dealing with medical images, by developing
three different models, with different depths and components, and
analyse the impact of these factors in the performance. One of the key
ingredients for the effectiveness of CNNs (and deep learning in general)
is the use of large amounts of data for training. This subtask scenario
is completely different, due to the small size of the dataset, implying a
significant risk of overfitting. We apply state-of-the-art techniques developed
to reduce overfitting in these networks to our models and evaluate
their effectiveness. Our best model achieves 65.31% accuracy on the test
set using only the training data provided.