Deep convolutional Triplet network for quantitative medical image analysis with comparative case study of gamma image classification

Alternative feature generation for quantitative image analysis is proposed. The proposed method reorganize Deep Convolutional Neural Networks to learn representation in Triplet Network. The features are compared with texture features using series of classifiers in a Gamma image classification task that contains visual information but has no known suitable features. Experiment show that features from the Triplet Network method outperform in the classification task suggesting a useful way of extracting feature for task without known suitable feature but advantageous for further investigation.