A Modified HSIFT Descriptor for Medical Image Classification of Anatomy Objects

Modeling low level features to high semantics in medical imaging is an important aspect filtering anatomy objects. Bag of Visual Words (BOVW) representations have been proven effective model these mid representations. Convolutional neural nets are learning systems that can automatically extract high-quality from raw images. However, their deployment the field still a bit challenging due lack training data. In this paper, learned obtained by convolutional networks compared with our proposed hand-crafted HSIFT features. The feature symmetric fusion Harris corner detector and Scale Invariance Transform process (SIFT) BOVW representation. SIFT enhanced as well classification technique adopting bagging surrogate split method. Quantitative evaluation shows outperforms discriminating image classes.