Region-specific bag-of-visual-words representations for retrieving brain tumors in contrast-enhanced MRI

The current paper proposes a content-based retrieval method for T1-weighted contrast-enhanced MRI (CE-MRI) of brain tumors using region-specific bag-of-visual-words (BoW) representations. Patches are densely sampled within and around the brain tumor regions, with normalization and whitening as the preprocessing methods. The proposed method is evaluated on a brain tumor CE-MRI dataset. The experimental results demonstrate that the mean average precision of the proposed method for retrieval can achieve 90.1%.