TITLE: Bag of Visual Words for Medical Images SPEAKER: Yu Qian (Middlesex University) ABSTRACT: Bag of visual Words (BoW) paradigm is extremely popular and successfully applied for image categorization. This method is to transform images into a set of ‘visual vocabulary’ and to represent the images using the statistics of appearance of each word as feature vectors, such as histogram. In this presentation, unlike the traditional BoW paradigm, sparse coding is employed instead of Vector Quantization (VQ) to train a video dictionary based on a set of local features. Furthermore, instead of using histograms, multiple scales of max pooling features are applied as the representations of image. Three applications for medical image representations based on SIFT sparse codes are finally presented, which are medical image classification, 3D brain image retrieval and echocardiogram video classification.