Search

Transfer learning technique is popularly employed for a lot of medical image classification tasks. Here based on convolutional neural network (CNN) and sparse coding process, we present a new deep transfer learning architecture for false positive reduction in lymph node detection task. We first convert the linear combination of the deep transferred features to the pre-trained filter banks. Next, a new point-wise filter based CNN branch is introduced to automatically integrate transfer features for the false and positive image classification purpose. To lower the scale of the proposed architecture, we bring sparse coding process to the fixed transferred convolution filter banks. On this basis, a two-stage training strategy with grouped sparse connection is presented to train the model efficiently. The model validity is tested on lymph node dataset for false positive reduction and our approach indicates encouraging performances compared to prior approaches. Our method reaches sensitivities of 71% /85% at 3 FP/vol. and 82% /91% at 6 FP/vol. in abdomen and mediastinum respectively, which compare competitively to previous approaches.

Many medical and real images are suffered from intensity inhomogeneity and weak edges. For higher image segmentation quality, lots of level set-based methods have been proposed. Some of them however cannot take advantage of image gradient information. And severe intensity inhomogeneity and weak edges are not disposed properly. To address these problems, a new level set method integrated with local direction gradient information is presented in this paper. Firstly, according to the two assumptions on image intensity inhomogeneity adopted by many existing methods, a new pixel classification model based on image gradient is introduced. Secondly, we employ variational level set method combined with image spatial information, which improves the anti-noise capability of the proposed method. Finally, considering the gray gradients in homogeneous regions are close to constants, an improved diffusion process is incorporated into the level set function to make the evolving curves stay around true image edges. To verify our method, different testing images including synthetic images, magnetic resonance imaging (MRI) and real-world images are introduced. The image segmentation results demonstrate that our method can deal with the relatively severe intensity inhomogeneity and obtain the comparatively ideal segmentation results efficiently.