No abstract

With recent advancements in machine learning, the research community has made tremendous advances towards the classification of neurological disorders from time-series functional MRI signals. However, existing classification techniques rely on hand-crafted features and classical machine learning models. In this paper, we propose an end-to-end model that utilizes the representation learning capability of deep learning to classify a neurological disorder from fMRI data. The proposed DeepFMRI model is comprised of three networks, namely (1) a feature extractor, (2) a similarity network, and (3) a classification network. The model takes fMRI raw time-series signals as input and outputs the predicted labels; and is trained end-to-end using back-propagation. Experimental results on the publicly available ADHD-200 dataset demonstrate that this innovative model outperforms previous state-of-the-art.

The cervical spine is a highly flexible anatomy and therefore vulnerable to injuries. Unfortunately, a large number of injuries in lateral cervical X-ray images remain undiagnosed due to human errors. Computer-aided injury detection has the potential to reduce the risk of misdiagnosis. Towards building an automatic injury detection system, in this paper, we propose a deep learning-based fully automatic framework for segmentation of cervical vertebrae in X-ray images. The framework first localizes the spinal region in the image using a deep fully convolutional neural network. Then vertebra centers are localized using a novel deep probabilistic spatial regression network. Finally, a novel shape-aware deep segmentation network is used to segment the vertebrae in the image. The framework can take an X-ray image and produce a vertebrae segmentation result without any manual intervention. Each block of the fully automatic framework has been trained on a set of 124 X-ray images and tested on another 172 images, all collected from real-life hospital emergency rooms. A Dice similarity coefficient of 0.84 and a shape error of 1.69 mm have been achieved.

No abstract

No abstract

Human activity recognition is an active field of research in computer vision with numerous applications. Recently, deep convolutional networks and recurrent neural networks (RNN) have received increasing attention in multimedia studies, and have yielded state-of-the-art results. In this research work, we propose a new framework which intelligently combines 3D-CNN and LSTM networks. First, we integrate discriminative information from a video into a map called a ‘motion map’ by using a deep 3-dimensional convolutional network (C3D). A motion map and the next video frame can be integrated into a new motion map, and this technique can be trained by increasing the training video length iteratively; then, the final acquired network can be used for generating the motion map of the whole video. Next, a linear weighted fusion scheme is used to fuse the network feature maps into spatio-temporal features. Finally, we use a Long-Short-Term-Memory (LSTM) encoder-decoder for final predictions. This method is simple to implement and retains discriminative and dynamic information. The improved results on benchmark public datasets prove the effectiveness and practicability of the proposed method.

Abstract. X-ray is a common modality for diagnosing cervical vertebrae injuries. Many injuries are missed by emergency physicians which later causes life threatening complications. Computer aided analysis of X-ray images has the potential to detect missed injuries. Segmentation of the vertebrae is a crucial step towards automatic injury detection system. Active shape model (ASM) is one of the most successful and popular method for vertebrae segmentation. In this work, we propose a new ASM search method based on random classification forest and a kernel density estimation-based prediction technique. The proposed method have been tested on a dataset of 90 emergency room X-ray images containing 450 vertebrae and outperformed the classical Mahalanobis distancebased ASM search and also the regression forest-based method.

Injuries in cervical spine X-ray images are often missed by emergency physicians. Many of these missing injuries cause further complications. Automated analysis of the images has the potential to reduce the chance of missing injuries. Towards this goal, this paper proposes an automatic localization of the spinal column in cervical spine X-ray images. The framework employs a random classification forest algorithm with a kernel density estimation-based voting accumulation method to localize the spinal column and to detect the orientation. The algorithm has been evaluated with 90 emergency room X-ray images and has achieved an average detection accuracy of 91% and an orientation error of 3.6 • . The framework can be used to narrow the search area for other advanced injury detection systems.