Activity

The performance was measured in terms of the Dice’s coefficient similarity, the area error and mean contour distance. The proposed method achieves yielded the best results in all the performance measures.Mammograms are commonly employed in the large scale screening of breast cancer which is primarily characterized by the presence of malignant masses. However, automated image-level detection of malignancy is a challenging task given the small size of the mass regions and difficulty in discriminating between malignant, benign mass and healthy dense fibro-glandular tissue. To address these issues, we explore a two-stage Multiple Instance Learning (MIL) framework. A Convolutional Neural Network (CNN) is trained in the first stage to extract local candidate patches in the mammograms that may contain either a benign or malignant mass. The second stage employs a MIL strategy for an image level benign vs. malignant classification. A global image-level feature is computed as a weighted average of patch-level features learned using a CNN. Our method performed well on the task of localization of masses with an average Precision/Recall of 0.76/0.80 and achieved an average AUC of 0.91 on the image-level classification task using a five-fold cross-validation on the INbreast dataset. Restricting the MIL only to the candidate patches extracted in Stage 1 led to a significant improvement in classification performance in comparison to a dense extraction of patches from the entire mammogram.The use of deep learning methods has dramatically increased the state-of-the-art performance in image object localization. However, commonly used supervised learning methods require large training datasets with pixel-level or bounding box annotations. Obtaining such fine-grained annotations is extremely costly, especially in the medical imaging domain. In this work, we propose a novel weakly supervised method for breast cancer localization. The essential advantage of our approach is that the model only requires image-level labels and uses a self-training strategy to refine the predicted localization in a step-wise manner. We evaluated our approach on a large, clinically relevant mammogram dataset. The results show that our model significantly improves performance compared to other methods trained similarly.In recent years, the conceptualisation of the brain as a “connectome” as summary measures derived from graph theory analyses, has become increasingly popular. Still, such approaches are inherently limited by the need to condense and simplify temporal fMRI dynamics and architecture into a purely spatial representation. We formulate a novel architecture based on Geometric Deep Learning which is specifically tailored to the one-step integration of spatial relationship between nodes and single-node temporal dynamics. We compare different spatiotemporal modelling mechanisms and demonstrate the effectiveness of our architecture in a binary prediction task based on a large homogeneous fMRI dataset made publicly available by the Human Connectome Project (HCP). As the idea of e.g. a dynamical network connectivity is beginning to make its way into the more mainstream toolset which neuroscientists commonly employ with neuroimaging data, our model can contribute to laying the groundwork for explicitly incorporating spatiotemporal information into every association and prediction problem in neuroscience.Recent neuroimaging studies have employed graph theory as a data-driven approach to describe topological organization of the brain under different neurological disorders or task conditions and across life span. In this exploratory study, we tested whether subtle differences in interoception related to intravesical fullness can alter brain topological architecture in healthy participants. 17 right-handed women underwent a series of resting state fMRI scans that included catheterization and partial bladder filling. Using a whole brain regions of interest (ROIs), we computed several graph theory metrics to assess the efficiency of brain-wide information exchange. Results showed that brain network’s topological properties significantly changed in many brain regions when we binary compared different interoceptive resting state conditions. Notably, we observed changes in global efficiency in the salience network, the central executive network, anterior dorsal attention network and the posterior default-mode network (DMN) as bladder became full and interoceptive signals intensified. Moreover, degree (the number of connections for each node), and betweenness centrality (how connected a particular region is to other regions) differed between the empty bladder, the catheterized empty bladder, and the catheterized and partially filled bladder. Comparing resting state data before and after an interoceptive task (repeated intravesical infusion and drainage) further showed increased average path length for the salience networks and decreased clustering coefficient of the DMN. These results suggest visceral interoception influences brain topological properties of resting state networks.Deep brain stimulation (DBS) is used to treat a range of neurologic conditions. Determining the anatomic location of the DBS lead and inferring the microelectrode recording track from co-registered pre-operative and post-operative scans is important for stereotactic surgery and neurophysiology research. read more Reslicing images with the DBS lead in-plane while maintaining mirror symmetry is not possible with current clinical navigation software. Therefore, we developed an open source software tool in Matlab for visualizing DBS lead placement and anatomic segmentation with computed tomography and magnetic resonance images. The code and graphical user interface are available at github.com/camplaboratory/DBS_reslice.Reconstructing the perceived faces from brain signals has become a promising work recently. However, the reconstruction accuracies rely on a large number of brain signals collected for training a stable reconstruction model, which is really time consuming, and greatly limits its application. In our current study, we develop a new framework that can efficiently perform high-quality face reconstruction with only a small number of brain signals as training samples. The framework consists of three mathematical models principle component analysis (PCA), linear regression (LR) and conditional generative adversarial network (cGAN). We conducted a functional Magnetic Resonance Imaging (fMRI) experiment in which two subjects’ brain signals were collected to test the efficiency of our proposed method. Results show that we can achieve state-of-the-art reconstruction performance from brain signals with a very limited number of fMRI training samples.