This letter investigates the inversion of rough surface parameters (the root mean square height and the correlation length) from microwave images by using deep convolutional neural networks (CNNs). Training data for the deep CNN are simulated numerically using computational electromagnetic method. As CNN is powerful in extracting image features, scattering field from rough surfaces is first converted to microwave images via interpolated fast Fourier transform and then fed into the CNN. In order to reduce overfitting, the regularization technique and dropout layer are used. The proposed CNN consists of five pairs of convolutional and maxpooling layers and two additional convolution layers for feature extraction and two fully connected layers for parameter regression. The experimental results demonstrated the feasibility using deep neural networks for the parameter inversion of rough surface from electromagnetic scattering fields. It suggests potential application of CNN for rough surface parameter inversion from microwave sensing data.
There is behavioural evidence that caloric vestibular stimulation (CVS) can alleviate central pain. Several such patients have also noted that it reduces tactile allodynia, an especially ill-understood phenomenon in these patients. The first aim is to use magnetoencephalography (MEG) to study neural activity associated with tactile allodynia in central post-stroke pain (CPSP). The second is to assess how this would be affected, if at all, by CVS. The third is to assess the ability of the VESTAL solution for MEG to detect anterior cingulate activation. A 58-year-old woman with CPSP, and marked unilateral tactile allodynia, participated in a MEG study with imaging pre- and post-CVS. Tactile simulation within the patient's allodynic area resulted in contralateral activation of the primary motor and anterior cingulate cortices, which had normalized 24 h post-CVS. We suggest that the unexpected primary motor cortex activation in response to light touch in the allodynic area arises from inappropriate activation of a normal mechanism, which may occur when a threat to homeostasis is present, to lower motor thresholds and allow for more rapid performance of corrective actions. We propose this may be mediated by the interoceptive cortex in the dorsal posterior insula.
Whole abdominal organ segmentation is important in diagnosing abdomen lesions, radiotherapy, and follow-up. However, oncologists' delineating all abdominal organs from 3D volumes is time-consuming and very expensive. Deep learning-based medical image segmentation has shown the potential to reduce manual delineation efforts, but it still requires a large-scale fine annotated dataset for training, and there is a lack of large-scale datasets covering the whole abdomen region with accurate and detailed annotations for the whole abdominal organ segmentation. In this work, we establish a new large-scale \textit{W}hole abdominal \textit{OR}gan \textit{D}ataset (\textit{WORD}) for algorithm research and clinical application development. This dataset contains 150 abdominal CT volumes (30495 slices). Each volume has 16 organs with fine pixel-level annotations and scribble-based sparse annotations, which may be the largest dataset with whole abdominal organ annotation. Several state-of-the-art segmentation methods are evaluated on this dataset. And we also invited three experienced oncologists to revise the model predictions to measure the gap between the deep learning method and oncologists. Afterwards, we investigate the inference-efficient learning on the WORD, as the high-resolution image requires large GPU memory and a long inference time in the test stage. We further evaluate the scribble-based annotation-efficient learning on this dataset, as the pixel-wise manual annotation is time-consuming and expensive. The work provided a new benchmark for the abdominal multi-organ segmentation task, and these experiments can serve as the baseline for future research and clinical application development.