The automatic and accurate segmentation of the prostate cancer from the multi-modal magnetic resonance images is of prime importance for the disease assessment and follow-up treatment plan. However, how to use the multi-modal image features more efficiently is still a challenging problem in the field of medical image segmentation. In this paper, we develop a cross-modal self-attention distillation network by fully exploiting the encoded information of the intermediate layers from different modalities, and the generated attention maps of different modalities enable the model to transfer significant spatial information that contains more details. Moreover, a novel spatial correlated feature fusion module is further employed for learning more complementary correlation and non-linear information of different modality images. We evaluate our model in five-fold cross-validation on 358 MRI images with biopsy confirmed. Without bells and whistles, our proposed network achieves state-of-the-art performance on extensive experiments.

The demand for accurate indoor localization has become more and more urgent in location-based services (LBS). Since the traditional indoor localization is based on a single fingerprint, which is susceptible to environmental changes, it is hard to obtain high accuracy. In order to fill this gap, we propose a new multiple fingerprints method used in the indoor environment without extra measurement, and it can reduce the uncertainty caused by a single fingerprint. The multiple fingerprints are based on the signal Subspace and the Received Signal Strength Index (RSSI) and others, where the signal Subspace represents the characteristic representation of the received array signal and the RSSI represents the signal strength. First, we propose an algorithm named Subspace gradient boost decision tree (Subspace-GBDT) to obtain a strong classifier, which can make successful prediction. Second, we propose another algorithm named double threshold assisted mode (DTAM) to combine multiple fingerprints. It can achieve accurate prediction by setting two different threshold parameters, the threshold of samples number and classifiers number, and reduce errors of matching. The experiment results show that the proposed method can obtain higher cumulative distribution function(CDF) value and localization accuracy than the traditional methods.

With the rise of location-based services and the rapidly growing requirements related to their applications, indoor localization based on channel state information–multiple-input multiple-output (CSI-MIMO) has become an important research topic. However, indoor localization based on CSI-MIMO has some disadvantages, including noise and high data dimensions. To overcome the above drawbacks, we proposed a novel method of indoor localization based on CSI-MIMO, named SICD. For SICD, a novel localization fingerprint was first designed which can reflect the time–frequency and space–frequency characteristics of CSI-MIMO under a single access point (AP). To reduce the redundancy in the data of CSI-MIMO amplitude, we developed a data dimensionality reduction algorithm. Moreover, by leveraging a log-normal distribution, we calculated the conditional probability of the naive Bayes classifier, which was used to predict the moving object’s location. Compared with other state-of-the-art methods, the results of the experiment confirm that the SICD effectively improves localization accuracy.

With the increasing demand of location-based services, neural network (NN)-based intelligent indoor localization has attracted great interest due to its high localization accuracy. However, deep NNs are usually affected by degradation and gradient vanishing. To fill this gap, we propose a novel indoor localization system, including denoising NN and residual network (ResNet), to predict the location of moving object by the channel state information (CSI). In the ResNet, to prevent overfitting, we replace all the residual blocks by the stochastic residual blocks. Specially, we explore the long-range stochastic shortcut connection (LRSSC) to solve the degradation problem and gradient vanishing. To obtain a large receptive field without losing information, we leverage the dilated convolution at the rear of the ResNet. Experimental results are presented to confirm that our system outperforms state-of-the-art methods in a representative indoor environment.