Accurate and automatic segmentation of medical images is a crucial step for clinical diagnosis and analysis. The convolutional neural network (CNN) approaches based on the U-shape structure have achieved remarkable performances in many different medical image segmentation tasks. However, the context information extraction capability of single stage is insufficient in this structure, due to the problems such as imbalanced class and blurred boundary. In this paper, we propose a novel Context Pyramid Fusion Network (named CPFNet) by combining two pyramidal modules to fuse global/multi-scale context information. Based on the U-shape structure, we first design multiple global pyramid guidance (GPG) modules between the encoder and the decoder, aiming at providing different levels of global context information for the decoder by reconstructing skip-connection. We further design a scale-aware pyramid fusion (SAPF) module to dynamically fuse multi-scale context information in high-level features. These two pyramidal modules can exploit and fuse rich context information progressively. Experimental results show that our proposed method is very competitive with other state-of-the-art methods on four different challenging tasks, including skin lesion segmentation, retinal linear lesion segmentation, multi-class segmentation of thoracic organs at risk and multi-class segmentation of retinal edema lesions.

Automated retinal layer segmentation of optical coherence tomography (OCT) images has been successful for normal eyes but becomes challenging for eyes with retinal diseases if the retinal morphology experiences critical changes. We propose a method to automatically segment the retinal layers in 3-D OCT data with serous retinal pigment epithelial detachments (PED), which is a prominent feature of many chorioretinal disease processes. The proposed framework consists of the following steps: fast denoising and B-scan alignment, multi-resolution graph search based surface detection, PED region detection and surface correction above the PED region. The proposed technique was evaluated on a dataset with OCT images from 20 subjects diagnosed with PED. The experimental results showed the following. 1) The overall mean unsigned border positioning error for layer segmentation is 7.87±3.36 μm, and is comparable to the mean inter-observer variability ( 7.81±2.56 μm). 2) The true positive volume fraction (TPVF), false positive volume fraction (FPVF) and positive predicative value (PPV) for PED volume segmentation are 87.1%, 0.37%, and 81.2%, respectively. 3) The average running time is 220 s for OCT data of 512 × 64 × 480 voxels.

Misalignment angle is an important error in the integration of the Strapdown Inertial Navigation System (SINS) and the Global Positioning System (GPS). Currently, the popular methods for SINS/GPS integrated system are based on the ground velocity. There are two major defects: (1) not all GPS receivers provide the velocity outputs; (2) the contained outliers in the velocity will degrade the performance of the alignment methods. In this paper, an in-motion coarse alignment method for SINS/GPS integrated system using position loci is proposed. Firstly, the observation vector for the in-motion coarse alignment with the position loci is derived. Different from the current popular methods, the constructed observation vector is just relative with the GPS positioning information, which can be obtained from any GPS receiver directly. Thus, the applicable range of the in-motion coarse alignment method is extended. Secondly, based on the advantages of the position loci, the proposed method can suppress the interferences of the outliers, which are contained in the GPS positioning outputs. Finally, the simulation and field tests are designed to verify the performance of the proposed method. The test results show that the proposed method is more robust than the current popular methods.

With the increase in the application of micro-electromechanical system (MEMS) accelerometers, their accurate calibration is increasingly essential to reduce the application error. Aiming at the high accuracy calibration requirements of scale factor, bias, and nonorthogonalities, we propose a novel calibration method to obtain highly accurate parameters, especially in the scenario without turntable. First, a twelve-parameter model of accelerometer output data is used for analysis. The static data characteristics of the six specified positions of the accelerometer were then analyzed. Finally, based on these characteristics, a novel calibration method performed using an iterated parameter correction process is proposed. Repeatable simulations under different biases show that the calibration accuracy of the proposed method is not affected by the bias value to be calibrated, and that the calibration results have consistent accuracy. The estimation error of bias and relative error of scale factor are less than 0.5 mg and 0.05%, respectively, and the accuracy of the non-orthogonal elements of the calibration matrix S and K are 0.003 and 0.006, respectively. In addition, the calibration results of the laboratory MEMS accelerometer and the triaxial accelerometer in ADIS16488 demonstrate the effectiveness of the proposed method.

Bias stability is an important performance indicator for MEMS gyroscopes. In this paper, a gyroscope zero rate output (ZRO) model under force-to-rebalance (FTR) closed-loop detection is presented, and the effect of circuit phase delay and various noises on the ZRO is analyzed. Based on the fact that the two feedback forces in the FTR system are insensitive to the phase delay of the sense mode, while they are sensitive to the phase delay of the drive mode, a method for quickly calculating the circuit phase delay is proposed, and an all-pass filter is used to realize one-time automatic compensation for the phase delay in the drive mode. The control system is implemented with an FPGA. The experimental results show that this method can be used to accurately calculate the circuit phase delay and that phase compensation can be used to effectively reduce the effect of quadrature error on ZRO. This technique provides bias instability and angle random walk performances of 0.338° h ⁻¹ and 0.061 ° (√h) ⁻¹ , respectively. In addition, the sensitivity of ZRO to temperature at 0 °C to 70 °C has reached 0.001°/s/°C.