Cardiac magnetic resonance image (MRI) segmentation has the features such as there is a lot of noise, the target areas are indistinguishable from the background, and the shape of the right ventricle is irregular. Although convolution operations are good at extracting local features, the U-shaped convolutional neural networks (CNN) structure hardly models long-distance dependency between pixels and can not achieve ideal segmentation results on cardiac MRI. To solve these problems, UConvTrans is proposed with a dual-flow U-shaped network by global and local information integration. First, the network applies the CNN branch to extract local features and capture global representations by Transformer branch, which retains local detailed features and suppresses the interference of noise and background features in cardiac MRI. Next, the bidirectional fusion module is proposed to fuse the features extracted by CNN and the Transformer with each other, enhancing the feature expression capability and improving the segmentation accuracy of the right ventricle. Besides, the parameters of network can be set flexibly because the transformer structure in the proposed method does not require pre-trained weights. The proposed method also strikes a better balance between precision and efficiency, which is evaluated on the MICCAI 2017 ACDC dataset. The results show that the network outperforms U-Net by 1.13% average dice coefficient while the parameter amount and the floating point operations are only 10% and 8% of the U-Net. Finally, the proposed method achieves a dice coefficient of 92.42% for the right ventricle, 91.64% for the myocardium, and 95.06% for the left ventricle respectively and wins the first place in the myocardium and left ventricle on test set.
To provide an accurate and objective feedback on rehabilitation training movements and to improve the motivation of rehabilitation patients in rehabilitation training, a motion evaluation method capable of processing continuous human rehabilitation training movement data is proposed. First, a motion segmentation method based on the Gaussian mixture model (GMM) is developed to extract single motion repetition from continuous repetitive motion sequences of the same motion. Then, based on relevant a priori knowledge, a multi-feature fusion motion evaluation method combining significant motion feature dynamic time warping (DTW) distance evluation and Gaussian mixture model likelihood evaluation is proposed to perform motion evaluation in both the overall motion feature and local joint information of rehabilitation exercises. The results show that the motion segmentation method can segment the motion data of continuous repetitive motions well, and the correct rate of segmented motions on the dataset reaches more than 95%. The multi-feature fusion motion evaluation method effectively improves the differentiation of motion evaluation between healthy samples and rehabilitation patient samples, so that the motion scores of healthy samples are mainly distributed in the range of 0.93—0.94 on a scale of 0—1, while the motion scores of patient samples are mainly distributed in the range of 0.81—0.89.
To simulate the biomechanical changes of knee joint before and after wearing the orthosis in patients with knee osteoarthritis (KOA) by constructing a finite element model of the knee joint and a customized additively manufactured knee orthosis, so as to verify the decompression effect of the orthosis, and aimed at orthopaedics, a quantitative research on the therapeutic effect of knee orthosis was conducted to evaluate the clinical efficacy of knee orthosis.This experiment was approved by the Ethics Committee of the Ninth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine. A woman with knee osteoarthritis was recruited, and the knee joint was subjected to optical body surface and CT scans. A custom-made additively manufactured knee orthosis was designed according to the principle of unilateral decompression. After boundary setting and other steps, software such as ANSYS was used to build a finite element model of the knee joint and a customized additively manufactured knee joint orthosis. By applying a compressive load of 1100 N to the knee joint along the direction of the lower limb load-bearing axis, a simulation and stress analysis was performed to study the deloading effect of a custom additively manufactured knee orthosis on the knee compartment. A finite element analysis of KOA characteristics was conducted for verification, considering the influence of cartilage, ligament, and lower limb skin on the bearing capacity of the knee joint. Compared with the case without wearing any orthosis, after wearing the customized knee orthosis, the varus angle of the knee joint was reduced, the medial pressure of the knee joint was shifted to the lateral side, and the pressure of the medial compartment of the knee joint was significantly reduced. The custom additively manufactured knee orthosis can reduce the pressure generated by the medial compartment of the knee joint during walking in patients with early and mid-stage medial compartment knee osteoarthritis, and the load reduction effect is significant.
To improve the reconstruction speed of parallel magnetic resonance imaging, an efficient reconstruction method named fSIDWT-SPIRiT is proposed based on shift-invariant discrete wavelets transform (SIDWT) and the iterative self-consistent parallel imaging reconstruction (SPIRiT) model. This method addresses the complex optimization problem containing data consistency term, calibration consistency term, and L1-norm regularization term. First, data consistency term and calibration consistency term are combined and processed, and then solved by a projected fast iterative shrinkage-thresholding algorithm to achieve fast parallel MRI reconstruction. Finally, simulation experiments are conducted using different human organ datasets. The results show that the proposed method is able to guarantee the image reconstruction quality with a faster convergence speed compared with other methods.
Laser speckle contrast imaging (LSCI) is a large measurement ranging, real-time, high spatial resolution optical imaging method. Exisiting reserches show that the low gray level (8-bit) and low resolution (752×480 pixels) camera can effectively monitor the flow of scattering media such as blood, but the defects such as large noise of scattered particles and small effective imaging area are difficult to be compensated by software, which seriously affects the imaging quality. Monitoring blood flow with a camera with a high gray level (16-bit) and a high resolution (2 048×2 048 pixels) will slow down the imaging speed of a single frame, and parallel computing can reduce the image processing time by 1/3. With the aid of an animal blood flow experiment, the results of high-gray-level and high-resolution spatial laser speckle contrast imaging (sLSCI), spatial approximate laser speckle contrast imaging (sLSCIa), and temporal laser speckle contrast imaging (tLSCI) were compared and analyzed using imaging speed and imaging quality as evaluation criteria. The parallel computation of high-gray-level and high-resolution sLSCI takes into consideration both imaging quality and imaging speed, which can meet the requirements of clinical real-time monitoring of blood flow.
In order to meet the needs of the intestinal robot to obtain power stably through wireless methods in different positions and postures, and to reduce the volume of the intestinal robot as much as possible, the transmitting coil is required to have the ability to generate spatial multi-dimensional magnetic fields. A hybrid transmitting coil structure is proposed, which combines Helmholtz coil pair and saddle-shaped coil pair. The structure is compact, and its symmetry can be used to generate a three-dimensional magnetic field without dead zone through rotation. The characteristics of the two types of coils and the combined characteristics are analyzed separately. The simulation and experimental results show that at the excitation of 2 A current, the minimum transmission efficiency of 3.44% and the received power of 1 204 mW can be obtained at the center of the coil. The minimum positional uniformity is 88.1%, and the three-dimensional power coverage can be achieved.
