Children's Pain Identification Based on Skin Potential Signal.

Pain management is a crucial concern in medicine, particularly in the case of children who may struggle to effectively communicate their pain. Despite the longstanding reliance on various assessment scales by medical professionals, these tools have shown limitations and subjectivity. In this paper, we present a pain assessment scheme based on skin potential signals, aiming to convert subjective pain into objective indicators for pain identification using machine learning methods. We have designed and implemented a portable non-invasive measurement device to measure skin potential signals and conducted experiments involving 623 subjects. From the experimental data, we selected 358 valid records, which were then divided into 218 silent samples and 262 pain samples. A total of 38 features were extracted from each sample, with seven features displaying superior performance in pain identification. Employing three classification algorithms, we found that the random forest algorithm achieved the highest accuracy, reaching 70.63%. While this identification rate shows promise for clinical applications, it is important to note that our results differ from state-of-the-art research, which achieved a recognition rate of 81.5%. This discrepancy arises from the fact that our pain stimuli were induced by clinical operations, making it challenging to precisely control the stimulus intensity when compared to electrical or thermal stimuli. Despite this limitation, our pain assessment scheme demonstrates significant potential in providing objective pain identification in clinical settings. Further research and refinement of the proposed approach may lead to even more accurate and reliable pain management techniques in the future.

Reconstructing 3D human pose and shape from a single image and sparse IMUs.

Model-based 3D pose estimation has been widely used in many 3D human motion analysis applications, in which vision-based and inertial-based are two distinct lines. Multi-view images in a vision-based markerless capture system provide essential data for motion analysis, but erroneous estimates still occur due to ambiguities, occlusion, or noise in images. Besides, the multi-view setting is hard for the application in the wild. Although inertial measurement units (IMUs) can obtain accurate direction without occlusion, they are usually susceptible to magnetic field interference and drifts. Hybrid motion capture has drawn the attention of researchers in recent years. Existing 3D pose estimation methods jointly optimize the parameters of the 3D pose by minimizing the discrepancy between the image and IMU data. However, these hybrid methods still suffer from the issues such as complex peripheral devices, sensitivity to initialization, and slow convergence. Methods: This article presents an approach to improve 3D human pose estimation by fusing a single image with sparse inertial measurement units (IMUs). Based on a dual-stream feature extract network, we design a model-attention network with a residual module to closely couple the dual-modal feature from a static image and sparse inertial measurement units. The final 3D pose and shape parameters are directly obtained by a regression strategy. Results: Extensive experiments are conducted on two benchmark datasets for 3D human pose estimation. Compared to state-of-the-art methods, the per vertex error (PVE) of human mesh reduces by 9.4 mm on Total Capture dataset and the mean per joint position error (MPJPE) reduces by 7.8 mm on the Human3.6M dataset. The quantitative comparison demonstrates that the proposed method could effectively fuse sparse IMU data and images and improve pose accuracy.

Automatic Cobb angle measurement method based on vertebra segmentation by deep learning.

The accuracy of the Cobb measurement is essential for the diagnosis and treatment of scoliosis. Manual measurement is however influenced by the observer variability hence affecting progression evaluation. In this paper, we propose a fully automatic Cobb measurement method to address the accuracy issue of manual measurement. We improve the U-shaped network based on the multi-scale feature fusion to segment each vertebra. To enable multi-scale feature extraction, the convolution kernel of the U-shaped network is substituted by the Inception Block. To solve the problem of gradient disappearance caused by the widening of the network structure from the Inception Block, we propose using Res Block. CBAM (Convolutional Block Attention Module) can help the network judges the importance of the feature map to modify learning weight. Also, to further enhance the accuracy of feature extraction, we add the CBAM to the U-shaped network bottleneck. Finally, based on the segmented vertebrae, the efficient automatic Cobb angle measurement method is proposed to estimate the Cobb angle. In the experiments, 75 spinal X-ray images are tested. We compare the proposed U-Shaped network with the state-of-the-art methods including DeepLabV3 + , FCN8S, SegNet, U-Net, U-Net + + , BASNet, and U2Net for vertebra segmentation. Our results show that compared to these methods, the Dice coefficient is improved by 32.03%, 33.58%, 12.42%, 5.65%, 4.55%, 4.42%, and 3.27%, respectively. The CMAE of the calculated Cobb measurement is 2.45°, which is lower than the average error of 5-7° of manual measurement. The experimental results indicate that the improved U-shaped network improves the accuracy of vertebra segmentation. The proposed efficient automatic Cobb measurement method can be used in clinics to reduce observer variability.

sEMG-Based Motion Recognition of Upper Limb Rehabilitation Using the Improved Yolo-v4 Algorithm.

The surface electromyography (sEMG) signal is widely used as a control source of the upper limb exoskeleton rehabilitation robot. However, the traditional way of controlling the exoskeleton robot by the sEMG signal requires one to specially extract and calculate for complex sEMG features. Moreover, due to the huge amount of calculation and individualized difference, the real-time control of the exoskeleton robot cannot be realized. Therefore, this paper proposes a novel method using an improved detection algorithm to recognize limb joint motion and detect joint angle based on sEMG images, aiming to obtain a high-security and fast-processing action recognition strategy. In this paper, MobileNetV2 combined the Ghost module as the feature extraction network to obtain the pretraining model. Then, the target detection network Yolo-V4 was used to estimate the six movement categories of the upper limb joints and to predict the joint movement angles. The experimental results showed that the proposed motion recognition methods were available. Every 100 pictures can be accurately identified in approximately 78 pictures, and the processing speed of every single picture on the PC side was 17.97 ms. For the train data, the mAP@0.5 could reach 82.3%, and mAP@0.5-0.95 could reach 0.42; for the verification data, the average recognition accuracy could reach 80.7%.

Pixel-level aflatoxin detecting in maize based on feature selection and hyperspectral imaging.

Aflatoxin is highly toxic and is easily found in maize, a little aflatoxin can induce liver cancer. In this paper, we used hyperspectral data in the pixel-level to build the aflatoxin classifying model, each of the pixel have 600 hyperspectral bands and labeled 'clean' or 'contaminated'. We use 3 method to extracted feature bands, one method is to select 4 hyperspectral bands from other articles: 390 nm, 440 nm, 540 nm and 710 nm, another method is to use feature extraction PCA to obtain first 5 pcs to shrink the hyperspectral volume, the third method is to use Fscnca, Fscmrmr, Relieff and Fishier algorithm to select top 10 feature bands. After feature band selection or extraction, we put the feature bands into Random Forest (RF) and K-nearest neighbor (KNN) to classify whether a pixel is polluted by aflatoxin. The highest accurate for feature selection is Relieff, it reached the accuracy of 99.38% with RF classifier and 98.77% in KNN classifier. PCA feature extraction with RF classifier also reached a high accuracy 93.83%. And the 600 bands without feature extraction reached the accuracy of 100%. Feature bands selected from other papers could reach an accuracy of 89.51%. The result shows that the feature extraction performs well on its own data set. And if the computing time is not taken into account, we could use full band to classify the aflatoxin due to its high accuracy.

Recognition of Human Activities Using Continuous Autoencoders with Wearable Sensors.

This paper provides an approach for recognizing human activities with wearable sensors. The continuous autoencoder (CAE) as a novel stochastic neural network model is proposed which improves the ability of model continuous data. CAE adds Gaussian random units into the improved sigmoid activation function to extract the features of nonlinear data. In order to shorten the training time, we propose a new fast stochastic gradient descent (FSGD) algorithm to update the gradients of CAE. The reconstruction of a swiss-roll dataset experiment demonstrates that the CAE can fit continuous data better than the basic autoencoder, and the training time can be reduced by an FSGD algorithm. In the experiment of human activities' recognition, time and frequency domain feature extract (TFFE) method is raised to extract features from the original sensors' data. Then, the principal component analysis (PCA) method is applied to feature reduction. It can be noticed that the dimension of each data segment is reduced from 5625 to 42. The feature vectors extracted from original signals are used for the input of deep belief network (DBN), which is composed of multiple CAEs. The training results show that the correct differentiation rate of 99.3% has been achieved. Some contrast experiments like different sensors combinations, sensor units at different positions, and training time with different epochs are designed to validate our approach.