Search

Objective To examine the heart sound signals of four common congenital heart diseases of ventricular septal defect, atrial septal defect, patent ductus arteriosus and patent foramen with associated pulmonary hypertension and propose a deep learning-based intelligent auscultation algorithm for automatic classification of heart sound signals. Methods The algorithm in this study was based upon digital signal processing technology of converting one-dimensional temporal signal classification into a two-dimensional image classification and further applying deep neural networks for automatic classification of heart sound signals. A total of 941 heart sound data samples collected from Children's Hospital of Zhejiang University School of Medicine were employed for training, validation and testing with a ratio of 8 1 1 for training, validation and testing sets respectively. Additionally, 107 heart sound data samples gathered from a clinical screening environment were collected to validate the effectiveness of intelligent auscultation algorithm in real-world clinical applications. Results In this study, discrete wavelet transformation was utilized to denoise the heart sound signals and there was a significant improvement in model performance. Compared to model without denoising, the denoised model achieved no- table enhancements in accuracy, sensitivity, specificity and F1 score on testing set with improvements of 15. 8%,32.6%,11.1% and 27.3% respectively. Furthermore, the authors compared the performance of several common classification neural network models, including Swin_transform, Vit, Mobilenet, Resenet and Vgg, with their respective F1 scores of 0.905,0.842,0.687,0.814 and 0.864. Finally, using Swin_transform model, tests on the external dataset of 107 cases yielded an accuracy of 0.833, a sensitivity of 0.872 and a specificity of 0. 801. Conclusion This study highlights the significant impact of noise and neural network structure on the performance of automatic classification models for CHD heart sound signals. Through the application of discrete wavelet transform for denoising heart sound signals, a substantial improvement in model performance is observed. Among various common classification neural network models, Swin_transform model exhibits the best classification performance. Additionally, the validation of intelligent auscultation algorithm on an external dataset of 107 cases demonstrates its effectiveness in real-world clinical applications, yielding favorable accuracy, sensitivity and specificity results. In summary, this study demonstrates the promising potential of deep learning-based intelligent auscultation algorithms for automatic classification of CHD heart sound signals. [ABSTRACT FROM AUTHOR]

Copyright of China Pulp & Paper is the property of China Pulp & Paper Magazines Publisher and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

To solve the green fruit recognition problem of persimmons and apples, a green fruit segmentation algorithm based on optimized SOLO (Segmenting Objects by Locations) was proposed in this study to achieve accurate segmentation of green fruits in complex environments. The proposed algorithm was a single-stage instance segmentation method, which avoided the disadvantage that detection before segmentation in two-stage methods relied on detection performance. By introducing the concept of instance category, each pixel in the instance was assigned a category according to the location and size of the instance, therefore, the instance segmentation was transformed into a classification problem. This study takes green persimmons and green apples as the research objects. The image collection locations are Shandong Normal University (Changqing Lake Campus) Houshan and the Longwangshan Apple Production Base in Fushan District, Yantai City, Shandong Province. The acquisition device is a Canon EOS 80D SLR camera with an image resolution of 6 000×4 000 pixels. Collect under natural light during the day (7:00-17:00) and under LED light at night (19:00-22:00). A total of 568 images of green persimmons and 498 images of green apples were collected in the experiment, including nighttime, overlap, backlighting, forward light, shading, and after rain. The collected images were annotated by LabelMe software and then were made into a dataset in COCO format. Specifically, first, split-attention networks (ResNeSt) were used to extract features of the target fruit as the backbone network of optimized SOLO, which enhanced the transfer, reuse, and fusion of features in the front and back layers. Then ResNeSt and Feature Pyramid Network (FPN) were combined to solve the multi-scale problem of green fruits. Because FPN defined allocation strategies for different scale features and assigned them to the pyramid levels optimally. Finally, the image features extracted by the ResNeSt+FPN structure were utilized for the subsequent prediction. The optimized SOLO segmentation algorithm was divided into two branches: category prediction and mask generation. While the semantic category was predicted by the category prediction branch, the object instance was segmented by the mask generation branch, in this way, the target fruit segmentation was completed. The experimental results showed that the average recall and precision of the optimized SOLO segmentation algorithm reached 94.84% and 96.16%, respectively, with an average segmentation time of 0.14 s per green target fruit image on Graphics Processing Unit (GPU). Besides, compared with four algorithms, which were the optimized Mask R-CNN fruit recognition algorithm, SOLO, Mask Region Convolutional Neural Network (Mask R-CNN), and Fully Convolutional Instance-aware Semantic Segmentation (FCIS), the recall of the optimized SOLO segmentation algorithm in this study was improved by 1.63, 1.74, 2.23, and 6.52 percentage points, the precision was improved by 1.10, 1.47, 2.61, and 6.75 percentage points, respectively, and the segmentation times were reduced by 0.06, 0.04, 0.11, and 0.13 s, respectively. The relevant results show that the green fruit optimization SOLO segmentation algorithm proposed by the study can meet the real-time performance of green fruit segmentation and improve the accuracy of green fruit segmentation. This research algorithm can provide theoretical reference for segmentation of other target fruits and vegetables to extend the application of orchard yield measurement and robot harvesting. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Data Acquisition & Processing / Shu Ju Cai Ji Yu Chu Li is the property of Editorial Department of Journal of Nanjing University of Aeronautics & Astronautics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Chinese Journal of Liquid Crystal & Displays is the property of Chinese Journal of Liquid Crystal & Displays and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The primal-dual interior point method requires strict initial feasible points in solving linear programming problems. The degree function is given based on the criterion of the initial feasible point in the primal-dual interior point method，and the fitness function in the evolutionary algorithm is designated by the degree function. The initial feasible point selection of the primal-dual interior point method for linear programming based on the evolutionary computation is proposed. It is taken by the harmony search. Numerical experiments show that the algorithm can find an initial feasible point of most given problems，thus validating the effectiveness of the algorithm. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In order to improve the recognition precision and speed for apple, and further improve the harvesting efficiency of apple harvesting robot, an apple recognition method based on combining K-means clustering segmentation with genetic radial basis function (RBF) neural network is proposed. Firstly, the captured apple image is transformed into L*a*b* color space, and then under this color space, the K-means clustering algorithm is used to segment the apple image. The color feature components and shape components of segmented image are extracted respectively. The color features include R, G, B, H, S and I, a total of 6 feature components; and the shape features include circular variance, density, ratio of perimeter square to area, and 7 Hu invariant moments, a total of 10 shape components. These extracted 16 features are used as the inputs of neural network to train RBF neural network, and get the apple recognition model. Due to some inherent defects the RBF neural network has, such as low learning rate, easily causing over fitting phenomenon, genetic algorithm (GA) is introduced to optimize the connection weights and the number of hidden layer neurons. In this study, a new optimization way is adopted, that is, the hybrid encoding of the number of hidden layer neurons and connection weights is carried out simultaneously. This moment, the learning of weights is not completed, and the least mean square (LMS) is used to further learn the connection weights. Finally, an optimized neural network model (GA-RBF-LMS) is established, which is to improve the operating efficiency and recognition precision. In the experiments, there are 150 images captured, and they have 229 apples; among them 50 images are selected as training samples, and the rest as testing samples. Every image for training sample has only one apple, so the testing samples have 179 apples. In order to get the precise model, fruits of apple are together with branches and leaves for training during the training process, which avoids the influence of branches or leaves shade on the recognition to some extent. So the training samples have 50 apples, 50 branches and 50 leaves, which are a total of 150 training samples, and the outputs of neural network include 3 classes. In order to compare with the traditional back propagation (BP) and RBF neural network, and GA-RBF algorithm, a series of experiments are carried out. After repeated trainings of 50 times, the results show that the successful training rate of the GA-RBF-LMS is the highest, which can reach 100% and get the minimum training error; but its running time is the longest, because the 2 optimizations of genetic algorithm and LMS are at the expense of the time. The recognition rates of the fruits with different growth postures, such as fruit without obscuration, overlapping fruit and covered fruit, are calculated respectively. After repeated experiments of 50 times, the results show that these 4 recognition models can achieve very good effect for recognizing the fruit without obscuration. For covered fruit and overlapping fruit, the recognition rate of GA-RBF-LMS is the highest, which can reach 95.38% and 96.17%, respectively. Looking from the overall, the recognition rate reaches 96.95%, recognizing 179 apples consumes 1.75 s, and the sum of square of error is the smallest. From the training time, the GA-RBF-LMS algorithm is the longest, whose average training time is 4.412 s for 150 training samples, but the training success rate can reach 100%, which saves the time wasted in human trying to construct the network structure. All of these illustrate that the GA-RBF-LMS neural network model has the higher operating efficiency and recognition precision, and it can be applied in target recognition for apple harvesting robot. [ABSTRACT FROM AUTHOR]