A convolutional neural network algorithm for breast tumor detection with magnetic detection electrical impedance tomography

Chen, Ruijuan; Zhao, Songsong; Wu, Weiwei; Sun, Zhihui; Wang, Huiquan; Han, Guojun

doi:10.1063/5.0041423

Cited by 7 publications

(4 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, by comparing the training results and test results of the CNN algorithm, RNN algorithm, and DNN algorithm, it was found that there was no significant difference between the training results and test results of the CNN algorithm (P > 0.05), while there was significant difference between the two results of DNN algorithm and RNN algorithm (P < 0.05), suggesting that the stability of CNN performance was higher than that of RNN and DNN algorithms. Chen et al [23] mentioned that the CNN algorithm had good stability in application testing. Xu et al [24] also proposed that the recognition accuracy and stability of the CNN algorithm were better than those of recognition methods based on time domain and frequency domain characteristics.…”

Section: Discussionmentioning

confidence: 99%

Machine Learning-Based Aerobic Exercise Recognition and Its Effect on Health Status of College Students

Zhang

Lai

et al. 2022

Mobile Information Systems

View full text Add to dashboard Cite

It was to explore the role of motion recognition technology based on the convolutional neural network (CNN) model in guiding the aerobic exercise of college students and its impact on their physical and mental health, providing a scientific and effective method for students to have a healthy body and mood. 256 college students were randomly divided into control group (appropriate aerobic exercise was selected by self-evaluation, and exercise was performed spontaneously) and observation group (aerobic exercise under the guidance of machine learning model-based motion recognition technology), 128 cases in each group. The changes in the physical and psychological status of the two groups of college students were evaluated after one year. The recognition rate of CNN algorithm test results (92.98%) and the recognition accuracy of stationary, walking, running, squatting, hand raising, seated position, lunge, and other movement states were higher than those of recurrent neural network (RNN) (52.22%) and deep neural network (DNN) (40.21%) algorithms ( P < 0.05 ); after 1 year of exercise, the 800/1,000 m running performance, standing long jump performance, 1 min sit-up performance, vital capacity, maximal ventilator volume per minute (MVV) and vital capacity/body mass index, stroke volume (SV), cardiac output (CO), ejection fraction (EF), heart rate (HR), ejection time (ET), mean systolic ejection rate (MSER), mean velocity of circumferential fibre shortening (MVCF) of the observation group were superior to those of the control group, and tension, anger, depression, fatigue, panic, energy, self-related emotions, and the improvement of total mood disturbance (TMD) score were also superior to those before exercise. The improvement effect of physical fitness, cardiopulmonary function, and mentality of college students in the observation group was better than that in the control group ( P < 0.05 ). Aerobic exercise can effectively improve the physical fitness, cardiopulmonary function, and adverse mentality of college students, and through the assistance of machine learning, exercise recognition technology can further improve the effect of aerobic exercise, which is worthy of application and promotion.

show abstract

Section: Discussionmentioning

confidence: 99%

Machine Learning-Based Aerobic Exercise Recognition and Its Effect on Health Status of College Students

Zhang

Lai

et al. 2022

Mobile Information Systems

View full text Add to dashboard Cite

show abstract

“…Chen et al introduced a CNN-based algorithm for the non-invasive and high-resolution detection of breast tumors through Magnetic Detection Electrical Impedance Tomography. The results indicated that the relative reconstruction error with the CNN algorithm could be reduced to 10% compared to the Truncated Singular Value Decomposition algorithm and the Backpropagation algorithm [30]. Li et al proposed a Densely Connected Convolutional Neural Network to improve image reconstruction in Electrical Resistance Tomography, effectively mitigating the issues of information and gradient vanishing and significantly enhancing the accuracy and visual quality of the reconstructed images [31].…”

Section: Convolution Patterns In Emt: a Projection-based Approachmentioning

confidence: 99%

Rotational Convolution Design in Convolutional Neural Networks for Direct 3D Electromagnetic Tomography

Zhao,

Liu

2024

Applied Sciences

View full text Add to dashboard Cite

The three-dimensional (3D) reconstruction of Electromagnetic Tomography (EMT) is an important task for many applications, such as the non-destructive testing of inner defects in rail systems. Additionally, image reconstruction algorithms utilizing deep learning methods have been verified to be useful in recent years. Therefore, the interpretability of deep learning is a question that is relevant to its application in other areas. This paper proposes an innovative rotational convolution pattern, Conv-P, for convolutional neural network (CNN) image reconstruction in a 3D EMT system. This pattern is based on the projection relationships inherent in tomographic imaging, where each convolution is performed on adjacent projections along the excitation rotation direction. The advantage of this pattern is that it can generate the convolution process by utilizing the 3D structural information from real sensors. To verify the effectiveness of this convolution pattern, we constructed a 3D dual-layer 16-coil EMT model and tested its image reconstruction performance. The results demonstrate that, compared with two common convolution patterns, Conv-P achieves a 4.7% and 4.1% increase in the Image Correlation Coefficient (CC), a 19.8% and 13.1% reduction in the Relative Image Error (IE), a 0.67% and 1.59% increase in the Peak Signal-to-Noise Ratio (PSNR), and a 3.24% and 0.74% increase in the Structural Similarity Index Measure (SSIM).

show abstract

“…CNN is a typical feed-forward neural network used to extract data containing spatial information, such as image and video recognition and classification [24]. The CNN comprises the convolutional, pooling, and fully connected layers.…”

Section: F Cnnmentioning

confidence: 99%

Inheritance and Innovation Development of Sports Based on Deep Learning and Artificial Intelligence

Zhou,

2023

IEEE Access

View full text Add to dashboard Cite

This work aims to speed up the intelligent construction of Sports and Leisure Characteristic Towns (SLCT) in the new era. AI and robot technologies have become the pillar of scientific research. First, the MobileNetV2 is simplified and improved using robot technology to design a lightweight Image Recognition (IR) system. The proposed IR system integrates the framework of Facial Recognition (FR) and Automatic License Plate Recognition (ALPR). Then, a hardware acceleration scheme for Convolution Neural Network (CNN) is designed to accelerate the calculation of the improved MobileNetV2-based IR system. Finally, an experiment is designed to verify the performance of the proposed IR system on FR and ALPR tasks. The results are summarized as follows. (1) For FR performance, the proposed IR system achieves 99.46% and 99.23% recognition accuracy for CelebFaces and CASIA-Web face data sets on the Xilinx VC707 platform, respectively. Thus, it can recognize facial images efficiently and timely on devices without Graphics Processing Unit (GPU). ( 2) Regarding ALPR performance, the proposed IR system uses NVIDIA Titan X GPU to process 960 × 640-pixel images with 30/Frames Per Second (FPS). Hence, it meets the requirements of real-time processing. (3) The proposed hardware acceleration scheme is tested on LeNet to generalize the acceleration effect. Consequently, the proposed convolution acceleration scheme can lend to common CNN. Compared with Eyeriss, the performance of the LeNet convolution layer is significantly improved in efficiency and resource consumption. Therefore, the proposed acceleration scheme is excellent and feasible. In conclusion, the lightweight neural network can obtain high accuracy and good generalization. The proposed hardware acceleration scheme can accelerate CNN operation. The findings lay the foundation for developing robot technology and constructing the intelligent SLCT. It contributes to the inheritance and innovative development of intelligent towns.

show abstract

A convolutional neural network algorithm for breast tumor detection with magnetic detection electrical impedance tomography

Cited by 7 publications

References 25 publications

Machine Learning-Based Aerobic Exercise Recognition and Its Effect on Health Status of College Students

Machine Learning-Based Aerobic Exercise Recognition and Its Effect on Health Status of College Students

Rotational Convolution Design in Convolutional Neural Networks for Direct 3D Electromagnetic Tomography

Inheritance and Innovation Development of Sports Based on Deep Learning and Artificial Intelligence

Contact Info

Product

Resources

About