• Laser & Optoelectronics Progress
  • Vol. 56, Issue 1, 011101 (2019)
Baohong Mi1, Wenxue Hong1、*, Jialin Song1, Shiming Wu2, and Hui Meng1
Author Affiliations
  • 1 College of Electrical Engineering, Yanshan University, Qinhuangdao, Hebei 0 66004, China
  • 2 Pain Clinic, Xinqiao Hospital, Army Medical University, Chongqing 400000, China
  • show less
    DOI: 10.3788/LOP56.011101 Cite this Article Set citation alerts
    Baohong Mi, Wenxue Hong, Jialin Song, Shiming Wu, Hui Meng. Myocardial Ischemia Pre-Diagnosis Method Based on Infrared Thermal Imaging and BP Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(1): 011101 Copy Citation Text show less
    Standard posture
    Fig. 1. Standard posture
    Process of myocardial ischemic pre-diagnosis based on infrared imaging data
    Fig. 2. Process of myocardial ischemic pre-diagnosis based on infrared imaging data
    Neck and shoulder positioning process is performed by template matching method
    Fig. 3. Neck and shoulder positioning process is performed by template matching method
    Randomly selected 6 subjects' precardiac localization results
    Fig. 4. Randomly selected 6 subjects' precardiac localization results
    Principle diagram of symmetrical set mirroring differential
    Fig. 5. Principle diagram of symmetrical set mirroring differential
    Two layer neural network topology
    Fig. 6. Two layer neural network topology
    Core size 3×3 different convolution kernel operation results. (a) Original image; (b) Sobel operator processed; (c) Laplace operator processed; (d) Gaussian operator processed
    Fig. 7. Core size 3×3 different convolution kernel operation results. (a) Original image; (b) Sobel operator processed; (c) Laplace operator processed; (d) Gaussian operator processed
    Core size 5×5 different convolution kernel operation results. (a) Original image; (b) Sobel operator processed; (c) Laplace operator processed; (d) Gaussian operator processed
    Fig. 8. Core size 5×5 different convolution kernel operation results. (a) Original image; (b) Sobel operator processed; (c) Laplace operator processed; (d) Gaussian operator processed
    Training set error curve processed by 3×3 Gaussian convolution kernel
    Fig. 9. Training set error curve processed by 3×3 Gaussian convolution kernel
    NuclearsizeSobeloperator /%Gaussiancore /%Laplaceoperator /%
    3×377.7895.5673.33
    5×575.5691.1166.67
    Table 1. Accuracy of different convolution experiments in BP neural network classifier test set
    Baohong Mi, Wenxue Hong, Jialin Song, Shiming Wu, Hui Meng. Myocardial Ischemia Pre-Diagnosis Method Based on Infrared Thermal Imaging and BP Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(1): 011101
    Download Citation