Red Blood Cell Aggregation Classification Based on Ultrasonic Radiofrequency Echo Signals by An Improved Convolutional Neural Network
Zerong Liao, Yufeng Zhang · 2022 3rd International Conference on Computer Vision, Image and Deep Learning & International Conference on Computer Engineering and Applications (CVIDL & ICCEA) · 2022
In order to solve the main problem that traditional machine learning methods rely on manual experience to extract features, the method based on ultrasonic RF signal convolutional neural network is proposed to extract features for red blood cell (RBC) classification. Considering that VGG16 network uses small convolution kernel, which not only reduces network parameters, but also improves classification accuracy. Therefore, a compact network model RBCA-VGG10 is proposed by deleting layers and adjusting the structure of VGG16. Compared with LeNet, AlexNet, GoogleNet, ResNet models, the average classification accuracies of RBCA-VGG10 are improved by 10.15%, 9.30%, 4.40%, 8.34%, respectively, and no overfitting. Compared with B-mode images method and the RF signals based on empirical wavelet transform method by support vector machine (SVM) classifier, the average classification accuracies of RBCA-VGG10 model are improved by 8.97% and 6.52%, respectively.