Can pre-trained convolutional neural networks be directly used as a feature extractor for video-based neonatal sleep and wake classification?

BMC Research Notes

Muhammad Irfan¹  Saadullah Farooq Abbasi¹  Saeed Akbarzadeh¹  Muhammad Awais¹  Chen Chen¹  Wei Chen²  Bin Yin³  Xi Long⁴  Laishuan Wang⁵  Chunmei Lu⁵  Xinhua Wang⁶ 

[1] Center for Intelligent Medical Electronics, Department of Electronic Engineering, School of Information Science and Technology, Fudan University, 200433, Shanghai, China;Center for Intelligent Medical Electronics, Department of Electronic Engineering, School of Information Science and Technology, Fudan University, 200433, Shanghai, China;China, and Human Phenome Institute Fudan University, Shanghai, China;Connected Care and Personal Health Department, Philips Research, Shanghai, China;Department of Electrical Engineering, Eindhoven University of Technology, Den Dolech 2, 5612 AZ, Eindhoven, The Netherlands;Department of Family Care Solutions, Philips Research, 5656 AE, Eindhoven, The Netherlands;Department of Neonatology, Children’s Hospital of Fudan University, 200032, Shanghai, China;Department of Neurology, Children’s Hospital of Fudan University, 200032, Shanghai, China;

关键词: Convolutional neural networks (CNNs);    Video electroencephalogram (VEEG);    Neonatal sleep;    Sleep and wake classification;    Feature extraction;   

DOI  :  10.1186/s13104-020-05343-4

来源: Springer

【 摘 要 】

ObjectiveIn this paper, we propose to evaluate the use of pre-trained convolutional neural networks (CNNs) as a features extractor followed by the Principal Component Analysis (PCA) to find the best discriminant features to perform classification using support vector machine (SVM) algorithm for neonatal sleep and wake states using Fluke® facial video frames. Using pre-trained CNNs as a feature extractor would hugely reduce the effort of collecting new neonatal data for training a neural network which could be computationally expensive. The features are extracted after fully connected layers (FCL’s), where we compare several pre-trained CNNs, e.g., VGG16, VGG19, InceptionV3, GoogLeNet, ResNet, and AlexNet.ResultsFrom around 2-h Fluke® video recording of seven neonates, we achieved a modest classification performance with an accuracy, sensitivity, and specificity of 65.3%, 69.8%, 61.0%, respectively with AlexNet using Fluke® (RGB) video frames. This indicates that using a pre-trained model as a feature extractor could not fully suffice for highly reliable sleep and wake classification in neonates. Therefore, in future work a dedicated neural network trained on neonatal data or a transfer learning approach is required.

【 授权许可】

CC BY