Towards automatic encoding of medical procedures using convolutional neural networks and autoencoders
作者：
Highlights：
• We proposed neural network based methods for encoding of Swiss clinical procedure code (CHOP).
• Proposed methods based on CNN and autoencoder are evaluated with logs from real production servers.
• Our methods outperform SVM, logistic regression on the task of relevance determination.
• Semantic enrichment and stepwise pre-training facilitate the model generalization.
• Models using autoencoder and semantic enrichment outperform CNN based models in our settings.
摘要
•We proposed neural network based methods for encoding of Swiss clinical procedure code (CHOP).•Proposed methods based on CNN and autoencoder are evaluated with logs from real production servers.•Our methods outperform SVM, logistic regression on the task of relevance determination.•Semantic enrichment and stepwise pre-training facilitate the model generalization.•Models using autoencoder and semantic enrichment outperform CNN based models in our settings.
论文关键词：Automatic encoding,Deep learning,Classification system,Autoencoder,Convolutional neural networks
论文评审过程：Received 20 October 2017, Revised 7 September 2018, Accepted 3 October 2018, Available online 29 October 2018, Version of Record 1 February 2019.
论文官网地址：https://doi.org/10.1016/j.artmed.2018.10.001
