Curriculum Contrastive Learning for COVID-19 FAQ Retrieval

Medical Frequently Asked Question (FAQ) retrieval aims to find the most relevant question-answer pairs for a given user query, which is of great significance for enhancing people medical health awareness and knowledge. However, due to medical data privacy and labor-intensive labeling, there is a lack of large-scale question-matching training datasets. Previous methods directly use the collected question-answer pairs on search engines to train retrieval models, which achieved poor performance. Inspired by recent advances in contrastive learning, we propose a novel contrastive curriculum learning framework for modeling user medical queries. First, we design different data augmentation methods to generate positive samples and different types of negative samples. Second, we propose a curriculum learning strategy that associates difficulty levels with negative samples. Through a contrastive learning process from easy to hard, our method achieves excellent results on two medical datasets. © 2022 IEEE.

RAPID DETECTION OF NOVEL CORONAVIRUS SARS-COV-2 BY SOLID-STATE NANOPORE

The COVID-19 pandemic spreads rapidly and globally. To quell the pandemic propagation, rapid and accurate detection of SARS-CoV-2 is urgently needed. Here, we present a nanopore coupled RT-LAMP method for SARSCoV-2 detection. After comparing all information from the nanopore experiment, we develop a method to use the event rate change of the amplicons translocation event to measure the amplification. As a result, our platform can distinguish positive from negative samples in 15 min with around 65 copies/reaction limit of detection and 100% specificity. We believe that the nanopore coupled RT-LAMP platform would provide a sensitive and specific detection for SARS-COV-2. © 2021 MicroTAS 2021 - 25th International Conference on Miniaturized Systems for Chemistry and Life Sciences. All rights reserved.

Fast Mixing of Hard Negative Samples for Contrastive Learning and Use for COVID-19

The new type of coronavirus pneumonia (COVID-19) is spreading around the world, and one of the main reasons is the low detection efficiency and lack of detection materials. Image recognition algorithms based on deep learning are an effective method to assist epidemic detection. However, COVID-19 medical image data has the problems of difficulty in obtaining case samples and high data labeling costs. Contrastive learning algorithm is the best performing method among self-supervised learning algorithms without label training. To solve the problem of insufficient annotation data for COVID-19 medical images, this paper proposes a contrastive learning algorithm that incorporates feature synthesis components of hard negative samples. The feature of hard negative samples is a key factor to promote network training among a large number of negative samples features. Generating more hard negative samples features and adding them to the contrastive loss function for calculation allows the comparison learning algorithm to learn better semantic features. At the same time, the contrastive learning algorithm proposed in this paper is used in the transfer learning of COVID-19 CT images, which can deal with the problem of large differences between the original domain and the target domain. The algorithm proposed in this paper uses common methods of self-supervised learning to evaluate the classification accuracy of 0.798 on the COVID-CT data set, which has a significant improvement. The training method of transfer learning has better classification performance than common methods, and the classification accuracy is 0.852. © 2021 ACM.