A review on Natural Language Processing Models for COVID-19 research.

This survey paper reviews Natural Language Processing Models and their use in COVID-19 research in two main areas. Firstly, a range of transformer-based biomedical pretrained language models are evaluated using the BLURB benchmark. Secondly, models used in sentiment analysis surrounding COVID-19 vaccination are evaluated. We filtered literature curated from various repositories such as PubMed and Scopus and reviewed 27 papers. When evaluated using the BLURB benchmark, the novel T-BPLM BioLinkBERT gives groundbreaking results by incorporating document link knowledge and hyperlinking into its pretraining. Sentiment analysis of COVID-19 vaccination through various Twitter API tools has shown the public's sentiment towards vaccination to be mostly positive. Finally, we outline some limitations and potential solutions to drive the research community to improve the models used for NLP tasks.

Deep imitation learning for 3D navigation tasks.

Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples.

Automatic screening and classification of diabetic retinopathy and maculopathy using fuzzy image processing.

Digital retinal imaging is a challenging screening method for which effective, robust and cost-effective approaches are still to be developed. Regular screening for diabetic retinopathy and diabetic maculopathy diseases is necessary in order to identify the group at risk of visual impairment. This paper presents a novel automatic detection of diabetic retinopathy and maculopathy in eye fundus images by employing fuzzy image processing techniques. The paper first introduces the existing systems for diabetic retinopathy screening, with an emphasis on the maculopathy detection methods. The proposed medical decision support system consists of four parts, namely: image acquisition, image preprocessing including four retinal structures localisation, feature extraction and the classification of diabetic retinopathy and maculopathy. A combination of fuzzy image processing techniques, the Circular Hough Transform and several feature extraction methods are implemented in the proposed system. The paper also presents a novel technique for the macula region localisation in order to detect the maculopathy. In addition to the proposed detection system, the paper highlights a novel online dataset and it presents the dataset collection, the expert diagnosis process and the advantages of our online database compared to other public eye fundus image databases for diabetic retinopathy purposes.

The evaluation of i-SIDRA - a tool for intelligent feedback - in a course on the anatomy of the locomotor system.

OBJECTIVE: This paper presents an empirical study of a formative mobile-based assessment approach that can be used to provide students with intelligent diagnostic feedback to test its educational effectiveness. METHOD: An audience response system called SIDRA was integrated with a neural network-based data analysis to generate diagnostic feedback for guided learning. A total of 200 medical students enrolled in a General and Descriptive Anatomy of the Locomotor System course were taught using two different methods. Ninety students in the experimental group used intelligent SIDRA (i-SIDRA), whereas 110 students in the control group received the same training but without employing i-SIDRA. RESULTS: In the students' final exam grades, a statistically significant difference was found between those students that used i-SIDRA as opposed to a traditional teaching methodology (T(162)=2.597; p=0.010). The increase in the number of correct answers during the feedback guided learning process from the first submission to the last submission in four multiple choice question tests was also analyzed. There were average increases of 20.00% (Test1), 11.34% (Test2), 8.88% (Test3) and 13.43% (Test4) in the number of correct answers. In a questionnaire rated on a five-point Likert-type scale, the students expressed satisfaction with the content (M=4.2) and feedback (M=3.5) provided by i-SIDRA and the methodology (M=4.2) used to learn anatomy. CONCLUSIONS: The use of audience response systems enriched with feedback such as i-SIDRA improves medical degree students' performance as regards anatomy of the locomotor system. The knowledge state diagrams representing students' behavior allow instructors to study their progress so as to identify what they still need to learn.

An Empirical Study of Neural Network-Based Audience Response Technology in a Human Anatomy Course for Pharmacy Students.

This paper presents an empirical study of a formative neural network-based assessment approach by using mobile technology to provide pharmacy students with intelligent diagnostic feedback. An unsupervised learning algorithm was integrated with an audience response system called SIDRA in order to generate states that collect some commonality in responses to questions and add diagnostic feedback for guided learning. A total of 89 pharmacy students enrolled on a Human Anatomy course were taught using two different teaching methods. Forty-four students employed intelligent SIDRA (i-SIDRA), whereas 45 students received the same training but without using i-SIDRA. A statistically significant difference was found between the experimental group (i-SIDRA) and the control group (traditional learning methodology), with T (87) = 6.598, p < 0.001. In four MCQs tests, the difference between the number of correct answers in the first attempt and in the last attempt was also studied. A global effect size of 0.644 was achieved in the meta-analysis carried out. The students expressed satisfaction with the content provided by i-SIDRA and the methodology used during the process of learning anatomy (M = 4.59). The new empirical contribution presented in this paper allows instructors to perform post hoc analyses of each particular student's progress to ensure appropriate training.

Neural network-based data analysis for medical-surgical nursing learning.

This paper presents the results of a project on neural network-based data analysis for knowledge clustering in a second-year course on medical-surgical nursing. Data was collected from 208 nursing students which performed one Multiple Choice Question (MCQ) test at the end of the first term. A total of 23 pattern groups were created using snap-drift. Data obtained can be integrated with an on-line MCQ system for training purposes. Findings about how students are classified suggest that the level of knowledge of the individuals can be addressed by customized feedback to guide them towards a greater understanding of particular concepts.

Snap-drift neural network for self-organisation and sequence learning.

This paper presents two novel neural networks based on snap-drift in the context of self-organisation and sequence learning. The snap-drift neural network employs modal learning that is a combination of two modes; fuzzy AND learning (snap), and Learning Vector Quantisation (drift). We present the snap-drift self-organising map (SDSOM) and the recurrent snap-drift neural network (RSDNN). The SDSOM uses the standard SOM architecture, where a layer of input nodes connects to the self-organising map layer and the weight update consists of either snap (min of input and weight) or drift (LVQ, as in SOM). The RSDNN uses a simple recurrent network (SRN) architecture, with the hidden layer values copied back to the input layer. A form of reinforcement learning is deployed in which the mode is swapped between the snap and drift when performance drops, and in which adaptation is probabilistic, whereby the probability of a neuron being adapted is reduced as performance increases. The algorithms are evaluated on several well known data sets, and it is found that these exhibit effective learning that is faster than alternative neural network methods.