Computational Modeling of Regional Dynamics of Pandemic Behavior using Psychologically Valid Agents (preprint)

Regional Psychologically Valid Agents (R-PVAs) are computational models representing cognition and behavior of regional populations. R-PVAs are developed using ACT-R—a computational implementation of the Common Model of Cognition. We developed R-PVAs to model mask-wearing behavior in the U.S. over the pre-vaccination phase of COVID-19 using regionally organized demographic, psychographic, epidemiological, information diet, and behavioral data. An R-PVA using a set of five regional predictors selected by stepwise regression, a psychological self-efficacy process, and context-awareness of the effective transmission number, Rt, yields good fits to the observed proportion of the population wearing masks in 50 U.S. states [R2 = 0.92].  An R-PVA based on regional Big 5 personality traits yields strong fits [R2 = 0.83].  R-PVAs can be probed with combinations of population traits and time-varying context to predict behavior. R-PVAs are a novel technique to understand dynamical, nonlinear relations amongst context, traits, states, and behavior based on cognitive modeling.

Effects of wearing a KF94 face mask on performance, perceptual, and physiological responses during a resistance exercise (preprint)

Wearing a face mask in indoor public places including fitness centers is an effective strategy to prevent the airborne transmission of COVID-19. However, only a few studies have been performed on wearing a mask during resistance exercise (RE) which is primarily performed in indoor fitness centers. This study aimed to investigate the effects of wearing a KF94 mask on exercise volume, perceptual parameters, and physiological responses during RE. Twenty young men participated in this randomized crossover trial. Participants performed moderate-intensity (1RM 60%) RE sessions in two different conditions (KF94 mask vs. no mask). Cardiorespiratory parameters, exercise volume, rating of perceived exertion (RPE), and dyspnea were measured during RE. Blood lactate concentration, blood pressure, arterial stiffness, and perceptual parameters were measured at pre-exercise and post-exercise. Exercise volume, ventilation volume, and ventilation efficiency parameters were lower with the KF94 mask than without the mask. However, RPE and dyspnea were higher with the KF94 mask than without the mask. Central arterial stiffness at post-exercise was higher with the KF94 mask than without the mask. Therefore, wearing a KF94 mask during RE affects exercise volume, perceptual parameters, and physiological responses, suggesting coaches need to modify RE manipulation variables while wearing a KF94 mask.

Adapting to Pandemic Challenges: Thai Dentists’ Response to COVID-19 Infection Control (preprint)

Background The COVID-19 pandemic has posed emotional challenges for dentists. This study aimed to evaluate the concerns, risk perceptions, and compliance with infection control practices among Thai dentists during and shortly after the COVID-19 pandemic. Insights from this assessment are intended to offer practical strategies to enhance dentists’ readiness for future outbreaks.Methods A questionnaire evaluating dentists’ perceptions of COVID-19 infection and precautionary measures was generated via Google Forms™ and distributed to Thai dentists during the late pandemic period (June-September 2022).Results Of the 467 respondents, 44.3% reported the highest concern level for infection risk when treating emergency patients potentially exposed to COVID-19. Regarding the infection risk in dental practice, 30.6% perceived the highest contracting risk, while 24.6% perceived the highest transmission risk. Notably, 49.7% expressed high confidence in the ability of the vaccine to reduce COVID-19 severity. Eighty-eight percent of the patients supported the continuation of pandemic-era precautionary measures even in the absence of disease. N95 mask usage in dental settings markedly increased during the outbreak. Many dental procedures were not adequately adapted to COVID-19 infection control measures, and personal protective equipment was insufficient. Factors influencing COVID-19 risk perception in dental treatment included gender, involvement in aerosol-generating procedures, and availability of protective equipment.Conclusion Thai dentists expressed significant concerns about the risk of contracting COVID-19 in their practice. The participants had strong confidence in the effectiveness of the vaccines in reducing symptom severity. The majority of the participants supported the continued implementation of pandemic-initiated precautionary measures. Sex, aerosol-generating procedures, and protective equipment availability were key factors influencing dentists’ risk perception. These insights underscore the need for improved infection control measures and resources in dental settings, both to address current concerns and to enhance preparedness for future health crises.

Modeling the Spread of COVID-19 in University Communities (preprint)

Mathematical and simulation models are often used to predict the spread of a disease and estimate the impact of public health interventions, and many such models have been developed and used during the COVID-19 pandemic. This paper describes a study that systematically compared models for a university community, which has a much smaller but more connected population than a state or nation. We developed a stochastic agent-based model, a deterministic compartment model, and a model based on ordinary differential equations. All three models represented the disease progression with the same susceptible-exposed-infectious-recovered (SEIR) model. We created a baseline scenario for a population of 14,000 students and faculty and eleven other scenarios for combinations of interventions such as regular testing, contact tracing, quarantine, isolation, moving courses online, mask wearing, improving ventilation, and vaccination. We used parameter values from other epidemiological studies and incorporated data about COVID-19 testing in College Park, Maryland, but the study was designed to compare modeling approaches to each other using a synthetic population. For each scenario we used the models to estimate the number of persons who become infected over a semester of 119 days. We evaluated the models by comparing their predictions and evaluating their parsimony and computational effort. The agent-based model (ABM) and the deterministic compartment model (DCM) had similar results with cyclic flow of persons to and from quarantine, but the model based on ordinary differential equations failed to capture these dynamics. The ABM's computation time was much greater than the other two models' computation time. The DCM captured some of the dynamics that were present in the ABM's predictions and, like those from the ABM, clearly showed the importance of testing and moving classes on-line.

Quantifying Media Influence on Covid-19 Mask-Wearing Beliefs (preprint)

How political beliefs change in accordance with media exposure is a complicated matter. Some studies have been able to demonstrate that groups with different media diets in the aggregate (e.g., U.S. media consumers ingesting partisan news) arrive at different beliefs about policy issues, but proving this from data at a granular level -- at the level of attitudes expressed in news stories -- remains difficult. In contrast to existing opinion formation models that describe granular detail but are not data-driven, or data-driven studies that rely on simple keyword detection and miss linguistic nuances, being able to identify complicated attitudes in news text and use this data to drive models would enable more nuanced empirical study of opinion formation from media messaging. This study contributes a dataset as well as an analysis that allows the mapping of attitudes from individual news stories to aggregate changes of opinion over time for an important public health topic where opinion differed in the U.S. by partisan media diet: Covid mask-wearing beliefs. By gathering a dataset of U.S. news media stories, from April 6 to June 8, 2020, annotated according to Howard 2020's Face Mask Perception Scale for their statements regarding Covid-19 mask-wearing, we demonstrate fine-grained correlations between media messaging and empirical opinion polling data from a Gallup survey conducted during the same period. We also demonstrate that the data can be used for quantitative analysis of pro- and anti-mask sentiment throughout the period, identifying major events that drove opinion changes. This dataset is made publicly available and can be used by other researchers seeking to evaluate how mask-wearing attitudes were driven by news media content. Additionally, we hope that its general method can be used to enable other media researchers to conduct more detailed analyses of media effects on opinion.

Enhancing COVID-19 Case Forecasting in the United States: A Comparative Analysis of ARIMA, SARIMA, and RNN Models with Grid Search Optimization (preprint)

The COVID-19 pandemic has resulted in a substantial number of fatalities in the United States since its onset in January 2020. In an effort to mitigate the spread of this highly infectious disease, a range of measures, including social distancing, mask-wearing, lockdowns, and vaccination campaigns, have been implemented. However, despite these extensive efforts, the persistent transmission of the virus can be attributed to a combination of vaccine hesitancy among certain individuals and the emergence of new viral strains. To effectively manage the ongoing pandemic, healthcare providers and government officials rely on infectious disease modeling to anticipate and secure the necessary resources. Accurate short-term case number forecasting is of paramount importance for healthcare systems. Since the beginning of the pandemic, numerous models have been employed to forecast the number of confirmed cases. In this study, we undertake a comparative analysis of six time-series techniques: Simple Moving Average (SMA), Exponentially Weighted Moving Average (EWMA), Holt-Winters Double Exponential Smoothing Additive (HWDESA), Autoregressive Integrated Moving Average (ARIMA), Seasonal Autoregressive Integrated Moving Average (SARIMA), and Recurrent Neural Network (RNN), with regard to their modeling and forecasting capabilities. SMA, EWMA, and HWDESA were employed for predictive modeling, while the ARIMA, SARIMA, and RNN models were utilized for case number forecasting. A comprehensive grid search was carried out to determine the optimal parameter combinations for both the ARIMA and SARIMA models. Our research findings demonstrate that the Holt-Winters Double Exponential model outperforms both the Exponentially Weighted Moving Average and Simple Moving Average in predicting the number of cases. On the other hand, the RNN model surpasses conventional time-series models such as ARIMA and SARIMA in terms of its forecasting accuracy. The finding of this study emphasizes the importance of accurately predicting the number of COVID-19 cases, given the substantial loss of lives caused by both the virus itself and the societal responses to it. Equipping healthcare managers with precise tools like Recurrent Neural Networks (RNNs) can enable them to forecast future cases more accurately and enhance their preparedness for effective response.

Mask-up: Investigating Biases in Face Re-identification for Masked Faces (preprint)

AI based Face Recognition Systems (FRSs) are now widely distributed and deployed as MLaaS solutions all over the world, moreso since the COVID-19 pandemic for tasks ranging from validating individuals' faces while buying SIM cards to surveillance of citizens. Extensive biases have been reported against marginalized groups in these systems and have led to highly discriminatory outcomes. The post-pandemic world has normalized wearing face masks but FRSs have not kept up with the changing times. As a result, these systems are susceptible to mask based face occlusion. In this study, we audit four commercial and nine open-source FRSs for the task of face re-identification between different varieties of masked and unmasked images across five benchmark datasets (total 14,722 images). These simulate a realistic validation/surveillance task as deployed in all major countries around the world. Three of the commercial and five of the open-source FRSs are highly inaccurate; they further perpetuate biases against non-White individuals, with the lowest accuracy being 0%. A survey for the same task with 85 human participants also results in a low accuracy of 40%. Thus a human-in-the-loop moderation in the pipeline does not alleviate the concerns, as has been frequently hypothesized in literature. Our large-scale study shows that developers, lawmakers and users of such services need to rethink the design principles behind FRSs, especially for the task of face re-identification, taking cognizance of observed biases.

Reconstructing Masked Faces using Variational Quantized Variational Auto Encoders and Recognition using DCNN-ELM Hybrid Framework (preprint)

The reconstruction of the face has historically been a significant issue in medical and forensic science. The presence of COVID-19 has added a significant new dimension. To model a new face, plastic surgery and informatics are employed, representing cyber forensics with challenges. The classic facial recognition techniques suffer from major drawbacks when face masks are widely used. As a result, new techniques are now being tried and tested to reconstruct a face from a collection of masked facial images. To determine the identification accuracy and other parameters/metrics, these faces are compared to real-world images of the same subject. Our research focuses on the task of post-mask face reconstruction, addressing the pressing need for precise and reliable techniques. We evaluate the effectiveness of three key algorithms: Edge Connect, Gated Convolution, and Hierarchical Variational Vector Quantized Autoencoders (HVQVAE). We use two synthetic datasets, MaskedFace-CelebA and MaskedFace-CelebAHQ, to rigorously assess the quality of reconstructed faces using metrics such as PSNR, SSIM, UIQI, and NCORR. Gated Convolution (GC) emerges as the superior choice in terms of image quality. To validate our findings, we employ five classifiers (Vgg16, Vgg19, ResNet50, ResNet101, ResNET152) and explore Extreme Learning Machine (ELM) and Support Vector Machine (SVM) as novel approaches for face recognition. A comprehensive ablation study reinforces our conclusion that Generative Convolution (GC) excels among the three models. Our research offers valuable insights into face reconstruction amid widespread mask usage, emphasizing innovative methodologies to address contemporary challenges in the field.

The Effect of N95 Respirators on Vital Parameters, PETCO2 and Physical Comfort among Healthcare Providers at the Pandemic Clinics (preprint)

Background: Wearing face shields and masks, which used to have very limited public use before the Covid-19 outbreak, has been highly recommended by organizations, such as CDC and WHO, during this pandemic period. The aim of this prospective study is to scrutinize the dynamic changes in vital parameters, change in end tidal CO2 levels, the relationship of these changes with taking a break, and the subjective complaints caused by respiratory protection while healthcare providers are performing their duties with the N95 mask. Methods The prospective cohort included 54 healthcare workers (doctors, nurses, paramedics) who worked in the respiratory unit of the emergency department (ED), performed their duties by wearing valved N95 masks, face shields. The vital parameters and end-tidal CO2 levels were measured at 0-4th-5th-and 9th hours of the work-shift. Results Only the decrease in diastolic BP between 0-9 hours was statistically significant (p=0.038). Besides, MAP values indicated a significant decrease between 0-9 hours and 5-9 hours (p= 0.024 and p=0.049, respectively). In terms of the vital parameters of the subjects working with and without breaks, only PETCO2 levels of those working uninterruptedly increased significantly at the 4th hour in comparison to the beginning-of-shift baseline levels (p=0.003). Conclusion Although the decrease in SBP and MAP values is assumed to be caused by increased fatigue due to workload and work pace as well as increase in muscle activity, the increase in PETCO2 levels in the ED healthcare staff working with no breaks between 0-4 hours should be noted in terms of PPE-induced hypoventilation.

Beneficial Effects of novel Aureobasidium Pullulans strains produced Beta-1,3-1,6 Glucans on Interleukin-6 and D-Dimer levels in COVID-19 patients; results of a randomized multiple-arm pilot clinical study (preprint)

Objective: Cytokine storm and Coagulopathy have been implicated as major causes of morbidity and mortality in COVID-19 patients. A black yeast Aureobasidium pullulans AFO-202 strain produced beta 1,3-1,6 glucan has been reported to offer potential immune enhancement and metabolism balancing, as well as mitigation of coagulopathy risks. The N-163 strain produced beta glucan is an efficient anti-inflammatory immune modulator. In this pilot clinical study, we report the beneficial effects of these two beta glucans on the biomarkers for cytokine storm and coagulopathy in COVID-19 patients. Methods: : A total of 24 RT-PCR positive COVID-19 patients were recruited (Age range: 18~62; 17 males and 7 females). Patients were randomly divided into three groups (Gr): Gr. 1 control (n=8); Gr. 2: AFO-202 beta glucan (n=8); and Gr. 3, a combination of AFO-202 and N-163 beta glucans (n=8). All three groups received the standard care while groups 2 and 3 received additional supplementation of beta glucans for 30 days. In addition to basic clinical parameters, we periodically evaluated D-Dimer, IL-6, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), the neutrophil to lymphocyte ratio (NLR), the lymphocyte to CRP ratio (LCR) and the leukocyte-CRP ratio (LeCR). Results: : The duration of hospital stay for all three groups was nearly equivalent. There was no mortality of the subjects in any of the groups. Intermittent oxygen was administered from day of admission for up to four to five days with mask (two to four Lpm) to two subjects in Gr. 2 and one subject in Gr. 3. None of the subjects required ventilation. The D-Dimer values in Gr. 1, which was on average 751 ng/ml at baseline, decreased to 143.89 ng/ml on day 15, but increased to 202.5 ng/ml on day 30, which in groups 2 and 3 decreased on day 15 and continued to remain at normal levels until day 30. IL-6 levels decreased on day 15 from an average of 7.395 pg/ml to 3.16 pg/ml in the control, 26.18 pg/ml to 6.94 pg/ml in Gr. 2 and 6.25 pg/ml to 5.22 pg/ml in Gr. 3. However, when measured on day 30, in Gr. 1, the IL-6 increased to 55.37 pg/ml while there was only slight marginal increase in Gr. 2 but within normal range, and the levels further decreased to less than 0.5 pg/ml in Gr. 3. The same trend was observed with ESR. LCR and LeCR increased significantly in Gr. 3. NLR decreased significantly in groups 2 and 3. There was no difference in CRP within the groups. Conclusion: In this exploratory study, consumption of Aureobasidium pullulans produced beta glucans for thirty days, results in a significant control of IL6, D-Dimer and NLR, a significant increase in LCR, LeCR and marginal control of ESR in COVID-19 patients. As these beta glucans are well known food supplements with decades of a track record for safety, based on these results, we recommend larger multi-centric clinical studies to validate their use as an adjunct in the management of COVID-19 and the ensuing long COVID-19 syndrome.

KNOWLEDGE OF PERSONAL PROTECTIVE EQUIPMENT USE OF SENIOR PARAMEDIC STUDENTS BEFORE GRADUATION DURING COVID-19 PANDEMIC PERIOD (preprint)

Introduction: This study is planned to determine the knowledge of personal protective equipment use of senior paramedic students before graduation during COVID-19 pandemic. Methods: The research is a descriptive study. The universe of the study consisted of 180 senior paramedic students studying at three universities in İzmir. When evaluating the study data, descriptive findings were expressed as percentage, mean, standard deviation and median. Results: 38.1% of the students are between 18-20 years old, 50.0% are between 21-23 years old and 11.9% are 23 years old and above. 58.3% of the students participating in the study are females and 41.7% are males. In the study, 74.4% of paramedic students stated that healthcare personnel working in ambulance or patient transport vehicles should use medical masks, gowns, gloves and eye protection while transporting suspected COVID-19- (SARS-CoV-2) patients to the health institution. 78.5% of the students answered that an N95/FFP2 mask should be used while intervening and taking samples in a patient suspected of having COVID-19. Discussion: As a result, it was found that during the COVID-19 pandemic period, the knowledge of personal protective equipment use of senior paramedic students before graduation is sufficient in some cases and not clear and sufficient in some cases. It is recommended that students should be given effective training on the use of PPE during the intervention of the patient with COVID-19 before graduation.

Covid-19 outbreak: does the use of a surgical mask impact the sense of smell? A crossover prospective comparative study (preprint)

- We hypothesized that the surgical mask could filter some odorant particles, leading to a transient hyposmia. - A crossover prospective comparative study between 2 groups of 10 healthy volunteers was conducted to evaluate the impact of wearing a surgical mask on sense of smell by comparing the results of sniffin’ sticks test (SST) with and without a surgical mask. - All the subjects, except one, had a significantly better total score (TDI) without a mask. - 4/20 (20%) were normosmic without a mask, while being recategorized as hyposmic with a mask. - Wearing a surgical mask may reduce the sense of smell, in a cohort of young normosmic patients.

A Study to Assess Safety and Improvement in Skin Barrier After Using 2 Facial Sunscreens in Adults With Sensitive Skin

Adherence to mask wearing and social distancing following vaccination and use of lateral flow testing during the COVID-19 pandemic in England and Wales: Results from a cross-sectional study nested within the prospective Virus Watch household community cohort study (preprint)

Background Personal protective behaviours (PPBs) played a crucial role in limiting the spread of infection during the COVID-19 pandemic, yet adherence to these behaviours varied at population level. Understanding the factors influencing adherence to protective behaviours is important, as PPBs will be a critical part of the response in future pandemics. Using behavioural science, we investigated the influences on adherence to PPBs, focusing on face mask wearing, social distancing, and lateral flow testing (LFT). Methods Two online surveys, the first gathering data on mask wearing and social distancing, and the second on lateral flow testing, were conducted in July and August 2021 with a sample from England and Wales (N = 20,488 (survey 1) and N = 26,613 (survey 2)). The survey questions were designed based on the Capability, Opportunity, Motivation (COM-B) model of Behavior. Multivariate models were used to examine associations between identified influences and adoption of these protective behaviours. Results Most respondents reported wearing a face mask in public indoor places (88.5%) and maintaining a 1+ metre distance (86.8%) all or most of the time. After two doses of COVID-19 vaccine, social distancing decreased with 48.3% reporting meeting friends or family and 38.3% visiting indoor places more frequently. Motivation, Opportunity and Capability factors were significantly associated with increased odds of wearing a face covering and social distancing. Among individuals who indicated using an LFT (comprising 68% of the total sample), 50.4% reported engaging in routine testing. For those who had never used an LFT, the predominant reason cited was a perceived lack of necessity for testing (55.3%). Statistically significant associations were found between routine testing and accurate interpretation of test results across all LFT belief-based statements (p < 0.05). Conclusions Findings indicated high levels of adherence to face masks, social distancing, and lateral flow testing, even amid reduced restrictions and high vaccination rates. Utilising a behavioural science framework, factors related to capability, opportunity, and motivation were found to significantly influence the use of these protective behaviours. Our recommendations can inform public health intervention design and guide the selection of implementation strategies for public health emergencies preparedness.

Women Focused Encounters for Resilience Independence Strength and Eudaimonia

The Exploration of Prolonged Mask Use on Appearance Anxiety during the COVID-19 Pandemic in Taiwan (preprint)

(1) Background: This study aimed to explore how prolonged mask use during the COVID-19 pandemic affected appearance anxiety; (2) Methods: It implemented a quantitative–qualitative mixed-research approach based on a convenience sample of young people in Taiwan; (3) Results: A total of 1,118 participants completed the online questionnaire. A total of 49 participants reached the 55-point threshold of Appearance Anxiety Scale. Higher score of appearance anxiety was correlated to higher state anxiety statistically (r = 0.4, p < 0.001). Ten participants were interviewed and qualitative content analysis revealed 5 main themes and 16 subthemes. Three of the main themes were related to the positive effects of wearing masks for pandemic prevention and reducing appearance anxiety: “Wearing masks indicates compliance with the epidemic prevention requirement,” “Wearing masks functions as protection against lack of confidence in appearance,” and “Wearing masks helps mitigate the fear of being judged by others.” The remaining two “Wearing masks shifts attention from appearance to figure” and “Prolonged mask wearing exacerbates the anxiety about taking it off” were related to the impacts of long-term mask use; (4) Conclusions: results enhance understanding of negative emotional experiences and anxiety related to physical appearance.

Perceived Risk of Developing Severe Illness or Complications Due to Covid-19 in the Brazilian Population (preprint)

Background The perception of the risk of developing serious illness as a result of COVID-19 was one of the first reports used to reflect the health condition of infected people after hospital admission. The effects of COVID-19 are more severe in individuals with chronic noncommunicable diseases (NCDs), indicating that the characteristics and implications of these diseases in people with COVID-19 need to be investigated.Methods This cross-sectional study was was carried out with 1961 people aged 18 or older living in Brazil. An open research approach (survey) was used for the sample design, which involved the use of an online questionnaire. Descriptive statistical analysis and logistic regression were applied to identify factors associated with the perceived risk of complications due to COVID-19.Results The sample was mostly composed of women (n = 1383; 70.5%), 18 to 39 years old (n = 1144; 58.3%), and white (n = 1140; 56.4%). It was possible to observe that people who perceived a risk of developing diseases or complications if they became infected with COVID-19 were more likely to have a chronic noncommunicable disease (NCD) (OR: 4.51; 95% CI: 3.61–5.65), self-perception of potential risk of becoming infected with COVID-19 (OR: 2.34; 95% CI: 1.87–2.93), self-perceived potential risk of the population becoming infected with COVID-19 (OR: 5.80; 95% CI: 3.30–10.74), wearing a protective mask (OR: 12.98; 95% CI: 5.8–31.35) during the pandemic period and having a religion (OR: 1.29; 95% CI: 1.02–1.63).Conclusions The study showed that the perception of the risk of developing a severe form of the disease was significant in certain groups, such as religious people or those with chronic noncommunicable diseases.

Masking Behaviors in Epidemiological Networks with Cognitively-plausible Reinforcement Learning (preprint)

The COVID-19 pandemic highlighted the critical role of human behavior in influencing infectious disease transmission and the need for models capturing this complex dynamic. We present an agent-based model integrating an epidemiological simulation of disease spread with a cognitive architecture driving individual mask-wearing decisions. Agents decide whether to mask based on a utility function weighting factors like peer conformity, personal risk tolerance, and mask-wearing discomfort. By conducting experiments systematically varying behavioral model parameters and social network structures, we demonstrate how adaptive decision-making interacts with network connectivity patterns to impact population-level infection outcomes. The model provides a flexible computational framework for gaining insights into how behavioral interventions like mask mandates may differentially influence disease spread across communities with diverse social structures. Findings highlight the importance of integrating realistic human decision processes in epidemiological models to inform policy decisions during public health crises.

A Graphical Approach to Treatment Effect Estimation with Observational Network Data (preprint)

We propose an easy-to-use adjustment estimator for the effect of a treatment based on observational data from a single (social) network of units. The approach allows for interactions among units within the network, called interference, and for observed confounding. We define a simplified causal graph that does not differentiate between units, called generic graph. Using valid adjustment sets determined in the generic graph, we can identify the treatment effect and build a corresponding estimator. We establish the estimator's consistency and its convergence to a Gaussian limiting distribution at the parametric rate under certain regularity conditions that restrict the growth of dependencies among units. We empirically verify the theoretical properties of our estimator through a simulation study and apply it to estimate the effect of a strict facial-mask policy on the spread of COVID-19 in Switzerland.

Facial Emotion Recognition Under Mask Coverage Using a Data Augmentation Technique (preprint)

Identifying human emotions using AI-based computer vision systems, when individuals wear face masks, presents a new challenge in the current Covid-19 pandemic. In this study, we propose a facial emotion recognition system capable of recognizing emotions from individuals wearing different face masks. A novel data augmentation technique was utilized to improve the performance of our model using four mask types for each face image. We evaluated the effectiveness of four convolutional neural networks, Alexnet, Squeezenet, Resnet50 and VGGFace2 that were trained using transfer learning. The experimental findings revealed that our model works effectively in multi-mask mode compared to single-mask mode. The VGGFace2 network achieved the highest accuracy rate, with 97.82% for the person-dependent mode and 74.21% for the person-independent mode using the JAFFE dataset. However, we evaluated our proposed model using the UIBVFED dataset. The Resnet50 has demonstrated superior performance, with accuracies of 73.68% for the person-dependent mode and 59.57% for the person-independent mode. Moreover, we employed metrics such as precision, sensitivity, specificity, AUC, F1 score, and confusion matrix to measure our system's efficiency in detail. Additionally, the LIME algorithm was used to visualize CNN's decision-making strategy.