A Lightweight CNN Architecture for Chest X-ray Images Analysis to Facilitate Early Stage Detection of Lung Diseases

Lung-related diseases are one of the significant causes of death among infants and children. However, the mortality rate can be reduced by the detection of lung abnormality at an early stage. Traditionally, radiologists identify irregularities by interpreting chest x-ray images which is time-consuming. Therefore, researchers have proposed many automated systems for diagnosing pneumonia and other lung-related diseases. Due to the remarkable performance of Convolutional Neural Networks(CNN) in image classification, it has gained immense popularity in chest x-ray image analysis. Most of the research has utilized famous pre-trained Imagenet models for more accurate analysis of Chest X-ray images. However, the problem with these architectures is that they have many parameters that increase the training time, which makes the detection process lengthy. This paper introduces a lightweight, compact, and well-tuned CNN architecture with far fewer parameters than the pre-trained model to analyze two of the most common lung diseases, pneumonia and Covid-19. We have evaluated our model on two benchmark datasets. Experimental results show that our lightweight CNN model has far fewer hyperparameters than other state-of-the-art models but achieves similar results. We have achieved an accuracy of 90.38% on the kermany dataset and 96.90% on the Covid-19 Radiography dataset. © 2022 IEEE.

Perceived Stress Analysis of Undergraduate Students during COVID-19: A Machine Learning Approach

Stress is a state of mind when an individual experiences emotional or physical tensions originating from any event that results in frustration, anger, or nervousness. Unfortunately, since the inception of the COVID-19 pandemic, it has been massively witnessed among university students due to persistent usage of e-learning gears for the last two years. Due to the severity of the observed stress, accurate and early prediction and detection should play a pivotal role in treating a student. In this work, a questionnaire-based dataset on Jordanian students has been analyzed using the 5-point Likert Scale. One of the most widely used psychological instrument Perceived Stress Scale (PSS) is used to identify the stressrelated symptoms of the students. Based on the dataset, several machine learning (ML) algorithms were applied for regression and classification analysis by which mental stress has been predicted and classified. After simulation in Python, the ML regressors were evaluated through the performance metrics such as R2 Score, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Percentage Absolute Error (MAPE), and ML classifiers were assessed by accuracy, precision, recall, and F1-Score. It is observed that Linear Regression (LR) performed the best among all the regression models whereas the Logistic Regression Classifier (LRC) portrayed the highest accuracy of 97.8% among all the classifiers. Therefore, ML-based stress analysis can significantly contribute to analyzing students' mental stress during COVID-19 in an automated manner.

Engineering Students' Readiness for Online Learning Amidst the COVID-19 Pandemic: Scale Validation and Lessons Learned from a Developing Country

The recent outbreak of the COVID-19 pandemic forced education institutes to shift to an internet-based online delivery mode. This unique situation accelerates a long-standing issue of digital inequality among the students in education and warrants a concentrated study to investigate students' readiness for learning in online environment. This study developed an instrument to meticulously measure the students' readiness for online learning in a pandemic situation. The proposed model consists of (a) motivation, (b) self-efficacy, and (c) situational factors. The proposed model was validated with the engineering students (for pilot study N = 68 and main study N = 988) from several universities in Bangladesh. To validate the underlying relationships between the latent constructs, an exploratory factor analysis (EFA) was performed followed by structural equation modelling (SEM) for the construct validity of the measurement model and to assess the model fit. The findings showed that besides motivation and self-efficacy, the situational factors describing the contextual dynamics emerging from the COVID-19 significantly influenced the student's online readiness. We argue that digital inequality is an important factor influencing student readiness for online learning.

Performance Assessment of Machine Learning Classifiers in Detecting Psychological Impact of Postgraduate Students due to COVID-19

Psychological health of postgraduate students has received significant attention with the emergence of COVID-19. However, the detection of these psychological disorders appears to be a prolonged process as rigorous clinical tests and diagnoses are involved. In this context, machine learning algorithms are capable of analyzing the hidden pattern of the data about students' mental health and classifying them into different levels. This study investigates the performances of these algorithms for detecting postgraduate students' psychological impact during this pandemic. An online survey dataset is employed from Mendeley data repository consisting of the responses of Malaysian students based on the questions of General Anxiety Disorder (GAD-7) and concerns about research progress, academic delay, and daily life. Among the classifiers, artificial neural network showcased the highest accuracy of 95.45% whereas Logistic Regression, Linear Kernel Support Vector Machine, and Gaussian Process exhibited accuracies of more than 90%.

Clinical characteristics and outcomes of acute kidney injury in critically ill adults

INTRODUCTION: Acute kidney injury (AKI) is a frequent diagnosis among critically ill patients. However, the incidence and outcome of COVID-19-induced kidney injury have been variably described. This study aimed to identify clinical characteristics, correlates, and outcomes experienced by AKI patients during the COVID-19 pandemic. METHODS: A retrospective analysis was performed using electronic health records of 331 patients (>18 years of age). AKI was defined based on the KDIGO guidelines. 226 patients were included as renal replacement therapy, kidney transplantation, and missing clinical information patients were excluded. The primary outcome was the incidence of AKI. Secondary outcomes were AKI recovery and in-hospital mortality. Cox regression models were used to analyze outcomes. RESULTS: 226 patients, 47.8% developed AKI. The incidence of AKI Stages 1, 2, and 3 were (34.3%, 36.1%, and 29.6%), respectively. Inpatient mortality was 13.7% and 51% had AKI. The COVID positive cohort (80%) was classified as AKI. AKI patients had higher BMI (30.8, IQR:24.1-36.5), SCr (1.8 mg/dL, IQR:0.8-2.1) PT (22.2, IQR:13.4-26.6), INR (2.2, IQR:1.3-2.7), RR (20.8 breaths/ min, IQR:16.5-24), acidosis 43 (39.8%), hypo-osmolality and hyponatremia 45 (41.7%). AKI patients had lower eGFR (58.8, IQR:27.8-88.2 (ml/min/1.73) and GCS (11.7, IQR:9-15). COVID+AKI patients had higher rates of sepsis (30%) and acidosis (35%) compared to non-COVID AKI. When considering AKI as outcomes, AKI patients had higher TOMV (0, IQR: 4-72) and ICU length of stay (78.5 hours) compared to non-AKI patients (24 hours). Modeling revealed the highest mortality hazard for AKI stage 3 (HR: 4.72). AKI groups used anti-infectives (81%), diuretics (42.4%), and vasopressors (38%) more frequently. In the recovery cohort, analgesics (93.5%), anti-infectives (90.3%), and intravenous fluids (100%) use were most common. In the mortality group analgesics (100%), anti-infectives (100%), and vasopressors (100%) were used most frequently. CONCLUSIONS: AKI incidence remains high and is associated with poorer outcomes. UO-based AKI classification was more sensitive than SCr alone. Despite particular medication classes correlating with the increased incidence of AKI, further investigation is warranted to examine a potential direct cause and effect relationship.

Protecting Health Care Workers from Infectious Diseases using Physical Proximity Networks (PPN)

The COVID-19 pandemic is a major global health threat, and Health Care Workers (HCWs) may have an increased risk of infection through occupational exposure. In the case of hospital outbreaks, contact tracing of close physical interaction needs to be performed. In this article, we propose an IoT-connected contact tracing system based on Bluetooth Low Energy (BLE) beacons for subject identification and data transmission. The proposed system consists of BLE receivers, BLE wearable tags, an edge gateway and a cloud server. The system records interaction information such as entering/exiting time of an HCW to isolation rooms in the hospital. The collected data will be further analyzed to inform infection prevention policies. The performance of the proposed system is assessed through qualitative and quantitative experimental results. Finally, the capabilities of the current system and future research directions are briefly discussed.

Protecting Health Care Workers from Infectious Diseases using Physical Proximity Networks (PPN)

The COVID-19 pandemic is a major global health threat, and Health Care Workers (HCWs) may have an increased risk of infection through occupational exposure. In the case of hospital outbreaks, contact tracing of close physical interaction needs to be performed. In this article, we propose an IoT-connected contact tracing system based on Bluetooth Low Energy (BLE) beacons for subject identification and data transmission. The proposed system consists of BLE receivers, BLE wearable tags, an edge gateway and a cloud server. The system records interaction information such as entering/exiting time of an HCW to isolation rooms in the hospital. The collected data will be further analyzed to inform infection prevention policies. The performance of the proposed system is assessed through qualitative and quantitative experimental results. Finally, the capabilities of the current system and future research directions are briefly discussed. © 2020 IEEE.