X-rays imaging analysis for early diagnosis of the thoracic disorders using Capsule Neural Network with Transfer learning

Over the past few years, millions of people around the world have developed thoracic ailments. MRI, CT scan, reverse transcription, and other methods are among those used to detect thoracic disorders. These procedures demand medical knowledge and are exceedingly pricy and delicate. An alternate and more widely used method to diagnose diseases of the chest is X-ray imaging. The goal of this study was to increase detection precision in order to develop a computationally assisted diagnostic tool. Different diseases can be identified by combining radiological imaging with various artificial intelligence application approaches. In this study, transfer learning (TL) and capsule neural network techniques are used to propose a method for the automatic detection of various thoracic illnesses utilizing digitized chest X-ray pictures of suspected patients. Four public databases were combined to build a dataset for this purpose. Three pre trained convolutional neural networks (CNNs) were utilized in TL with augmentation as a preprocessing technique to train and evaluate the model. Pneumonia, COVID19, normal, and TB (Tb) were the four class classifiers used to train the network to categorize. © 2023 IEEE.

Analysis of Emerging Preprocessing Techniques Combined with Deep CNN for Lung Disease Detection

Data preprocessing is one of the pertinent steps while classifying images via CNN models. The efficiency of any model depends on the quality of the dataset it deals with. A clean dataset provides an efficient platform for a model to tackle classification and segmentation issues. Our paper focuses on three emerging data preprocessing techniques: Real ESRGAN, Swin IR, and GFPGAN over the lung disease dataset. We have used three models: Mobile net, Densenet201, and NasNet, to carry out classification tasks on Chest X-Ray images of six different types of lung disease: Bacterial Pneumonia, Viral pneumonia, Lung opacity, Covid, Tuberculosis, and Normal. Analysis of the aforementioned preprocessing techniques followed by classification via three CNN models (Mobile net, Densenet, and NasNet) are carried out on lung disease dataset, and their accuracy prediction, Training, and validation loss are extensively compared. © 2022 IEEE.

Application of Voting Based Approach on Deep Learning Algorithm for Lung Disease Classification

With the advent of the Deep learning era, an unprecedented change has come in the field of medical image analysis via CAD (Computer-Aided Diagnosis) [1]-[13] system. With feature extraction capability, Deep learning effectively performs the classification task and thus enhances the prediction of Medical images. CNN models like Dense net 121, Resnet-50, Alex net, Mobile Net, and Inceptionv3 have been used to identify lung disease, but no model was able to predict the particular class of illness with 100% accuracy. In this paper, we apply the voting method, which combines three models Dense net 121, Mobile Net, NasNet, and predictions are made on the basis of the majority of optimistic predictions. Our approach was found to improve the prediction of a particular class of lung disease like Bacterial Pneumonia, Covid, and Lung Opacity. Normal lungs are predicted with 100% accuracy with our approach. © 2022 IEEE.

Smoking May Increase the Risk of COVID-19 Infection: Evidence from In Silico Analysis

Introduction: SARS-CoV2, first reported in December 2019 in Wuhan as COVID-19 causing respiratory illness, rapidly evolved into a pandemic owing to its very high infectivity. There is insufficient evidence about if and how smoking affects the risk of COVID-19 infection, and the reports on whether smoking increases or reduces the risk of respiratory infections, are contradictory. Therefore, the current study was designed to determine the effects of nicotine consumption on the infectivity of COVID-19. Methods: We performed in silico computer simulation-based study. The structures of SARS-CoV2spike ectodomain, and its receptor ACE2, were obtained from PDB. The structure of nicotine and its metabolites NNK and NNAL were obtained from the PubChem chemical database. After optimization, they were interacted using AutoDock 4.2, to see the effect of nicotine, NNK, or NNAL presence on the docking of viral spike protein to its receptor ACE2. Results: ACE2 vs spike protein interaction results were used as a control (ZDOCK score 1498.484, with four hydrogen bonds). The NNK+ACE2 vs spike protein docking formed 10 hydrogen bonds with the highest ZDOCK score of 1515.564. NNAL+ ACE2 vs spike protein interaction formed eleven hydrogen bonds with the ZDOCK score of 1499.371. Nicotine+ACE2 vs spike protein docking showed the lowest ZDOCK score of 1496.302 and formed 8 hydrogen bonds. Whereas, NNK+spike vs ACE2 interaction had a ZDOCK score of 1498.490 and formed eight hydrogen bonds. NNAL+spike vs ACE2 docking formed eleven hydrogen bonds with a ZDOCK score of 1498.482. And Nicotine+spike vs ACE2 interaction showed a ZDOCK score of 1498.488 and formed 9 hydrogen bonds. Conclusions: The binding of nicotine to either spike of virus or its receptor ACE2 is not affecting the viral docking with the receptor. But binding of NNK, a metabolite of nicotine, is facilitating the viral docking with its receptor indicating that smoking may increase the risk of COVID-19 infection.

Thoracic Disease Detection using Deep Learning

Thoracic diseases are the most common radiological disorders worldwide especially in India. It is a life-threatening infectious disease affecting breathing organ like thorax and one or both lungs in human body commonly caused by bacteria. Physicians and radiologists are quiet using physical and visual graphical manners in order to diagnose the chest X-rays. Patient's diagnoses are entirely dependent on the consultant given by that chest expert. However, there might be emergency circumstances where radiology experts are too busy or may not be accessible. The timely and early diagnosis of thoracic diseases is very important. To resolve this situation, an algorithm that accept poster anterior (PA) chest X-rays images which classify whether the thorax is infected or not. If a thorax is infected, the proposed model will figure out which type of thoracic disorder is available on that PA view X-ray image. The proposed model can significantly improve the efficiency of doctors by early detection of the diseases using Computer aided diagnosis (CAD) wielding deep learning. Thus, an intelligent and automatic system is required to diagnose the chest radiograph to detect the various thorax related diseases. This research employ a web oriented identification system using deep learning based convolutional neural network algorithms for the detection, classification and early stage diagnosis of chest radiograph into healthy and thoracic disorders patients including COVID-19. The deep learning model is trained and tested on different radiographs which contain normal and numerous thorax disorders patient. Moreover, after developing the neural network model for the early diagnosis of the numerous thoracic disorders, a graphical user interface (web) based disease screening system also described for visualized the accurate diagnosis X-ray images in respective target disease classes. © 2021 IEEE.