Assessing an AI-based Smart Imagery Framing and Truthing (SIFT) system to assist radiologists annotating lung abnormalities on chest X-ray images for development of deep learning models

To assess a Smart Imagery Framing and Truthing (SIFT) system in automatically labeling and annotating chest X-ray (CXR) images with multiple diseases as an assist to radiologists on multi-disease CXRs. SIFT system was developed by integrating a convolutional neural network based-augmented MaskR-CNN and a multi-layer perceptron neural network. It is trained with images containing 307,415 ROIs representing 69 different abnormalities and 67,071 normal CXRs. SIFT automatically labels ROIs with a specific type of abnormality, annotates fine-grained boundary, gives confidence score, and recommends other possible types of abnormality. An independent set of 178 CXRs containing 272 ROIs depicting five different abnormalities including pulmonary tuberculosis, pulmonary nodule, pneumonia, COVID-19, and fibrogenesis was used to evaluate radiologists' performance based on three radiologists in a double-blinded study. The radiologist first manually annotated each ROI without SIFT. Two weeks later, the radiologist annotated the same ROIs with SIFT aid to generate final results. Evaluation of consistency, efficiency and accuracy for radiologists with and without SIFT was conducted. After using SIFT, radiologists accept 93% SIFT annotated area, and variation across annotated area reduce by 28.23%. Inter-observer variation improves by 25.27% on averaged IOU. The consensus true positive rate increases by 5.00% (p=0.16), and false positive rate decreases by 27.70% (p<0.001). The radiologist's time to annotate these cases decreases by 42.30%. Performance in labelling abnormalities statistically remains the same. Independent observer study showed that SIFT is a promising step toward improving the consistency and efficiency of annotation, which is important for improving clinical X-ray diagnostic and monitoring efficiency. © 2023 SPIE.

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.

Analyzing Chest X-Ray to Detect the Evidence of Lung Abnormality Due to Infectious Disease

Analyzing chest X-ray is the must especially when are required to deal of infectious disease outbreak, and COVID-19. The COVID-19 pandemic has had a large effect on almost every facet of life. As COVID-19 was a disease only discovered in recent history, there is comparatively little data on the disease, how it is detected, and how it is cured. Deep learning is a powerful tool that can be used to learn to classify information in ways that humans might not be able to. This allows computers to learn on relatively little data and provide exceptional results. This paper proposes a customized convolutional neural network (CNN) for the detection of COVID-19 from chest X-rays called basicConv. This network consists of five sets of convolution and pooling layers, a flatten layer, and two dense layers with a total of approximately 9 million parameters. This network achieves an accuracy of 95.8%, which is comparable to other high-performing image classification networks. This provides a promising launching point for future research and developing a network that achieves an accuracy higher than that of the leading classification networks. It also demonstrates the incredible power of convolution. This paper is an extension of a 2022 Honors Thesis (Henderson, Joshua Elliot, "Convolutional Neural Network for COVID-19 Detection in Chest X-Rays” (2022). Honors Thesis. 254. https://red.library.usd.edu/honors-thesis/254 ). © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Severe COVID-19 pneumonia leads to post-COVID-19 lung abnormalities on follow-up CT scans.

Purpose: To investigate the association of the maximal severity of pneumonia on CT scans obtained within 6-week of diagnosis with the subsequent development of post-COVID-19 lung abnormalities (Co-LA). Methods: COVID-19 patients diagnosed at our hospital between March 2020 and September 2021 were studied retrospectively. The patients were included if they had (1) at least one chest CT scan available within 6-week of diagnosis; and (2) at least one follow-up chest CT scan available ≥ 6 months after diagnosis, which were evaluated by two independent radiologists. Pneumonia Severity Categories were assigned on CT at diagnosis according to the CT patterns of pneumonia and extent as: 1) no pneumonia (Estimated Extent, 0%); 2) non-extensive pneumonia (GGO and OP, <40%); and 3) extensive pneumonia (extensive OP and DAD, >40%). Co-LA on follow-up CT scans, categorized using a 3-point Co-LA Score (0, No Co-LA; 1, Indeterminate Co-LA; and 2, Co-LA). Results: Out of 132 patients, 42 patients (32%) developed Co-LA on their follow-up CT scans 6-24 months post diagnosis. The severity of COVID-19 pneumonia was associated with Co-LA: In 47 patients with extensive pneumonia, 33 patients (70%) developed Co-LA, of whom 18 (55%) developed fibrotic Co-LA. In 52 with non-extensive pneumonia, 9 (17%) developed Co-LA: In 33 with no pneumonia, none (0%) developed Co-LA. Conclusions: Higher severity of pneumonia at diagnosis was associated with the increased risk of development of Co-LA after 6-24 months of SARS-CoV-2 infection.

Grayscale Image Statistical Attributes Effectively Distinguish the Severity of Lung Abnormalities in CT Scan Slices of COVID-19 Patients.

Grayscale statistical attributes analysed for 513 extract images taken from pulmonary computed tomography (CT) scan slices of 57 individuals (49 confirmed COVID-19 positive; eight confirmed COVID-19 negative) are able to accurately predict a visual score (VS from 0 to 4) used by a clinician to assess the severity of lung abnormalities in the patients. Some of these attributes can be used graphically to distinguish useful but overlapping distributions for the VS classes. Using machine and deep learning (ML/DL) algorithms with twelve grayscale image attributes as inputs enables the VS classes to be accurately distinguished. A convolutional neural network achieves this with better than 96% accuracy (only 18 images misclassified out of 513) on a supervised learning basis. Analysis of confusion matrices enables the VS prediction performance of ML/DL algorithms to be explored in detail. Those matrices demonstrate that the best performing ML/DL algorithms successfully distinguish between VS classes 0 and 1, which clinicians cannot readily do with the naked eye. Just five image grayscale attributes can also be used to generate an algorithmically defined scoring system (AS) that can also graphically distinguish the degree of pulmonary impacts in the dataset evaluated. The AS classification illustrated involves less overlap between its classes than the VS system and could be exploited as an automated expert system. The best-performing ML/DL models are able to predict the AS classes with better than 99% accuracy using twelve grayscale attributes as inputs. The decision tree and random forest algorithms accomplish that distinction with just one classification error in the 513 images tested.

Residual Lung Abnormalities Following COVID-19 Hospitalization: Interim Analysis of the UKILD Post-COVID Study.

RATIONALE: Shared symptoms and genetic architecture between COVID-19 and lung fibrosis suggests SARS-CoV-2 infection may lead to progressive lung damage. OBJECTIVES: The UKILD Post-COVID study interim analysis was planned to estimate the prevalence of residual lung abnormalities in people hospitalized with COVID-19 based on risk strata. METHODS: The Post-HOSPitalisation COVID Study (PHOSP-COVID) was used for capture of routine and research follow-up within 240 days from discharge. Thoracic CTs linked by PHOSP-COVID identifiers were scored for percentage of residual lung abnormalities (ground glass opacities and reticulations). Risk factors in linked CT were estimated with Bayesian binomial regression and risk strata were generated. Numbers within strata were used to estimate post-hospitalization prevalence using Bayesian binomial distributions. Sensitivity analysis was restricted to participants with protocol driven research follow-up. MEASUREMENTS AND MAIN RESULTS: The interim cohort comprised 3700 people. Of 209 subjects with linked CTs (median 119 days, interquartile range 83-155), 166 people (79.4%) had >10% involvement of residual lung abnormalities. Risk factors included abnormal chest X-ray (RR 1·21 95%CrI 1·05; 1·40), percent predicted DLco<80% (RR 1·25 95%CrI 1·00; 1·56) and severe admission requiring ventilation support (RR 1·27 95%CrI 1·07; 1·55). In the remaining 3491 people, moderate to very-high risk of residual lung abnormalities was classified in 7·8%, post-hospitalization prevalence was estimated at 8.5% (95%CrI 7.6%; 9.5%) rising to 11.7% (95%CrI 10.3%; 13.1%) in sensitivity analysis. CONCLUSIONS: Residual lung abnormalities were estimated in up to 11% of people discharged following COVID-19 related hospitalization. Health services should monitor at-risk individuals to elucidate long-term functional implications. This article is open access and distributed under the terms of the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/).

Lung Abnormalities in Severe Acute Respiratory Syndrome Coronavirus 2 Infection: An Analysis of Paired Computed Tomography Scans Before and After Infection.

Background: Studies on the pulmonary consequences of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection are impeded by limited access to pre-SARS-CoV-2 examinations. Methods: We invited Copenhagen General Population Study participants with a confirmed SARS-CoV-2 polymerase chain reaction (PCR) test during the first and second coronavirus disease 2019 waves in Denmark for a repeat chest computed tomography (CT) scan. Paired CT scans were independently assessed for interstitial and noninterstitial abnormalities by 2 trained radiologists. A semiquantitative CT score (ranging from 0 to 20) was used to quantify the extent of interstitial abnormalities. Results: Of 111 SARS-CoV-2-infected individuals, 102 (91.2%) experienced symptoms and 12 (11.2%) were hospitalized. Follow-up examination was performed at median of 5.4 (interquartile range, 4.1-7.8) months after a positive SARS-CoV-2 PCR test. Of 67 individuals with paired CT scans, ground glass opacities and reticulation were present in 31 (46.3%) individuals post-SARS-CoV-2 compared to 23 (34.1%) pre-SARS-CoV-2 (mean CT score, 3.0 vs 1.3; P = .011). Results were similar for nonhospitalized individuals. We did not detect development of bronchiectasis, emphysema, or nodules. Conclusions: SARS-CoV-2 infection in predominantly nonhospitalized individuals with mild disease was associated with a small increase in only interstitial lung abnormalities.

Chronic Obstructive Pulmonary Disease Combined with Interstitial Lung Disease.

Although chronic obstructive pulmonary disease (COPD) and interstitial lung disease (ILD) have distinct clinical features, both diseases may coexist in a patient because they share similar risk factors such as smoking, male sex, and old age. Patients with both emphysema in upper lung fields and diffuse ILD are diagnosed with combined pulmonary fibrosis and emphysema (CPFE), which causes substantial clinical deterioration. Patients with CPFE have higher mortality compared with patients who have COPD alone, but results have been inconclusive compared with patients who have idiopathic pulmonary fibrosis (IPF). Poor prognostic factors for CPFE include exacerbation, lung cancer, and pulmonary hypertension. The presence of interstitial lung abnormalities, which may be an early or mild form of ILD, is notable among patients with COPD, and is associated with poor prognosis. Various theories have been proposed regarding the pathophysiology of CPFE. Biomarker analyses have implied that this pathophysiology may be more closely associated with IPF development, rather than COPD or emphysema. Patients with CPFE should be advised to quit smoking and undergo routine lung function tests, and pulmonary rehabilitation may be helpful. Various pharmacologic agents and surgical approaches may be beneficial in patients with CPFE, but further studies are needed.

Classification network of Chest X-ray images based on residual network in the context of COVID-19

As COVID-19 spreads across the globe, more cases are being confirmed around the world, making it imperative that we take a better approach to fighting the outbreak. To stop the spread of the disease and better screen for cases, we need a more sensitive and efficient test that can classify images of lung abnormalities in patients. In this paper, residual network is used to classify the collected chest radiographs. Feature extraction and classification were carried out on the original chest X-ray images, which were divided into the following three categories: normal lung, bacterial pneumonia and virus pneumonia. This can quickly rule out normal and routine infections, screen out large numbers of cases, and reduce the burden on health care workers who need to further examine cases. At the same time, our results are also very good, with an accuracy of 94%, which has practical classification significance. © 2022 IEEE.

A Traditional Machine Learning Approach for COVID-19 Detection from CT Images

COVID-19 has severe effects on several body organs, especially the lung. These severe effects result in features in the COVID-19 patients' Computed Tomography (CT) images distinct from other viral pneumonia. Although the primary diagnosis of COVID-19 is not primarily screened by CT, machine learning-based diagnosis systems early detect the COVID-19 lung abnormalities. Feature extraction is crucial for the success of traditional machine learning algorithms. Traditional machine learning algorithms utilize hand-crafted features to identify and classify patterns in an image. This paper utilizes the Gabor filters as the primary feature extractor for automated COVID-19 classification from lung CT images. We use a publicly available COVID-19 data-set of chest CT images to validate the performance and accuracy of the proposed model. The Gabor filter and other feature extractors with Random Forest classifiers achieved over 81% classification accuracy, the sensitivity of 81%, Specificity of 82%, and F1 score of 81%. © 2021 IEEE.

Characteristic and quantifiable COVID-19 like abnormalities in CT- and PET/CT-1 imaged lungs of SARS-CoV-2 infected crab eating macaques (Macaca fascicularis)

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is causing an exponentially increasing number of coronavirus disease 19 (COVID-19) cases globally. Prioritization of medical countermeasures for evaluation in randomized clinical trials is critically hindered by the lack of COVID-19 animal models that enable accurate, quantifiable, and reproducible measurement of COVID-19 pulmonary disease free from observer bias. We first used serial computed tomography (CT) to demonstrate that bilateral intrabronchial instillation of SARS CoV-2 into crab-eating macaques (Macaca fascicularis) results in mild-to-moderate lung abnormalities qualitatively characteristic of subclinical or mild-to-moderateCOVID-19 (e.g., ground-glass opacities with or without reticulation, paving, or alveolar consolidation, peri-bronchial thickening, linear opacities) at typical locations (peripheral>central, posterior and dependent, bilateral, multi-lobar). We then used positron emission tomography (PET) analysis to demonstrate increased FDG uptake in the CT-defined lung abnormalities and regional lymph nodes. PET/CT imaging findings appeared in all macaques as early as 2 days post exposure, variably progressed, and subsequently resolved by 6-12 days post exposure. Finally, we applied operator-independent, semi-automatic quantification of the volume and radiodensity of CT abnormalities as a possible primary endpoint for immediate and objective efficacy testing of candidate medical countermeasures.

Impact of Interstitial Lung Diseases and COVID-19 on the body

Medical imaging techniques are often used in treatment and follow-ups for diagnosed diseases. Image scans provide quick acquisition of images and clear and precise information, along with a magnified view of a particular portion of the body. Chest images can demonstrate various lung disorders, such as, COVID-19, Interstitial Lung Diseases (ILD) and Chronic lung disease, Pneumonia, Bronchiectasis, Cystic Fibrosis, etc. However, subtle changes in the volume and character of lung abnormalities can be difficult to assess even by expert radiologists. This is where Artificial Intelligence (AI) comes in. AI can aid traditional medical imaging technology by offering computational prowess that process images with greater speed and precision. This work presents a solution that performs AI-empowered analysis of Chest image scans for diagnosis, tracking and prognosis of various lung diseases. © 2021 IEEE

Persistent interstitial lung abnormalities in post-COVID-19 patients: a case series.

A new concept of multisystem disease has emerged as a long-term condition following mild-severe COVID-19 infection. The main symptoms of this affection are breathlessness, chest pain, and fatigue. We present here the clinical case of four COVID-19 patients during hospitalization and 60 days after hospital discharge. Physiological impairment of all patients was assessed by spirometry, dyspnea score, arterial blood gas, and 6-minute walk test 60 days after hospital discharge, and computed tomographic scan 90 days after discharge. All patients had fatigue, which was not related to hypoxemia or impaired spirometry values, and interstitial lung alterations, which occurred in both mechanically ventilated and non-mechanically ventilated patients. In conclusion, identifying the prevalence and patterns of permanent lung damage is paramount in preventing and treating COVID-19-induced fibrotic lung disease. Additionally, and based on our preliminary results, it will be also relevant to establish long-term outpatient programs for these individuals.