Lung Opacities Research Articles

Detecting Covid-19 and Other Lung Diseases with Deep Learning

Since December 2019, Covid-19, also known as severe acute respiratory syndrome (SARSCOV-2), has infected about 531 million individuals worldwide causing devastating consequences. Scientists, health professionals, and radiologists have all invested huge amounts of time, effort and resources in developing faster and accurate methods for detecting the virus as it spreads all over the world. The current gold standards for detecting covid-19 such as RT-PCR, POC, Droplet-based digital PCR (dPCR) and Immunoassays are not only expensive for low-income countries but also require substantial human expert knowledge which in turn makes them time-consuming and susceptible to manipulation. The study presented in this paper aims to address these issues by proposing a model to automatically detect COVID-19 in chest X-rays using Deep Learning techniques. Specifically, the model uses convolutional neural networks (CNNs), a deep learning technique, which has been shown to produced very good results when applied to medical image diagnosis problems. Our proposed model will therefore not only provide a cheaper, faster and accurate method of detecting COVID-19 but will also provide an easy to use web and mobile platform for non-experts. While current detection techniques such the RT-PCR are still considered as the most effective, our CNN-based model was also able to produce good results when applied to the COVID-19 radiography dataset. For instance, the model was able to detect COVID-19 from X-ray images with an accuracy rate of 90 percent. Furthermore, the model was also able to automatically detect other respiratory diseases such as viral pneumonia and lung opacity with an accuracy rate of over 90 percent. Given the potential of the automated detection of COVID-19 established in this study, future work will extend this work by incorporating automated search techniques such as genetic algorithms in the fine tuning of the model’s parameters in order to obtain an even higher accuracy rate.

A Hybrid Model Based on Deep Features and Ensemble Learning for the Diagnosis of COVID-19: DeepFeat-E

COVID-19, which has been declared a pandemic disease, has affected the lives of millions of people and caused a major epidemic. Despite the development of vaccines and vaccination to prevent the transmission of the disease, COVID-19 case rates fluctuate worldwide. Therefore, rapid and reliable diagnosis of COVID-19 disease is of critical importance. For this purpose, a hybrid model based on transfer learning methods and ensemble classifiers is proposed in this study. In this hybrid approach, called DeepFeat-E, the diagnosis process is performed by using deep features obtained from transfer learning models and ensemble classifiers consisting of classical machine learning methods. To test the proposed approach, a dataset of 21,165 X-ray images including 10,192 Normal, 6012 Lung Opacity, 1345 Viral Pneumonia and 3616 COVID-19 were used. With the proposed approach, the highest accuracy was achieved with the deep features of the DenseNet201 transfer learning model and the Stacking ensemble learning method. Accordingly, the test accuracy was 90.17%, 94.99% and 94.93% for four, three and two class applications, respectively. According to the results obtained in this study, it is seen that the proposed hybrid system can be used quickly and reliably in the diagnosis of COVID-19 and lower respiratory tract infections.

A Neutrosophic Set Approach on Chest X-rays for Automatic Lung Infection Detection

COVID cases and its variants is noted enormously in the past three years. In many medical cases, lung infections such as viral pneumonia, bacterial pneumonia have been initially interpreted as COVID-19. Hence, the proposed work is concentrating on differentiating these lung infection types. This work focuses on using neutrosophic approach of classifying into True (T), False (F) and Indeterminacy (I) set membership to reduce the fuzziness and retain more significant information for feature extraction of the opacity to differentiate the types of lung infections. Initially, the images are preprocessed by alpha-mean and beta-enhancement operation to reduce the indeterminacy and enhancing the image components as the range of lung opacity levels to determine the types. Then, these neutrosophic set enhanced images are fed to various deep learning models like ResNet-50, VGG-16 and XGBoost for classification. Experiments are conducted on ActualMed COVID-19 Chest X-ray and COVID-19 radiography dataset and a comparative analysis on several domain set of images such as the original image, neutrosophic domain (T, I, F) and enhanced neutrosophic domain (alpha, beta) are trained and tested through transfer learning by tuning the various validation parameters. On experimental analysis, an enhanced neutrosophic image achieves a better accuracy of 97.33% among the other domain sets.

MTMC-AUR2CNet: Multi-textural multi-class attention recurrent residual convolutional neural network for COVID-19 classification using chest X-ray images

Coronavirus disease (COVID-19) has infected over 603 million confirmed cases as of September 2022, and its rapid spread has raised concerns worldwide. More than 6.4 million fatalities in confirmed patients have been reported. According to reports, the COVID-19 virus causes lung damage and rapidly mutates before the patient receives any diagnosis-specific medicine. Daily increasing COVID-19 cases and the limited number of diagnosis tool kits encourage the use of deep learning (DL) models to assist health care practitioners using chest X-ray (CXR) images. The CXR is a low radiation radiography tool available in hospitals to diagnose COVID-19 and combat this spread. We propose a Multi-Textural Multi-Class (MTMC) UNet-based Recurrent Residual Convolutional Neural Network (MTMC-UR2CNet) and MTMC-UR2CNet with attention mechanism (MTMC-AUR2CNet) for multi-class lung lobe segmentation of CXR images. The lung lobe segmentation output of MTMC-UR2CNet and MTMC-AUR2CNet are mapped individually with their input CXRs to generate the region of interest (ROI). The multi-textural features are extracted from the ROI of each proposed MTMC network. The extracted multi-textural features from ROI are fused and are trained to the Whale optimization algorithm (WOA) based DeepCNN classifier on classifying the CXR images into normal (healthy), COVID-19, viral pneumonia, and lung opacity. The experimental result shows that the MTMC-AUR2CNet has superior performance in multi-class lung lobe segmentation of CXR images with an accuracy of 99.47%, followed by MTMC-UR2CNet with an accuracy of 98.39%. Also, MTMC-AUR2CNet improves the multi-textural multi-class classification accuracy of the WOA-based DeepCNN classifier to 97.60% compared to MTMC-UR2CNet.

Severe Pneumocystis Pneumonia Associated with Ulcerative Colitis

A 56-year-old woman with a 9-month history of ulcerative colitis (UC) who had been treated with prednisolone (PSL) 40 mg/day and azathioprinefor 2 months was referred to our hospital because of fever and dyspnea. Computed tomography (CT) of the chest revealed diffuse ground-glass opacities in both lungs and consolidation in both lower lobes. PCR of the patient's bronchoalveolar lavage fluid was positive for pneumocystis jirovecii, and she was diagnosed with pneumocystis pneumonia (PCP). The patient developed refractory respiratory failure that required invasive mechanical ventilation. After sulfamethoxazole-trimethoprim administration and methyl prednisolone pulse therapy, her clinical condition improved dramatically. Few reports are available on severe PCP associated with UC. It is important to keep in mind the possibility of complication by PCP in UC patients.

Multiple Myeloma with Lung Plasmacytoma

Multiple myeloma (MM) is a clonal proliferation of malignant plasma cells mainly affecting the bone marrow. Most common sites of extramedullary dissemination reported in the literature are skin, liver, kidneys and central nervous system. The presentation of MM with lung plasmacytoma is found to be very uncommon. We present a case of multiple myeloma associated with lung plasmacytoma. A 53-years-old man with history of shortness of breath and lung opacity visualized on a chest X-ray. Careful integration of clinical manifestations with radiological and biological data led to the diagnosis of multiple myeloma with pulmonary plasmacytoma.

The association of clinically relevant variables with chest radiograph lung disease burden quantified in real-time by radiologists upon initial presentation in individuals hospitalized with COVID-19

ObjectivesWe aimed to correlate lung disease burden on presentation chest radiographs (CXR), quantified at the time of study interpretation, with clinical presentation in patients hospitalized with coronavirus disease 2019 (COVID-19). Material and methodsThis retrospective cross-sectional study included 5833 consecutive adult patients, aged 18 and older, hospitalized with a diagnosis of COVID-19 with a CXR quantified in real-time while hospitalized in 1 of 12 acute care hospitals across a multihospital integrated healthcare network between March 24, 2020, and May 22, 2020. Lung disease burden was quantified in real-time by 118 radiologists on 5833 CXR at the time of exam interpretation with each lung annotated by the degree of lung opacity as clear (0%), mild (1–33%), moderate (34–66%), or severe (67–100%). CXR findings were classified as (1) clear versus disease, (2) unilateral versus bilateral, (3) symmetric versus asymmetric, or (4) not severe versus severe. Lung disease burden was characterized on initial presentation by patient demographics, co-morbidities, vital signs, and lab results with chi-square used for univariate analysis and logistic regression for multivariable analysis. ResultsPatients with severe lung disease were more likely to have oxygen impairment, an elevated respiratory rate, low albumin, high lactate dehydrogenase, and high ferritin compared to non-severe lung disease. A lack of opacities in COVID-19 was associated with a low estimated glomerular filtration rate, hypernatremia, and hypoglycemia. ConclusionsCOVID-19 lung disease burden quantified in real-time on presentation CXR was characterized by demographics, comorbidities, emergency severity index, Charlson Comorbidity Index, vital signs, and lab results on 5833 patients. This novel approach to real-time quantified chest radiograph lung disease burden by radiologists needs further research to understand how this information can be incorporated to improve clinical care for pulmonary-related diseases.. An absence of opacities in COVID-19 may be associated with poor oral intake and a prerenal state as evidenced by the association of clear CXRs with a low eGFR, hypernatremia, and hypoglycemia.

Enhancing deep learning techniques for the diagnosis of the novel coronavirus (COVID-19) using X-ray images

Deep learning techniques combined with radiological imaging provide precision in the diagnosis of diseases that can be utilised for the classification and diagnosis of several diseases in the medical sector. Several research studies have focused on binary classification of COVID-19, and there is limited research focusing on multiclass classification of COVID-19. The purpose of this study is to develop a model that can enhance the multiclass classification of COVID-19 using raw chest X-ray images. The study involved using convolutional neural networks as the classifier. Five pre-trained deep learning models including VGG16, MobileNet, EfficientNetB0, NasNetMobile and ResNet50V2 are used to distinguish between COVID-19 infection and other lung diseases. Data Augmentation and Normalization techniques have been used to improve the models’ performance and avoid training problems. The study findings revealed that it is possible to distinguish between COVID-19 infection and other lung diseases using pre-trained deep learning models. The proposed technique successfully classifies five classes (normal, COVID-19, lung opacity, viral pneumonia, and bacterial pneumonia). It is found that NasNetMobile model outperformed the rest of the models and achieved the highest results. It achieved an overall accuracy, sensitivity, specificity and precision of 91%, 91%, 97.7% and 91%, respectively. The VGG16 model produced better results in detecting COVID-19 infection, resulting in an accuracy of 95.8%. The suggested technique is more accurate in comparison to the other newly developed techniques presented in the literature. This provides healthcare staff with a powerful tool for the diagnosis of COVID-19 based on deep learning.

EGPA Phenotyping: Not Only ANCA, but Also Eosinophils

Eosinophilic Granulomatosis with Polyangiitis (EGPA) is a small-vessel necrotizing vasculitis. The anti-neutrophil cytoplasmic antibodies' (ANCA) role in defining clinical EGPA phenotypes is well established. Although the role of eosinophils in disease pathogenesis has been clearly demonstrated, the value of blood eosinophil count (BEC) as a biomarker of disease phenotypes is currently uncertain. We retrospectively analyzed EGPA patients referred to our Immunology Clinic. Demographic, laboratory and clinical features were retrieved from clinical records, and a Logistic Regression was fitted to evaluate the predictive power of all baseline clinical and laboratory features to define EGPA phenotypes. 168 patients were recruited. BEC ≤ 1500 cells/mL was predictive of a clinical involvement characterized by asthma, chronic rhinosinusitis with nasal polyps (CRSwNP) and lung opacities (OR 0.18, 95% CI 0.07-0.43; respiratory-limited phenotype); BEC > 3500/mL was predictive of extrapulmonary organ involvement (OR 3.5, 95% CI 1.7-7.1; systemic phenotype). BEC was also predictive of peripheral nervous system (PNS) involvement, with a positive trend with increasing BEC (<1500/mL: OR 0.17, 95%CI, 0.06-0.47; >3500/mL: OR 2.8, 95% CI, 1.5-5.28). ANCA positivity was also predictive of extrapulmonary involvement (OR 4.7, 95% CI 1.9-11.99). according to BEC and irrespective of the ANCA status, two EGPA phenotypes could be identified, named systemic and respiratory-limited phenotypes, with different organ involvement and possibly different prognoses.

COVID-19: Clinical Course and Outcome of hospitalized Hemodialysis Patients: Single-center Experience.

There is a paucity of data regarding the consequences of coronavirus disease 2019 (COVID-19) infection in patients with maintenance hemodialysis (MHD). Our objective was to identify the clinical manifestations and prognostic factors and to assess the impact of treatment schemes on the outcomeMaterials and methods: Here we present retrospectively collected data from medical records of patients on MHD hospitalized with COVID-19 infection from 1st June to 30th November 2020Result: Around 69 patients were admitted with a median age of 51 years. About 81% had hypertension, 41% had diabetes, and 24% had body mass index (BMI) ≥ 23 kg/m2 . Of all who died, 73.33% had dialysis vintage of &lt;12 months (p = 0.06). Common presenting symptoms were fatigue (67%), fever (58%), cough (42%), and dyspnea (35%). Milder, severe, and critical disease was found in 35, 45, and 20% of patients, respectively. About 54 patients were living 4 weeks after discharge. Around 15 patients died, that includes all who received invasive ventilatory support. Nonsurvivors were older and had lower oxygen saturation on admission, lower hemoglobin (Hb), and worst lactate dehydrogenase (LDH), interleukin (IL)-6, and D-dimer values than survivors, which were statistically significant. Use of remdesivir and anticoagulant improves chances of survival (p-value 0.035 and 0.034, respectively) Conclusion: About one-third of patients had mild disease. Those with critical disease displayed high mortality. Older age, male gender, short dialysis vintage, lower oxygen saturation on admission, anemia, leucocytosis, higher inflammatory markers [except C-reactive protein (CRP)], bilateral lung opacity, and requirement of the mechanical ventilator are poor prognostic factors. CRP, ferritin, and lymphopenia are not good prognostic markers unlike in the general population. These findings need to be verified in larger cohorts.

Schematic approach to the diagnosis of multifocal lung opacities in the emergency department

Radiologists in the emergency department must be prepared to deal with any type of disease in any organ at any time. Many entities involving the chest can result in patients’ presenting at the emergency department. This chapter deals with entities that manifest with multifocal lung opacities and that can be mistaken for pneumonia. To facilitate their identification, this chapter approaches these entities by considering their most characteristic distribution on chest X-rays, the main diagnostic modality used for thoracic problems in the emergency department. Our schematic approach includes the key findings in patients’ personal histories, clinical examination, laboratory tests, and imaging studies that can be available during the initial workup.

A New Respiratory Diseases Detection Model in Chest X-Ray Images Using CNN

Convolutional Neural Network (CNN)-based deep learning techniques have recently demonstrated increased potential and effectiveness in image recognition applications, such as those involving medical images. Deep-learning models can recognize targets with performance comparable to radiologists when used with CXR. The primary goal of this research is to examine a deep learning technique used on the radiography dataset to detect COVID-19 in X-ray medical images. The proposed system consists of several stages, from pre-processing, passing through the feature reduction using more than one technique, to the classification stage based on a proposed model. The test was applied to the COVID-19 Radiography dataset of normal and three lung infections (COVID-19, Viral Pneumonia, and Lung Opacity). The proposed CNN model has shown its ability to classify COVID, normal, and other lung infections with perfect accuracy of 99.94%. Consequently, the AI-based early-stage detection algorithms will be enhanced, increasing the accuracy of the X-ray-based modality for the screening of various lung diseases.

Drug-induced interstitial lung disease due to over-the-counter cold medicine taken daily for 25 years: a case report

IntroductionWe report a rare case of drug-induced interstitial lung disease due to over-the-counter cold medicine taken daily for 25 years to clear the patient’s head.Case presentationA 77-year-old Japanese man presented to our hospital with a worsening cough that started 5 years ago. Chest radiographs and computed tomography images showed bilateral opacities, and transbronchial lung biopsy specimens showed an organizing pneumonia pattern. He reported taking the same over-the-counter cold medicine daily for the past 25 years to clear his head. We suspected that the cold medicine caused the lung opacities and asked him to stop taking them. His cough, general fatigue, and chest infiltrate gradually diminished. However, 6 months later, he resumed the same treatment because of a cold. The following month, he presented with severe worsening cough and chest radiographical findings. We diagnosed drug-induced interstitial lung disease. He improved by stopping the cold medicine again and taking prednisolone.ConclusionsOver-the-counter cold medicines are easily accessible at the drugstore. In cases of diffuse lung disease, we should consider drug-induced interstitial lung disease due to over-the-counter cold medicine, which patients have been taking not only for weeks or months but also years.

Pneumonia X‐ray detection with anchor‐free detection framework and data augmentation

AbstractPneumonia has become one of the main causes of human death. However, it is a tall order to efficiently and accurately diagnose pneumonia for clinicians. Therefore, A novel method based on anchor‐free detection framework is proposed to automatically locate lung opacities on chest radiographs in this study. We conducted extensive sets of experiments on the dataset of the Radiological Society of North America (RSNA) pneumonia detection challenge from the Kaggle competition. The results show superior performances for our method compared with previous studies. The best method achieved 52.9% in average precision (AP) and 97.5% in average recall (AR). For better interpretability of the results, visualization techniques are applied to provide visual explanations for our method. The visualization of these randomly selected samples shows that the method has excellent performance for lung opacity detection. Our method achieves better discriminative results and is suitable for the pneumonia diagnosis.

Abordaje esquemático del diagnóstico de las opacidades pulmonares multifocales en la urgencia

Radiologists in the emergency department must be prepared to deal with any type of disease in any organ at any time. Many entities involving the chest can result in patients’ presenting at the emergency department. This chapter deals with entities that manifest with multifocal lung opacities and that can be mistaken for pneumonia. To facilitate their identification, this chapter approaches these entities by considering their most characteristic distribution on chest X-rays, the main diagnostic modality used for thoracic problems in the emergency department. Our schematic approach includes the key findings in patients’ personal histories, clinical examination, laboratory tests, and imaging studies that can be available during the initial workup.

Effects of Pirfenidone on Idiopathic Pulmonary Fibrosis Progression and Safety: Results of Multicenter Prospective Observational Study

The aim of this study was to determine the effectiveness of pirfenidone in patients with idiopathic pulmonary fibrosis (IPF) seen in clinical practice. Fifty-five adults with IPF were enrolled in this multicenter, open-label, non-randomized, non-controlled, interventional clinical study. All patients received pirfenidone 2403 mg/day (three 267 mg capsules three times daily) for 26 weeks. After 26 weeks of treatment, the mean change in absolute forced vital capacity (FVC) was 128.8 mL (95% confidence interval [CI] -26.8, 284.4) and the mean change in relative predicted FVC was -0.10% (95% CI -3.18, 2.99). Stable disease (defined as improvement of ≥0% or a decline of <10% to 0% of the corresponding FVC value) was observed in most patients (relative FVC, 90.9%; absolute FVC, 83.6%). There was no statistically significant change in the mean high-resolution computed tomography fibrosis score or lung opacity score at week 26 compared with baseline. Treatment-emergent adverse events were reported in 80% of patients during the treatment period; most of them were mild or moderate in severity. No serious pirfenidone-related adverse events were observed during the study period. Pirfenidone was generally safe and effective for controlling functional decline and stabilizing disease in patients with IPF encountered in clinical practice in Russia.

Description of normal pulmonary radiographic findings in 55 apparently healthy juvenile Kemp's ridley sea turtles (Lepidochelys kempii).

A total of 55 digital radiographic studies from 53 individual juvenile Kemp's ridley sea turtles (Lepidochelys kempii) were retrospectively used to determine the normal radiographic anatomy of the lower respiratory tract in sea turtles that had been stranded due to hook-and-line injury and were otherwise clinically healthy. There were three or four projections available for each study: dorsoventral (DV), rostrocaudal (RoCd), and left and/or right lateral. The DV and RoCd were most conducive for assessing global lung volume and symmetry of lung volume. The DV and lateral views were most helpful for evaluating the main bronchus and its branching channels and for assessing lung margination. The RoCd view was most useful for assessing the symmetry of the lung opacity. The lateral views were most helpful for assessing the ventral margin of each lung lobe. On the lateral view, the main bronchus lay ventrally and coursed horizontally through the lung from cranial to caudal. On the DV view, the bronchus lay medially and was observed to be curvilinear coursing caudomedially. On the RoCd view, the main bronchus was located ventromedially. The RoCd view demonstrated the channels and niches end-on resulting in a reticulated or honeycomb appearance. The channels were seen as uniform striations coursing perpendicular to the main bronchus on the lateral views (vertical striations coursing dorsal to ventral) and DV views (horizontal striations coursing medially to laterally).

In Search of an Efficient and Reliable Deep Learning Model for Identification of COVID-19 Infection from Chest X-ray Images.

The virus responsible for COVID-19 is mutating day by day with more infectious characteristics. With the limited healthcare resources and overburdened medical practitioners, it is almost impossible to contain this virus. The automatic identification of this viral infection from chest X-ray (CXR) images is now more demanding as it is a cheaper and less time-consuming diagnosis option. To that cause, we have applied deep learning (DL) approaches for four-class classification of CXR images comprising COVID-19, normal, lung opacity, and viral pneumonia. At first, we extracted features of CXR images by applying a local binary pattern (LBP) and pre-trained convolutional neural network (CNN). Afterwards, we utilized a pattern recognition network (PRN), support vector machine (SVM), decision tree (DT), random forest (RF), and k-nearest neighbors (KNN) classifiers on the extracted features to classify aforementioned four-class CXR images. The performances of the proposed methods have been analyzed rigorously in terms of classification performance and classification speed. Among different methods applied to the four-class test images, the best method achieved classification performances with 97.41% accuracy, 94.94% precision, 94.81% recall, 98.27% specificity, and 94.86% F1 score. The results indicate that the proposed method can offer an efficient and reliable framework for COVID-19 detection from CXR images, which could be immensely conducive to the effective diagnosis of COVID-19-infected patients.

Interpretable Differential Diagnosis of Non-COVID Viral Pneumonia, Lung Opacity and COVID-19 Using Tuned Transfer Learning and Explainable AI.

The coronavirus epidemic has spread to virtually every country on the globe, inflicting enormous health, financial, and emotional devastation, as well as the collapse of healthcare systems in some countries. Any automated COVID detection system that allows for fast detection of the COVID-19 infection might be highly beneficial to the healthcare service and people around the world. Molecular or antigen testing along with radiology X-ray imaging is now utilized in clinics to diagnose COVID-19. Nonetheless, due to a spike in coronavirus and hospital doctors' overwhelming workload, developing an AI-based auto-COVID detection system with high accuracy has become imperative. On X-ray images, the diagnosis of COVID-19, non-COVID-19 non-COVID viral pneumonia, and other lung opacity can be challenging. This research utilized artificial intelligence (AI) to deliver high-accuracy automated COVID-19 detection from normal chest X-ray images. Further, this study extended to differentiate COVID-19 from normal, lung opacity and non-COVID viral pneumonia images. We have employed three distinct pre-trained models that are Xception, VGG19, and ResNet50 on a benchmark dataset of 21,165 X-ray images. Initially, we formulated the COVID-19 detection problem as a binary classification problem to classify COVID-19 from normal X-ray images and gained 97.5%, 97.5%, and 93.3% accuracy for Xception, VGG19, and ResNet50 respectively. Later we focused on developing an efficient model for multi-class classification and gained an accuracy of 75% for ResNet50, 92% for VGG19, and finally 93% for Xception. Although Xception and VGG19's performances were identical, Xception proved to be more efficient with its higher precision, recall, and f-1 scores. Finally, we have employed Explainable AI on each of our utilized model which adds interpretability to our study. Furthermore, we have conducted a comprehensive comparison of the model's explanations and the study revealed that Xception is more precise in indicating the actual features that are responsible for a model's predictions.This addition of explainable AI will benefit the medical professionals greatly as they will get to visualize how a model makes its prediction and won't have to trust our developed machine-learning models blindly.

Is the generalizability of a developed artificial intelligence algorithm for COVID-19 on chest CT sufficient for clinical use? Results from the International Consortium for COVID-19 Imaging AI (ICOVAI)

ObjectivesOnly few published artificial intelligence (AI) studies for COVID-19 imaging have been externally validated. Assessing the generalizability of developed models is essential, especially when considering clinical implementation. We report the development of the International Consortium for COVID-19 Imaging AI (ICOVAI) model and perform independent external validation.MethodsThe ICOVAI model was developed using multicenter data (n = 1286 CT scans) to quantify disease extent and assess COVID-19 likelihood using the COVID-19 Reporting and Data System (CO-RADS). A ResUNet model was modified to automatically delineate lung contours and infectious lung opacities on CT scans, after which a random forest predicted the CO-RADS score. After internal testing, the model was externally validated on a multicenter dataset (n = 400) by independent researchers. CO-RADS classification performance was calculated using linearly weighted Cohen’s kappa and segmentation performance using Dice Similarity Coefficient (DSC).ResultsRegarding internal versus external testing, segmentation performance of lung contours was equally excellent (DSC = 0.97 vs. DSC = 0.97, p = 0.97). Lung opacities segmentation performance was adequate internally (DSC = 0.76), but significantly worse on external validation (DSC = 0.59, p < 0.0001). For CO-RADS classification, agreement with radiologists on the internal set was substantial (kappa = 0.78), but significantly lower on the external set (kappa = 0.62, p < 0.0001).ConclusionIn this multicenter study, a model developed for CO-RADS score prediction and quantification of COVID-19 disease extent was found to have a significant reduction in performance on independent external validation versus internal testing. The limited reproducibility of the model restricted its potential for clinical use. The study demonstrates the importance of independent external validation of AI models.Key Points• The ICOVAI model for prediction of CO-RADS and quantification of disease extent on chest CT of COVID-19 patients was developed using a large sample of multicenter data.• There was substantial performance on internal testing; however, performance was significantly reduced on external validation, performed by independent researchers. The limited generalizability of the model restricts its potential for clinical use.• Results of AI models for COVID-19 imaging on internal tests may not generalize well to external data, demonstrating the importance of independent external validation.