The role of chest radiography in confirming covid-19 pneumonia

Abstract

Contributorship and guarantor: MJ conceptualised this article, undertook the literature search and is guarantor. JC used her knowledge of chest radiology to inform the content of the article and sourced the images. JP used his knowledge of assessment and management of covid-19 patients to inform the article. MJ, JC, and JP drafted and checked the final article. Patient consent:We considered seeking individual consent to use radiological images. However, as patients with covid-19 are ill and an infectious risk, obtaining consent was not possible. We approached the hospital ethics committee chair, Trust R&D, and Data Protection lead for permission to use anonymised radiological images without specific consent. They agreed this was acceptable.

DeepCOVID-XR: An Artificial Intelligence Algorithm to Detect COVID-19 on Chest Radiographs Trained and Tested on a Large U.S. Clinical Data Set

“…Although there have been promising results using AI for detection of COVID-19 on CT imaging (5-7), the use of CT for this purpose is limited by concerns regarding cost/time, radiation exposure, and decontamination procedures for equipment (8). In contrast, chest radiographs can be acquired rapidly and portably, involve trivial radiation exposure, and pose a lower risk for viral spread (9,10). But previously reported AI algorithms for identification of COVID-19 using chest radiographs have been limited by small datasets and the use of publicly available images of variable quality and questionable validity (11)(12)(13)(14)(15)(16).…”

“…This serves as an important sanity check to reinforce confidence in algorithm predictions.The explainable errors in algorithm predictions likely represent limitations of chest imaging in the radiologic diagnosis of COVID-19 rather than the algorithm itself. Prior clinical studies showed COVID-19 pneumonia produces characteristic features on chest imaging, but up to 56% of symptomatic patients can demonstrate normal chest imaging, especially early in their disease course(9,27,(29)(30)(31). Imaging is therefore inappropriate to "rule out" disease.…”

Transfer Learning for COVID-19 Pneumonia Detection and Classification in Chest X-ray Images

“…phenomenon known as consolidation ( Figure 1b) [9]. Hence, utilizing chest radiography, or combining it with laboratory and clinical assessment, can be an efficient tool for the early and accurate diagnosis of COVID-19 pneumonia.…”

Development and Validation of the Patient History COVID-19 (PH-Covid19) Scoring System: A Multivariable Prediction Model of Death in Mexican Patients with COVID-19

“…This last study, however, included only patients admitted to ICUs who had a high prevalence of comorbidities, which are independent risk factors for ICU admission[3].COVID-19 patients included in a meta-analysis[23]. Chest X-ray, on the other hand, may be normal in up to 63% of patients with early COVID-19 pneumonia[24].Nonetheless, the low proportion of pneumonia in non-survivors suggests non-optimal diagnosis of pneumonia may be occurring in Mexico. The lack of an operational definition may have contributed since clinicians could have defined pneumonia differently based on clinical and/or radiographical findings.…”