Omnia Health is part of the Informa Markets Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Thoracic imaging in COVID-19

Article-Thoracic imaging in COVID-19

The following article is available in full, including figures and data, on Cleveland Clinic Journal of Medicine August 2020 as part of its COVID-19 Curbside Consults.

The typical findings of COVID-19 on chest radiography and computed tomography (CT) include bilateral, multifocal parenchymal opacities (ground-glass opacities with or without consolidation, and “crazy paving”). In most cases, the opacities are predominantly in the peripheral and lower lung zones, and several have rounded morphology.

However, these imaging findings are not pathognomonic for COVID-19 pneumonia and can be seen in other viral and bacterial infections, as well as with noninfectious causes such as drug toxicity and connective tissue disease. Most radiology professional organizations and societies recommend against routine screening CT to diagnose or exclude COVID-19.

The lungs are the most common site of infection in COVID-19, and progression to respiratory failure is the most common cause of death. In this brief summary we describe the role of thoracic imaging in COVID-19.

Chest radiography in COVID-19

Chest radiography is considered an appropriate initial imaging diagnostic test for most patients with lower respiratory tract infection, including those suspected of having COVID-19. The radiographic abnormalities in COVID-19 mirror those on computed tomography (CT), demonstrating bilateral, peripheral, and mid-lower-lung-zone-predominant consolidation.

However, in patients who have a high pretest probability of COVID-19, atypical findings such as diffuse interstitial changes or unilateral focal consolidation should not dissuade the radiologist from suspecting an infection, including COVID-19, as a possible diagnosis.

In patients with progressive disease, the density and extent of parenchymal changes typically increase over time. The severity of chest radiographic findings peaks 10 to 12 days after the onset of symptoms.

Unfortunately, most bacterial pneumonias also present as consolidation, and it is difficult to distinguish them from viral infections on chest radiography. The subtleties of rounded morphology and “crazy paving” associated with COVID-19 can only be appreciated on CT and not on plain chest radiographs. Cavitation within an airspace consolidation likely suggests a superadded infection.

Moreover, chest radiography has a high false-negative rate, especially in the early stage of infection, and should not be used as a screening tool to rule out COVID-19. In fact, baseline radiography has a lower sensitivity (69%) than initial reverse transcriptase polymerase chain reaction (RT-PCR) testing (91%).

Radiography 'through class' to avoid spreading the virus

During the pandemic, our hospital (as well as many others in the United States) has employed a method of obtaining portable radiographs in cases of confirmed or suspected COVID-19 through the glass wall of the patient’s room in the intensive care unit and in the emergency department.

With some minor technical modifications, the chest radiographs taken “through glass” are comparable to those obtained by the standard method. This technique has the potential to reduce the consumption of personal protective equipment by radiology technicians and to reduce the risk of machine contamination.

Read the full article.

COVID-19 India Update

How Qure.ai is fighting COVID-19

Article-How Qure.ai is fighting COVID-19

Amongst the health tech companies actively involved in the COVID-19 fight in recent months, Qure.ai, headquartered in Mumbai, is undertaking a truly worldwide effort, from Malawi to Oman, thanks largely to its strong foundations developing AI solutions for radiology.  

Qure.ai’s COVID-19 “story” started four years ago, when the startup was established. A risk score had then been developed in its automated screening for tuberculosis that took more than eight findings into account, looking at factors that included opacity and fibrosis, with the clinician making the final diagnosis.

When the coronavirus disease emerged, they realised that the same technology applied to tuberculosis could be extended to COVID-19 to assist clinical teams for triaging or progression monitoring of COVID-19 patients.

Qure.ai later offered an additional solution for remote patient management and disease monitoring purposes. Both solutions are now available on a monthly subscription basis. 

Technology aside, Qure.ai’s experience in optimising and deploying workflows on the ground, combined with live sites in 20 countries using its solutions, meant that it was well positioned to offer COVID-19 support when the disease broke out.

Chest X ray screening repurposed

At the end of February 2020 and the beginning of March in the early months of the pandemic, there was limited access to PCR test kits.

Qure.ai’s hypothesis was that it was difficult for clinical teams to determine whether patients should go home upon being tested or admitted immediately to hospital.

The startup repurposed and offered its x-ray AI solution qXR to existing customers, expanding their ability to detect COVID-19 as part of their existing TB screening efforts.

The solution was able to identify findings indicative of COVID-19 in a chest X-Ray, for instance an opacity, and also those that are counter-indicative such as fibrosis, in which case the coronavirus was not likely. Between the two buckets, there were 8-10 findings.

qXR was also able to pinpoint the location of the findings within the lung: most cases are bilateral and settling in the lower parts of the organ.

Based on these factors, a COVID-19 score was offered.

qXr-chest-xray-opacity.png

The pivot to disease monitoring

The situation changed when cases of COVID-19 and access to test kits both grew, and so Qure.ai quickly responded. Customers asked for help with monitoring progression of the disease, rather than its detection, in the absence of any definitive treatment or drug therapy.

Quantifying the extent of the infection is a time-consuming step, and with the results being often inconsistent Qure.ai customers demanded a solution that was data-driven and replicable across multiple days.

Clinical protocol in Europe provided early guidance on this: the European College of Radiology for example released guidelines recommending the use of bedside imaging such as mobile X Rays or ultrasound for daily tracking of patients. 

qXR began quantifying what was detected, for example the percentage of lung impacted by an opacity, which was then adopted by Western hospitals in their progression monitoring.

These hospitals include Italian San Raffaele Hospital in Milan, for instance, to monitor disease progression, while Bolton NHS Foundation Trust, for instance, has also adopted qXR to help monitor COVID-19 progression in patients.

Additional users are based in Italy, Portugal, France, Mexico, the US, India and Pakistan, with talks ongoing with Australia, Peru and other countries.

Accurate data

Qure.ai has been sourcing diverse and large data sets from multiple countries for its algorithms. The current data access with the company is more than 7M scans and split almost equally between the Americas, Europe and Asia as the source of the data. This includes data from diverse patient settings and multiple equipment vendors.

This data sourcing is a significant effort and investment as in most countries, Qure.ai is not legally allowed to use client data to train an AI model for COVID-19 triage or diagnosis, owing to GDPR or privacy rules as the data belongs to the hospital or patient.

Qure.ai is therefore having to sign separate agreements with healthcare providers across the globe for data sharing agreements to build its AI models.

Algorithmic updates

The first version of the Qure.ai qXR algorithm was launched three years ago. With COVID, version 2.1 was brought out, and very soon the company shall launch qXR version 3.

While it’s continuously working on new updates, in accordance with a strict and stringent process, not every iteration goes out to the market.

There are 15-16 validation data sets, each from a different country, and enhanced improvement must be seen in all for it to be brought to production.

Every major change requires going back to the regulator, entailing documentation, cost and time. A new version every three months isn’t feasible.

Qure.ai is looking to continually improve on other fronts too. Two examples: avoiding false negatives through looking at X Rays of individuals who are asymptomatic and COVID-19 positive, and using CT scans as a secondary reference to improve accuracy.

Contact tracing and triaging

Qure.ai also repurposed tools aimed at contract tracing - qScout - with the goal of helping health systems monitor people while in isolation.

Through managing symptoms and triaging people remotely using the AI-powered tool, Qure.ai ensures that patients are only brought to the hospital at the right time, healthcare systems manage their workload and avoid placing a burden on limited infrastructure.

These remote monitoring tools were adapted from dashboards originally applied to help tuberculosis screening partners using Qure.ai’s AI solutions, especially in Africa and India.

Built in under 30 days, qScout was initially deployed at the Ministry of Health in Oman, for potential monitoring of 15,000 patients across the country using AI to minimise the intervention of medical personnel in early quarantine stages. The tool has now been in live use for approximately 3 months and has supported more than 40,000 people in self-monitoring and triage using its proprietary chatbot. 

The medical test program was integrated with an app already in use by the Ministry of Health, Tarassud Plus, which diagnoses, follows up and tracks the medical condition of individuals infected with COVID-19.