Precision diagnostics in pediatric chest involvement of COVID-19: Added value of an artificial intelligence tool in chest CT analysis

Principal Investigators

The European Commission wants to understand the added value of artificial intelligence (AI) in health. In this sense, and using the opportunity presented by the COVID-19 pandemic, it has decided to create a project to analyse the value of AI in chest CT. This project will initially involve 12 hospitals across Europe, which will test the added value of using AI in radiology departments. SJD Barcelona Children's Hospital has been selected by the European Commission to carry out this project and to date it is the only fully pediatric hospital involved.

Chest Computed Tomography (CT) + Artificial Intelligence (AI): a potential complimentary approach to RT-PCR for COVID-19 detection and lesion characterization.

Rationale

  • When radiography cannot provide a clear diagnosis of COVID-19 pneumonia, chest CT has proven to be a useful tool, as it is more sensitive and specific and can also classify the degree of COVID-19 involvement and detect possible complications such as pulmonary embolism
  • However, chest CT can be difficult to interpret for untrained personnel, and a lack of experience in diagnosing and assessing the severity of COVID-19 pneumonia may lead to inaccurate conclusions.
  • Based on previous experience of using AI to support medical imaging, it can be considered likely that the use of pulmonary AI tools with chest CT for COVID-19 patients would enable both (a) an increase in the speed of response to the disease and (b) a quantification of pulmonary changes and therefore the drawing of more precise conclusions on the involvement and extent of the disease.
  • In addition to the expected benefits for adult patients, it must also be evaluated whether the same benefits can be achieved for pediatric patients, who are known to have a different, as yet poorly understood, course of the disease to adults.

Main questions:

  • Does AI enable us to distinguish COVID-19 cases from other lung diseases?
  • Does AI improve the diagnosis and classification of COVID-19 pediatric patients?
  • Can we quantify COVID-19 lung disease in a standard way and can that information be shared in a comprehensible way?

Project design

Chest CT has proven to be a very accurate diagnostic and prognostic tool in COVID-19 lung involvement, as 96% of patients with COVID-19 present chest imaging abnormalities. Thus, a CT+AI approach could be used as a first diagnostic tool with subsequent RT-PCR as a confirmation tool.

As part of a European consortium from 12 referral hospitals and institutions for COVID-19, we are helping to address the COVID-19 outbreak in collaboration with the company InferVision.

InferVision has developed an AI solution (InferRead CT Lung COVID-19) based on CT lung scans to help radiologists make diagnostic judgments on COVID-19 patients' conditions quickly and accurately, providing not only confirmation of real COVID-19 cases but also information about the extent of the lesion and a solid basis to evaluate treatment options. Specifically, InferRead CT Lung COVID-19 allows:

  • Quick identification and triage of pneumonia caused by COVID-19.
  • Comparisons with prior studies to quantify the disease progression in confirmed cases.
  • Auto-generation of a structured report.

The European consortium provides a unique setting to validate InferRead CT Lung COVID-19 with thousands of real COVID-19 pediatric and adult cases for disease prevention and control.

Scientific objectives

In collaboration with InferVision, the European consortium will validate the AI-based solution InferRead CT Lung COVID-19 in CT lung scans from real COVID-19 pediatric and adult cases. This approach may be used as a complement to other diagnostic techniques such as RT-PCR and to quantify the disease progression in confirmed cases.

SJD Barcelona Children's Hospital will be the only exclusively pediatric center that participates in studying the added value of AI algorithms in diagnosis and thoracic prognosis.

Data and samples

This project will study the patients admitted to SJD Barcelona Children's Hospital who have clinical-radiological criteria for chest CT, either to complete a COVID-19 diagnosis or because their underlying disease indicates it. We will study the findings on chest CT and compare them with the results that the AI algorithms indicate.

Group members

Who is behind this research

The project has been directly assigned by the European Commission to the consortium hospitals as well as to the company InferVision.

InferVision is a leading global AI high-tech enterprise dedicated to applying deep learning technology to assist medical image diagnosis with efficient and accurate solutions. Based in Beijing, InferVision was founded in 2015 and has rapidly expanded globally to North America, Asia-Pacific and Europe. It has also established cooperative business partnerships with hospitals and institutions world-wide and successfully validated their AI and deep learning technologies in real cases.

SJD Barcelona Children's Hospital has been collaborating with InferVision since 2017, validating AI algorithms in emergency plain chest radiography.

Collaborators

European consortium of referral hospitals and institutions:

  • Spain: SJD Barcelona Children's Hospital, Vall d'Hebrón University Hospital, University Hospital Parc Tauli, Dexeus University Hospital.
  • France: CHU Nimes.
  • Germany: University Medical Center of Mainz, University Hospital Jena.
  • Austria: Xcoorp GmbH Tele-radiology.
  • Switzerland: University Hospital Zurich, University Hospital Bern.
  • Italy: Campus Bio-Medico University Hospital, O3-Enterprise Teleradiology.