Integrating Deep Learning CT-scan Model, Biological and Clinical Variables to Predict Severity of Asthma in Children (BREATHE)

Institute of Hospitalization and Scientific Care (IRCCS)

Status

Enrolling

Conditions

Asthma in Children

Study type

Observational

Funder types

Other

Identifiers

NCT05140889

08073521

Details and patient eligibility

About

Artificial intelligence (AI) offers substantial opportunities for healthcare, supporting better diagnosis, treatment, prevention and personalized care. Analysis of health images is one of the most promising fields for applying AI in healthcare, contributing to better prediction, diagnosis and treatment of diseases.

Deep learning (DL) is currently one of the most powerful machine learning techniques. DL algorithms are able to learn from raw (or with little pre-processing) input data and build by themselves sophisticated abstract feature representations (useful patterns) that enable very accurate task decision making. Recently, DL has shown promising results in assisting lung disease analysis using computed tomography (CT) images.

Current severe asthma guidelines recommend high-resolution and multidetector CT as a tool for disease evaluation. CT scans contain prognostic information, as the presence of bronchial wall thickening, air trapping, bronchial luminal narrowing, and bronchiectasis are associated with longer disease duration and disease severity in adults. Only a small number of studies have reported chest CT findings in children with severe asthma, and their relationship to clinical and pathobiological parameters yielded inconsistent results. Thus, to which extent CT scans add prognostic information beyond what can be inferred from clinical and biological data is still unresolved in children.

The project is expected to build an DL-severity score to prognoses severe evolution for children with asthma, using a DL model to capture CT scan prognosis information.

Full description

The aims of this project are:

to build a large database of clinical, biological and radiological data collected from pediatric patients with severe asthma;
to design and train a forecaster model based on DL techniques to predict asthma severity in children;
to estimate transition probabilities between asthma severity levels using a multi-state Markov model taking into account qualitative and quantitative information obtained from CT imaging.

Our evaluation of DL-severity and existing clinical scores in childhood asthma is expected to reveal that emerging methodologies assisted by DL techniques can provide accurate severity predictions, when compared with existing clinical scores. Such an accurate prediction model would allow pediatricians to identify features that are the most indicative of severity and progression of asthma and would be employed to formulate intervention strategies and early medical attention for children.

Enrollment

25 estimated patients

Sex

All

Ages

6 to 17 years old

Volunteers

No Healthy Volunteers

Inclusion criteria