Multicenter Validation Study of an Artificial Intelligence Tool for Automatic Classification of Chest X-rays

NCT04991987 | Unknown

• 1 other identifier: 6025
• Study Type: observational
• Target: 385
• Locations: 1 country
• Sites: 1

Brief Summary

A current problem in Radiology Departments is the constant increase in the number of studies performed. Currently the largest volume of studies belongs to plain x-rays. This problem is intensified by the shortage of specialists with dedication and experience in their interpretation. In the field of computer science, an area of study called Artificial Intelligence (AI) has emerged, which consists of a computer system that learns to perform specific routine tasks, and can complement or imitate human work. Since 2018, Hospital Italiano de Buenos Aires has been running the TRx program, which consists of the development of an AI-based tool to detect pathological findings in chest x-rays. The intended use of this tool is to assist non-imaging physicians in the diagnosis of chest x-rays by automatically detecting radiological findings. The present multicenter study seeks to externally validate the performance of an AI tool (TRx v1) as a diagnostic assistance tool for chest x-rays.

Conditions

Pneumothorax; Pleural Effusion; Bone Fracture; Consolidation; Opacity

Keywords

Radiography; Thorax; Artificial Intelligence; Machine Learning; Deep learning

Outcome Measures

Primary Outcomes (1)

• Concordance between AI tool and reference standard

  The concordance between the category assigned by the professionals and that assigned by the algorithm will be analyzed. For this purpose, a diagnostic test will be evaluated for the detection of abnormality (i.e., the test is positive when at least one of the four types of findings is observed). Considering the specialists' diagnosis as a reference standard, the confusion matrix will be constructed and the diagnostic metrics of the AI tool (sensitivity, specificity and predictive values) will be calculated. The 95% confidence intervals will be calculated using exact binomial distribution.

  5 months

Secondary Outcomes (4)

• Receiver Operating Characteristic curves

  5 months

• Qualitative analysis

  5 months

• Inter-observer concordance index

  5 months

• Analysis by institution

  5 months

Eligibility Criteria

• Age: 18 Years+
• Sex: all
• Age Groups: Adult (18-64), Older Adult (65+)
• Sampling Method: Probability Sample

Study Population

The unit of study will be the chest X-rays provided by the participating centers, maintaining the confidentiality of the patient in question (without any sensitive data such as name, surname, ID card number or date of birth). The images will be obtained retrospectively from their respective institutional databases.

You may qualify if:

• X-rays that meet the following requirements will be included:
• Chest X-ray
• Belong to patients over 18 years of age.
• Advocacy and digital acquisition
• Study conducted in the aforementioned institutions and stored in their respective Picture Archiving and Communication System

You may not qualify if:

• X-rays that are excluded:
• Poor technique (low contrast, veiled, off-center)
• Presence of abnormal position of the patient during acquisition.

Sponsors & Collaborators

• Hospital Italiano de Buenos Aires: lead

Study Sites (1)

Hospital Italiano de Buenos Aires

Buenos Aires, 1199, Argentina

Location

Related Publications (6)

• Kesselman A, Soroosh G, Mollura DJ; RAD-AID Conference Writing Group. 2015 RAD-AID Conference on International Radiology for Developing Countries: The Evolving Global Radiology Landscape. J Am Coll Radiol. 2016 Sep;13(9):1139-1144. doi: 10.1016/j.jacr.2016.03.028. Epub 2016 May 25.

  PMID: 27233909 BACKGROUND

• Chartrand G, Cheng PM, Vorontsov E, Drozdzal M, Turcotte S, Pal CJ, Kadoury S, Tang A. Deep Learning: A Primer for Radiologists. Radiographics. 2017 Nov-Dec;37(7):2113-2131. doi: 10.1148/rg.2017170077.

  PMID: 29131760 BACKGROUND

• Erickson BJ, Korfiatis P, Akkus Z, Kline TL. Machine Learning for Medical Imaging. Radiographics. 2017 Mar-Apr;37(2):505-515. doi: 10.1148/rg.2017160130. Epub 2017 Feb 17.

  PMID: 28212054 BACKGROUND

• Balthazar P, Harri P, Prater A, Safdar NM. Protecting Your Patients' Interests in the Era of Big Data, Artificial Intelligence, and Predictive Analytics. J Am Coll Radiol. 2018 Mar;15(3 Pt B):580-586. doi: 10.1016/j.jacr.2017.11.035. Epub 2018 Feb 6.

  PMID: 29402532 BACKGROUND

• Calvert JS, Price DA, Chettipally UK, Barton CW, Feldman MD, Hoffman JL, Jay M, Das R. A computational approach to early sepsis detection. Comput Biol Med. 2016 Jul 1;74:69-73. doi: 10.1016/j.compbiomed.2016.05.003. Epub 2016 May 12.

  PMID: 27208704 BACKGROUND

• Mosquera C, Diaz FN, Binder F, Rabellino JM, Benitez SE, Beresnak AD, Seehaus A, Ducrey G, Ocantos JA, Luna DR. Chest x-ray automated triage: A semiologic approach designed for clinical implementation, exploiting different types of labels through a combination of four Deep Learning architectures. Comput Methods Programs Biomed. 2021 Jul;206:106130. doi: 10.1016/j.cmpb.2021.106130. Epub 2021 May 2.

  PMID: 34023576 BACKGROUND

Related Links

• Weakly Supervised Learning of Deep Convolutional Neural Networks \[Internet\]. 2016 Institute of Electrical and Electronics Engineers, Conference on Computer Vision and Pattern Recognition. 2016.
• Guest Editorial Deep Learning in Medical Imaging: Overview and Future Promise of an Exciting New Technique \[Internet\]. Vol. 35, Institute of Electrical and Electronics Engineers, Transactions on Medical Imaging. 2016. p. 1153-9.
• Dataset shift in machine learning. Neural Information Processing. 2008.

MeSH Terms

Conditions

Pneumothorax; Pleural Effusion; Fractures, Bone

Condition Hierarchy (Ancestors)

Pleural Diseases; Respiratory Tract Diseases; Wounds and Injuries

Study Officials

• Sonia E Benitez, MD, MSc

  Hospital Italiano de Buenos Aires

  PRINCIPAL INVESTIGATOR

Study Design

• Study Type: observational
• Observational Model: OTHER
• Time Perspective: PROSPECTIVE
• Sponsor Type: OTHER
• Responsible Party: SPONSOR

Study Record Dates

• First Submitted: July 28, 2021
• First Posted: August 5, 2021
• Study Start: July 1, 2021
• Primary Completion: February 28, 2022
• Study Completion: July 31, 2022
• Last Updated: August 5, 2021

Record last verified: 2021-07

Locations

AR (1)