Artificial intelligence outperforms pulmonologists in the interpretation of pulmonary function tests. Issue 4 (11th April 2019)

• Record Type: Journal Article
• Authors: Topalovic, Marko; Das, Nilakash; Burgel, Pierre-Régis; Daenen, Marc; Derom, Eric; Haenebalcke, Christel; Janssen, Rob; Kerstjens, Huib A.M.; Liistro, Giuseppe; Louis, Renaud; Ninane, Vincent; Pison, Christophe; Schlesser, Marc; Vercauter, Piet; Vogelmeier, Claus F.; Wouters, Emiel; Wynants, Jokke; Janssens, Wim
• Other Names: author non-byline.; De Pauw R. author non-byline.; Depuydt C. author non-byline.; Haenebalcke C. author non-byline.; Muyldermans S. author non-byline.; Ringoet V. author non-byline.; Stevens D. author non-byline.; Bayat S. author non-byline.; Benet J. author non-byline.; Catho E. author non-byline.; Claustre J. author non-byline.; Fedi A. author non-byline.; Ferjani M.A. author non-byline.; Guzun R. author non-byline.; Isnard M. author non-byline.; Nicolas S. author non-byline.; Pierret T. author non-byline.; Pison C. author non-byline.; Rouches S. author non-byline.; Wuyam B. author non-byline.; Corhay J.L. author non-byline.; Guiot J. author non-byline.; Ghysen K. author non-byline.; Renaud L. author non-byline.; Sibille A. author non-byline.; De La Barriere H. author non-byline.; Charpentier C. author non-byline.; Corhut S. author non-byline.; Hamdan K.A. author non-byline.; Schlesser M. author non-byline.; Wirtz G. author non-byline.; Alabadan E. author non-byline.; Birsen G. author non-byline.; Burgel P.R. author non-byline.; Chohra A. author non-byline.; Hamard C. author non-byline.; Lemarié B. author non-byline.; Lothe M.N. author non-byline.; Martin C. author non-byline.; Sainte-Marie A.C. author non-byline.; Sebane L. author non-byline.; Berk Y. author non-byline.; de Brouwer B. author non-byline.; Janssen R. author non-byline.; Kerkhoff J. author non-byline.; Spaanderman A. author non-byline.; Stegers M. author non-byline.; Termeer A. author non-byline.; van Grimbergen I. author non-byline.; van Veen A. author non-byline.; van Ruitenbeek L. author non-byline.; Vermeer L. author non-byline.; Zaal R. author non-byline.; Zijlker M. author non-byline.; Aumann J. author non-byline.; Cuppens K. author non-byline.; Degraeve D. author non-byline.; Demuynck K. author non-byline.; Dieriks B. author non-byline.; Pat K. author non-byline.; Spaas L. author non-byline.; Van Puijenbroek R. author non-byline.; Weytjens K. author non-byline.; Wynants J. author non-byline.; Adam V. author non-byline.; Berendes B.J. author non-byline.; Hardeman E. author non-byline.; Jordens P. author non-byline.; Munghen E. author non-byline.; Tournoy K. author non-byline.; Vercauter P. author non-byline.; Alame T. author non-byline.; Bruyneel M. author non-byline.; Gabrovska M. author non-byline.; Muylle I. author non-byline.; Ninane V. author non-byline.; Rozen D. author non-byline.; Rummens P. author non-byline.; Van Den Broecke S. author non-byline.; Froidure A. author non-byline.; Gohy S. author non-byline.; Liistro G. author non-byline.; Pieters T. author non-byline.; Pilette C. author non-byline.; Pirson F. author non-byline.; Kerstjens H. author non-byline.; Van den Berge M. author non-byline.; Ten Hacken N. author non-byline.; Duiverman M. author non-byline.; Koster D. author non-byline.; Vosse B. author non-byline.; Conemans L. author non-byline.; Maus M. author non-byline.; Bischoff M. author non-byline.; Rutten M. author non-byline.; Agterhuis D. author non-byline.; Sprooten R. author non-byline.; Beutel B. author non-byline.; Jerrentrup A. author non-byline.; Klemmer A. author non-byline.; Viniol C. author non-byline.; Vogelmeier C. author non-byline.; Bode H. author non-byline.; Dooms C. author non-byline.; Gullentops D. author non-byline.; Janssens W. author non-byline.; Nackaerts K. author non-byline.; Rutens D. author non-byline.; Wauters E. author non-byline.; Wuyts W. author non-byline.; Derom E. author non-byline.; Dobbelaere S. author non-byline.; Loof S. author non-byline.; Serry G. author non-byline.; Putman B. author non-byline.; Van Acker L. author non-byline.; Vandeweygaerde Y. author non-byline.; Criel M. author non-byline.; Daenen M. author non-byline.; Gubbelmans R. author non-byline.; Klerkx S. author non-byline.; Michiels E. author non-byline.; Thomeer M. author non-byline.; Vanhauwaert A. author non-byline.;
• Is Part Of: European respiratory journal. Volume 53:Issue 4(2019)
• Publication Date: 2019-04-11
• Subjects: Respiratory organs -- Diseases -- Periodicals; Respiration -- Periodicals; 616.2
• DOI: 10.1183/13993003.01660-2018
• Languages: English
• ISSNs: 0903-1936

Abstract

The interpretation of pulmonary function tests (PFTs) to diagnose respiratory diseases is built on expert opinion that relies on the recognition of patterns and the clinical context for detection of specific diseases. In this study, we aimed to explore the accuracy and interrater variability of pulmonologists when interpreting PFTs compared with artificial intelligence (AI)-based software that was developed and validated in more than 1500 historical patient cases. 120 pulmonologists from 16 European hospitals evaluated 50 cases with PFT and clinical information, resulting in 6000 independent interpretations. The AI software examined the same data. American Thoracic Society/European Respiratory Society guidelines were used as the gold standard for PFT pattern interpretation. The gold standard for diagnosis was derived from clinical history, PFT and all additional tests. The pattern recognition of PFTs by pulmonologists (senior 73%, junior 27%) matched the guidelines in 74.4±5.9% of the cases (range 56–88%). The interrater variability of κ=0.67 pointed to a common agreement. Pulmonologists made correct diagnoses in 44.6±8.7% of the cases (range 24–62%) with a large interrater variability (κ=0.35). The AI-based software perfectly matched the PFT pattern interpretations (100%) and assigned a correct diagnosis in 82% of all cases (p<0.0001 for both measures). The interpretation of PFTs by pulmonologists leads to marked variations and errors. AI-based software provides more accurate interpretations and may serve as a powerful decision support tool to improve clinical practice. There is poor accuracy and substantial disagreement between pulmonologists when interpreting complex pulmonary function data. Automating interpretation with artificial intelligence provides a powerful decision support tool in clinical practice. http://ow.ly/Tj9h30nxw4U