Pneumologie 2023; 77(S 01): S73-S74
DOI: 10.1055/s-0043-1761042
Abstracts

ArtiQ.PFT: AI-powered decision support for the diagnosis of lung diseases: Does it help the pulmonologist to be more accurate?

P Desbordes
1   Artiq NV
,
M Steinwender
2   Department of Lung Diseases
,
E Helfenstein
3   Lungenpraxis Hirslanden – Klinik St Anna
,
C Marginean
4   Country Clinical Hospital Mures – Clinic of Pulmonary Diseases Tg Mures
,
N Henzi
5   Department of Lung Diseases – Spital Limmattal
,
P Cerkl
6   Department of Pulmonology – Landeskrankenhaus Hohenems
,
M Spielmanns
7   Pulmonary Medicine and Sleep Medicine Center – Zuercher Rehazentren Clinic – Wald
,
S Keusch
8   Lungdocs
,
D Stolz
9   Pulmonary Division, University Hospital Basel
,
M Simon
10   Department of Bronchology – Leon Daniello Clinic of Pulmonology
,
D Gompelmann
11   Department of Medicine II – Division of Pulmonology – Medical University of Vienna
,
M Topalovic
1   Artiq NV
› Institutsangaben
› Weitere Informationen
Auch verfügbar aufeRef
 
 

    Background ArtiQ.PFT (ArtiQ NV, Belgium) is an Artificial Intelligence (AI) based software that helps physicians in the interpretation of pulmonary function tests (PFTs). It provides an automated description of the lung function according to the latest guidelines and calculates disease probabilities to support the diagnostic process.

    Methods A retrospective survey has quantified the diagnostic accuracy of ArtiQ.PFT and the impact of the next steps suggested by the software. 60 cases (Healthy, Asthma, COPD, ILD, NMD, OBD, PVD, TD) are presented to 20 pulmonologists from 3 different countries (Austria, Switzerland, and Romania). For each case, the full PFT (spirometry, body plethysmography, diffusion) and a short anamnesis are provided. Pulmonologists provide a primary and up to 3 differential diagnoses per case as well as they next step they would take, once without and once with the support of ArtiQ.PFT.

    Results Experts gave on average 2.1 diagnoses (primary and differential) and this number was similar whether they were supported by AI or not. For all considered diagnoses the use of AI improves the diagnosis prediction by 18.0% (60.0% without vs 78.0% with AI). Similarly, when looking only at the primary diagnosis, the use of AI improves diagnostic accuracy by 18.9% (44.1% vs 63.0%). Finally, physicians have slightly increased confidence in their diagnosis when using ArtiQ.PFT (3.5 vs 3.6 on a scale of 1 – 5). Interestingly, 15% of the suggested next steps differ. Proposals for “immediate further investigation” and “plan follow-up” are replacing “back to general practitioner” and “start treatment”.

    Conclusions The results indicate that the AI-based decision support helps improve the accuracy of diagnostic prediction by pulmonologists when interpreting PFTs. In addition, the next steps are influenced by the AI support as well. Further research is needed to evaluate if the results differ between diseases; eg difference between common and more rare diseases.


     

    Publikationsverlauf

    Artikel online veröffentlicht:
    09. März 2023

    © 2023. Thieme. All rights reserved.

    Georg Thieme Verlag
    Rüdigerstraße 14, 70469 Stuttgart, Germany