{
  "abstract": "Introduction Interstitial Lung Disease (ILD) prevalence is rising, yet ILD often remains misdiagnosed or diagnosed late, particularly due to referral delays (3.8±2.1 years between first spirometry and diagnosis). 1 In primary care, spirometry is the predominant test for assessing respiratory symptoms, but ILD is often overlooked. ArtiQ.Spiro, an AI-based support tool for spirometry interpretation, shows strong potential to assist clinicians in identifying possible ILD. We examine the prevalence of AI-flagged ILD cases in a real-world primary care setting to evaluate AI’s potential to prompt earlier specialist referrals.Methods The Hillingdon Confederation runs a central spirometry hub for four GP practices and uses ArtiQ.Spiro to support spirometry execution and interpretation. We applied the AI model to demographic and clinical spirometry data from 2,794 consecutive individuals undergoing spirometry between October 2022 and May 2025. Interpretation followed the ERS/ATS 2021 standard. The AI-predicted ILD cases in primary care were assessed and contextualised based on spirometry patterns. Prediction reliability was supported by prior external validation 2 demonstrating strong ILD screening performance (AUC=0.9, NPV=92.3%, PPV=60.5%).Results The model predicted ILD in 7.8% of cases (n=219; see the table for characteristics). These predictions are associated with higher age compared to the total cohort, and reduced FVC z-scores with normal to elevated FEV1/FVC z-scores (0.41±0.93), as expected in ILD patients when compared to a healthy population. Among the ILD predicted cases, 55.7% exhibited a possible restrictive pattern, 43.4% a normal pattern, and 0.9% a possible mixed disorder. Predicted ILD cases include 24.6% of the patients with obesity and restrictive pattern, indicating the model’s predictions are not driven by obesity-related restriction. While the previously validated PPV suggests some overprediction, high NPV provides confidence that most true ILD cases are being identified.Conclusions ArtiQ. Spiro can support ILD diagnosis by flagging at-risk patients across both restrictive and normal spirometry patterns that might otherwise be overlooked. Further work is needed to evaluate how integrating ArtiQ.Spiro into diagnostic pathways could impact secondary care by weighing the benefits of earlier ILD detection against the risks of unnecessary specialist referrals.References Topalovic. ERJ, 2022. https://doi.org/10.1183/13993003.congress-2022.1217Sunjaya. ERJ Open Research 2025. https://doi.org/10.1183/23120541.00116-2025",
  "authors": [
    {
      "affiliations": [
        "ArtiQ, a Clario company, Leuven, Belgium"
      ],
      "name": "C Vairelles"
    },
    {
      "affiliations": [
        "ArtiQ, a Clario company, Leuven, Belgium"
      ],
      "name": "M Rakić"
    },
    {
      "affiliations": [
        "ArtiQ, a Clario company, Leuven, Belgium"
      ],
      "name": "E Smets"
    },
    {
      "affiliations": [
        "ArtiQ, a Clario company, Leuven, Belgium"
      ],
      "name": "P Desbordes"
    },
    {
      "affiliations": [
        "Harefield Respiratory Research Group, Guy’s and St Thomas’ NHS Foundation Trust, London, UK"
      ],
      "name": "SSC Kon"
    },
    {
      "affiliations": [
        "The Confederation, Hillingdon CIC, London, UK"
      ],
      "name": "S Sehdev"
    },
    {
      "affiliations": [
        "Department of Respiratory Medicine, The Hillingdon Hospitals NHS Foundation Trust, London, UK"
      ],
      "name": "WD-C Man"
    },
    {
      "affiliations": [
        "ArtiQ, a Clario company, Leuven, Belgium"
      ],
      "name": "M Topalović"
    }
  ],
  "title": "S155 Potential of AI to detect ILD in primary care settings: a real-world evaluation",
  "uid": "b49b247d-0ce4-5173-9d3c-1f2ab8825370"
}