{
  "abstract": "Introduction Progressive interstitial lung disease (ILD) is the leading cause of mortality in patients with systemic sclerosis (SSc). Early recognition of progression is crucial to enable timely treatment. However, reliable markers to early intercept or predict progression are lacking. Artificial intelligence (AI)-assisted analysis of chest high-resolution CT (HRCT) scans might provide a valuable tool for this purpose. Our study aims to compare AI-assisted analysis with traditional visual evaluation in detecting early signs of SSc-ILD progression and to assess whether such signs are predictive of subsequent functional deterioration.Material and Methods A retrospective longitudinal study was conducted on 33 patients with SSc-ILD. Clinical data were collected, along with pulmonary function tests (PFT) and high-resolution CT scans at two timepoints (T1, T2) 1 year (IQR 0.75-2.5 years) apart. PFT were further performed at a third timepoint (T3), 12 months (IQR 9-14.5 months) after T2. HRCT scans were blindly evaluated by AI-assisted analysis (Thoracic VCAR, AW SERVER v.3.2, GE Healthcare) and by two chest radiologists for semi-quantitative analysis of fibrosis and ground glass opacities (GGO). 1 Correlations of CT parameters with PFT and prediction of ILD functional progression defined using Erice functional criteria2 were analysed.Results An overall significant reduction of forced vital capacity (FVC)% was observed from T1 to T2 [88% (IQR 71-101)] to T2 [81% (IQR 69-93); p=0.04), while no differences were detected in other PFT parameters. Both visual and AI-assisted HRCT analyses showed a significant increase in fibrosis over time (visual: 1.5 to 1.6, p=0.0005; AI: 2.2 to 2.4%, p=0.0007), while only AI detected a significant increase in GGO (15.5 to 15.8%, p=0.04) and a reduction of normal lung parenchyma (75.6 to 73%, p=0.009). At both T1 and T2, AI-derived HRCT parameters strongly correlated with PFT values, whereas visual scoring showed weaker correlations. After adjusting for age, sex, and disease duration, early (T1-T2) increases in GGO detected by AI predicted further functional progression at T3 (p=0.04, OR 9e+7, 95% CI 1.37-5.9e+15); in particular, an increase >50 mL (8.2%) predicted progression with 81% sensitivity and 77% specificity ( figure 1). Visual analysis had no predictive value.Conclusions The AI-assisted quantitative HRCT analysis, but not visual scoring, can detect SSc-ILD progression and may identify patients at risk of functional decline.References Kazerooni EA, et al. AJR Am J Roentgenol. 1997;169(4):977–83.George PM, et al. Lancet Respir Med. 2020;8(9):925–34.Abstract P.232 Figure 1ROC curve of ground glass opacities volume increases from T1 to T2 at AI analysis of chest CT to predict functional SSc-ILD progression at T3; AUC-ROC of 0.85, 95% CI 0.68-1.00",
  "authors": [
    {
      "affiliations": [
        "Humanitas University and Department of Rheumatology and Clinical Immunology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Francesca Motta"
    },
    {
      "affiliations": [
        "Humanitas University and Department of Rheumatology and Clinical Immunology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Antonio Tonutti"
    },
    {
      "affiliations": [
        "Humanitas University and Department of Radiology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Federica Catapano"
    },
    {
      "affiliations": [
        "Humanitas University and Department of Radiology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Gianluca Sellaro"
    },
    {
      "affiliations": [
        "Humanitas University and Respiratory Unit, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Francesco Amati"
    },
    {
      "affiliations": [
        "Humanitas University and Respiratory Unit, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Anna Stainer"
    },
    {
      "affiliations": [
        "Department of Radiology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Calogero Messana"
    },
    {
      "affiliations": [
        "Humanitas University and Respiratory Unit, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Stefano Aliberti"
    },
    {
      "affiliations": [
        "Department of Medical-Surgical Sciences and Translational Medicine, School of Medicine and Psychology, Sapienza, Rome, Italy"
      ],
      "name": "Marco Francone"
    },
    {
      "affiliations": [
        "Humanitas University and Department of Rheumatology and Clinical Immunology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Carlo Selmi"
    },
    {
      "affiliations": [
        "Humanitas University and Department of Rheumatology and Clinical Immunology, IRCCS Humanitas Research Hospital, Milan, Italy"
      ],
      "name": "Maria De Santis"
    }
  ],
  "title": "P.232 Early diagnosis of SSc-ILD progression: a comparative study of high-resolution chest CT assessment with visual and artificial intelligence-assisted scoring",
  "uid": "b1a9ade9-cfeb-509e-8e12-eebec6c886fd"
}