Semi-Automated Quantitative CT–Enhanced GAP Index for Prognostic Assessment in Idiopathic Pulmonary Fibrosis

Background Idiopathic pulmonary fibrosis (IPF) is marked by spatially heterogeneous fibrotic remodeling and highly variable clinical trajectories, which limit the precision of existing prognostic tools. Widely used indices, such as GAP (Gender–Age–Physiology), rely on physiological measurements and do not explicitly capture anatomical disease burden. Advances in artificial intelligence (AI)–assisted 3D quantitative Computed Tomography (CT) imaging enable objective, volumetric characterization of lung fibrosis and provide a foundation for radiology-integrated prognostic modeling. Methods We developed and evaluated an AI-assisted quantitative CT framework for prognostic assessment in patients with IPF. Whole-lung segmentation was performed using a pretrained deep learning model (LungMask), followed by semi-automated segmentation of fibrotic abnormalities with radiologist-guided refinement. Quantitative imaging biomarkers, including honeycombing volume, non-honeycomb fibrotic lung volume, preserved lung volume, and the pathological lung volume-to-total lung volume ratio, were derived from 3D CT analysis. This radiological parameter was integrated into the conventional GAP index to create a radiology-enhanced prognostic model. Prognostic performance was assessed using receiver operating characteristic (ROC) analysis, and Kaplan–Meier survival analysis. Results A total of 116 patients with IPF were included in the study (mean age: 64.2 ± 9.2 years; 69.8% male). During a median follow-up of 40.5 months, 44% of patients died. Patients who died had significantly lower baseline pulmonary function parameters and greater fibrotic lung involvement on AI-assisted quantitative CT analysis than survivors (all p < 0.05). The pathological lung volume-to-total lung volume ratio showed good prognostic performance for predicting mortality (AUC = 0.795, 95% CI: 0.713–0.877, p < 0.001). Integrating this radiological parameter into the conventional GAP index further improved prognostic performance (AUC = 0.805, 95% CI: 0.725–0.885, p < 0.001) compared with the GAP index alone (AUC = 0.712, 95% CI: 0.618–0.807). Kaplan–Meier analysis showed significant survival differences among modified GAP risk groups, with median survival of 100 months for patients with a score ≤ 6 and 48 months for those with a score ≥ 7 (log-rank p < 0.001). Conclusions This study presents an interpretable, AI-assisted quantitative CT imaging framework that integrates objective measures of fibrotic lung burden into an established clinical prognostic index. By unifying volumetric CT biomarkers with physiological scoring, the proposed model improves prognostic precision while maintaining clinical usability. This work highlights the potential of AI-assisted quantitative imaging as a scalable, translational tool for personalized risk stratification in fibrotic lung disease.