Idiopathic pulmonary fibrosis (IPF) is a chronic and progressive interstitial lung disease characterized by a highly heterogeneous clinical trajectory, which complicates prognostic assessment and therapeutic planning. In this thesis, we investigate a multimodal machine learning framework for predicting short-term functional progression in IPF, defined as a ≥10% decline in forced vital capacity (FVC) at one-year follow-up. Baseline chest computed tomography (CT) scans and structured patient data are integrated to assess whether imaging-derived representations provide additional prognostic value beyond clinical variables. From CT scans, convolutional neural networks (CNNs) are used as feature extractors to obtain high-dimensional slice-level embeddings, which are aggregated at the patient level through different pooling strategies.