Personalized Lung Care

Over the last few years, we have focused on diagnosing and managing interstitial lung diseases (ILDs) using state-of-the-art AI-based approaches. Our work spans multimodal data integration, mechanistic informed-AI, and AI-driven clinical decision support, with a particular emphasis on fibrotic ILDs (fILDs).

Interstitial lung diseases

Interstitial lung diseases (ILDs) comprise a large and diverse group of chronic lung disorders that lead to progressive scarring of lung tissue and, ultimately, respiratory failure. Despite their clinical importance, diagnosis remains challenging and often delayed, even with multidisciplinary expert evaluation, resulting in inconsistent treatment decisions and suboptimal outcomes.

Our research focuses on developing AI methods to improve the diagnosis, prognosis, and personalized management of ILDs. We analyze multimodal data, including CT imaging, clinical records, demographics, and molecular biomarkers, to better characterize disease patterns and support clinical decision-making. Building on the SNSF-funded INTACT project, we have developed AI-based imaging tools for automated lung tissue characterization and radiological pattern recognition at a near-expert level. Current work extends these approaches by integrating longitudinal clinical and biological data to distinguish inflammatory and fibrotic disease processes and to predict individual disease trajectories. This research is supported by the Swiss National Science Foundation (SNSF) within PRISM-fILD.

Lung Ultrasound

We are also expanding our AI research to lung ultrasonography (LUS), a non-irradiating, bedside imaging modality with demonstrated sensitivity for detecting pneumonia. We investigate deep learning approaches for pneumonia classification from LUS images, leveraging contrastive and multimodal learning to align representations across imaging modalities — ultrasound, CT, and chest X-ray — to improve diagnostic accuracy and enable cross-modal knowledge transfer.