The primary goal of our study was to investigate the potential association between disease patterns in lung CT data, their changes over time, and the individual outcomes of patients diagnosed with Idiopathic Pulmonary Fibrosis (IPF). We aimed to employ deep learning algorithms to segment and quantify the lung texture patterns in CT scans and examine the potential relationship between these patterns and survival. The goal was to identify specific lung texture patterns that may be indicative of patient outcomes. Therefore, our study aimed to provide...

The study enrolled 55 patients who had been diagnosed with IPF, and each patient underwent at least two subsequent CT scans. Using an image analysis deep learning algorithm, the lung texture patterns in each scan were automatically segmented, and lung voxel was assigned to one of six distinct disease patterns based on their appearance in the CT scan. These patterns included ground-glass opacity, reticular pattern, emphysema, nodular pattern, honeycombing, and consolidation. The volume of the total voxels containing these six patterns was calculated for each...

Our study included a cohort of 55 IPF patients, among whom 21 patients passed away before the censoring date, and the average time between their scans and death was 18.9 months. The remaining 34 patients had a censoring data on average 73.9 months after the scans. Our analysis revealed that both the static and dynamic signature patterns were significantly associated with patient survival. Specifically, we found that the honeycombing and consolidation patterns were significantly associated with survival, with a hazard ratio of 5.39 and 5.35,...

In conclusion, our study demonstrated that deep learning algorithms can be used to accurately quantify and analyze disease patterns in CT scans that are associated with individual outcomes of patients with IPF. Our findings suggest that both the static and dynamic pattern load profiles have significant correlations with survival, with honeycombing and consolidation patterns showing a strong association with patient outcomes. However, it is important to note that our study had some limitations, such as a small sample size. Therefore, further research with larger sample...

J. Pan: Nothing to disclose K-H. Nenning: Nothing to disclose S. Röhrich: Nothing to disclose N. Sverzellati: Nothing to disclose V. Poletti: Nothing to disclose J. Hofmanninger: Nothing to disclose A. Makropoulos: Nothing to disclose H. Prosch: Nothing to disclose G. Langs: Shareholder: Contextflow GmbH

-