Clinical Report: Leveraging Liver Analysis for Colorectal Neoplasia Prediction
Overview
This study evaluates the use of AI-based liver analysis from routine CT scans to predict colorectal neoplasia, demonstrating a significant potential for opportunistic screening. The best-performing model achieved an AUROC of 0.810, indicating promising accuracy in identifying colorectal neoplasia.
Background
Colorectal cancer (CRC) is a leading cause of cancer-related mortality, with low screening participation rates contributing to its prevalence. Non-invasive screening methods are crucial for increasing participation and reducing mortality. This study explores the gut-liver axis as a novel biomarker source for CRC risk prediction, leveraging existing CT imaging data.
Data Highlights
Model
AUROC
Sensitivity
Specificity
XGBoost
0.810
74.1%
72.3%
Clinical-only model
0.457
N/A
N/A
Subclassification (CRC vs. adenoma)
0.674
N/A
N/A
Key Findings
AI-based liver analysis can predict colorectal neoplasia using routine CT scans.
The XGBoost model achieved an AUROC of 0.810, significantly outperforming clinical-only models.
Sensitivity and specificity for detecting colorectal neoplasia were 74.1% and 72.3%, respectively.
Subclassification between CRC and adenoma showed lower accuracy with an AUROC of 0.674.
This approach utilizes existing CT scans, potentially increasing screening participation without additional burden.
Clinical Implications
The findings suggest that integrating AI-based liver analysis into routine practice could enhance colorectal neoplasia screening efforts. Clinicians may consider leveraging existing CT imaging data to identify patients at risk for CRC, thereby facilitating earlier intervention.
Conclusion
AI-driven analysis of liver features from routine CT scans presents a promising adjunct to traditional colorectal cancer screening methods. This approach could significantly improve early detection and risk stratification in clinical settings.
by Anna Hinterberger, Jonas Bohn, Darya Trofimova, Nicolas Knabe, Julia Dettling, Tobias Norajitra, Fabian Isensee, Johannes Betge, Stefan O. Schönberg, Dominik Nörenberg, Sergio Grosu, Sonja Loges, Ralf Floca, Jakob Nikolas Kather, Klaus Maier-Hein, Freba Grawe
