Algebraic topology-based machine learning using MRI predicts outcomes in primary sclerosing cholangitis

Area of Science:

• Medical imaging analysis
• Machine learning applications in hepatology
• Topological data analysis

Background:

• Primary sclerosing cholangitis (PSC) is a chronic liver disease with unpredictable outcomes.
• Predicting hepatic decompensation in PSC patients remains a clinical challenge.
• Magnetic resonance imaging (MRI) offers potential for prognostic assessment.

Purpose of the Study:

• To develop a machine learning model utilizing algebraic topology.
• To extract predictive MRI features for hepatic decompensation in PSC.
• To enable early identification of patients at high risk for liver failure.

Main Methods:

• Retrospective multicenter study of adult large duct PSC patients with MRI.
• Application of a topological data analysis-inspired nonlinear framework.
• Training and validation of a machine learning model using persistence images from contrast-enhanced T1-weighted MRI.

Main Results:

• The model was trained on a derivation cohort and validated on an independent cohort (n=115).
• In the validation cohort, the model achieved an area under the receiver operating characteristic curve of 0.84 for predicting early hepatic decompensation.
• The algebraic topology-based machine learning approach demonstrated predictive capability for clinical outcomes.

Conclusions:

• Algebraic topology-based machine learning is a viable method for predicting outcomes in PSC.
• This approach shows promise for predicting hepatic decompensation and other outcomes in chronic liver diseases.
• Integrating advanced ML techniques with medical imaging can improve prognostic accuracy in liver disease management.