Knowledge-Driven Interpretation of Multi-View Data in Medicine.

Area of Science:

• Artificial Intelligence
• Medical Informatics
• Computational Biology

Background:

• Clinical decision support systems (CDSS) often lack interpretability, hindering trust and adoption.
• Multi-view data integration presents challenges in identifying key predictive factors for complex diseases.
• Detecting mild neurocognitive disorder (MND) early is crucial for timely intervention and management.

Purpose of the Study:

• To develop a novel, interpretable approach for clinical decision support using multi-view data.
• To introduce a Bayesian network structure learning method for identifying significant associations.
• To evaluate the method's effectiveness in detecting mild neurocognitive disorder in the elderly.

Main Methods:

• A Bayesian network structure learning algorithm was employed with minimal domain knowledge.
• The method integrated diverse data sources: demography, medical/family history, lifestyle, and biomarkers.
• The approach was validated on a real-life dataset from Singapore for mild neurocognitive disorder detection.

Main Results:

• The proposed method achieved comparable accuracy to benchmark studies for mild neurocognitive disorder detection.
• The approach demonstrated superior interpretability in identifying relevant factors and their relationships.
• Significant associations were highlighted among demography, medical history, lifestyle, and biomarker data.

Conclusions:

• The developed Bayesian network approach enhances interpretability in clinical decision support.
• This method facilitates informed clinical decisions by revealing key disease-associated factors.
• The approach shows promise for early and accurate detection of mild neurocognitive disorder in elderly populations.