Uncertainty-Gated SHAP-Dirichlet Refinement for MRI-Based Depression Classification

Key points

• We propose uncertainty-gated SHAP-Dirichlet refinement, a post-hoc reliability layer that improves predictions without retraining the classifier.1
• The method retrieves similar labeled SHAP explanations, constructs Dirichlet class evidence, and selectively refines uncertain predictions through probabilistic fusion, while allowing unreliable cases to be deferred for expert review.1
• We evaluate five classifiers on 504 SRPBS subjects (252 controls and 252 MDD patients) using 159 anatomical features and repeated nested cross-validation.1
• These findings support selective explanation-guided refinement as a practical approach to improving prediction reliability and facilitating expert triage in MDD classification.1

Sources (1)

• [1]Uncertainty-Gated SHAP-Dirichlet Refinement for MRI-Based Depression Classification
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 01:25 AM

  Structural MRI-based classification of major depressive disorder (MDD) remains challenging due to subtle, heterogeneous anatomical changes.

  We propose uncertainty-gated SHAP-Dirichlet refinement, a post-hoc reliability layer that improves predictions without retraining the classifier.

Extractive summary: sentences quoted from the sources.