Classifying Clinical Outcome of Epilepsy Patients with Ictal Chirp Embeddings

2025-08-18

AI Summary

This study addresses the clinical challenge of stratifying epilepsy surgical outcomes using ictal chirp signal features across three dimensions: distinguishing surgical success from failure, classifying cases by procedural difficulty (high vs. low), and identifying “optimal cases”—those achieving success with minimal difficulty. Methodologically, we introduce a novel paradigm integrating t-SNE manifold embedding with SHAP-based interpretability analysis to construct spatially localized feature influence maps, thereby elucidating how chirp attributes drive local clustering in the latent manifold. We evaluate multiple classifiers—including random forest, SVM, logistic regression, and k-NN—under stratified 5-fold cross-validation. Results show that random forest and k-NN achieve 88.8% accuracy in optimal-case detection. The derived interpretability maps clearly visualize outcome-specific spatial separations in the embedded space, substantially enhancing model transparency, clinical trustworthiness, and decision-support utility.

Technology Category

• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Machine Learning: Learning with Manifolds
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

This study presents a pipeline leveraging t-Distributed Stochastic Neighbor Embedding (t-SNE) for interpretable visualizations of chirp features across diverse outcome scenarios. The dataset, comprising chirp-based temporal, spectral, and frequency metrics. Using t-SNE, local neighborhood relationships were preserved while addressing the crowding problem through Student t-distribution-based similarity optimization. Three classification tasks were formulated on the 2D t-SNE embeddings: (1) distinguishing clinical success from failure/no-resection, (2) separating high-difficulty from low-difficulty cases, and (3) identifying optimal cases, defined as successful outcomes with minimal clinical difficulty. Four classifiers, namely, Random Forests, Support Vector Machines, Logistic Regression, and k-Nearest Neighbors, were trained and evaluated using stratified 5-fold cross-validation. Across tasks, the Random Forest and k-NN classifiers demonstrated superior performance, achieving up to 88.8% accuracy in optimal case detection (successful outcomes with minimal clinical difficulty). Additionally, feature influence sensitivity maps were generated using SHAP explanations applied to model predicting t-SNE coordinates, revealing spatially localized feature importance within the embedding space. These maps highlighted how specific chirp attributes drive regional clustering and class separation, offering insights into the latent structure of the data. The integrated framework showcases the potential of interpretable embeddings and local feature attribution for clinical stratification and decision support.

Problem

• Classifying epilepsy patient outcomes using chirp embeddings
• Distinguishing clinical success from failure with minimal difficulty
• Identifying optimal surgical cases through interpretable visualization techniques

Innovation

• t-SNE embeddings for clinical outcome visualization
• Random Forest and k-NN classifiers for optimal detection
• SHAP explanations generating feature influence sensitivity maps