Cross-Sample Augmented Test-Time Adaptation for Personalized Intraoperative Hypotension Prediction

2025-12-12

AI Summary

Intraoperative hypotension (IOH) events are sparse and highly patient-specific, rendering conventional test-time adaptation unreliable due to insufficient or non-representative adaptation data. Method: We propose a cross-sample augmentation framework for test-time adaptation. Its core innovations include (i) a novel cross-patient IOH sample retrieval mechanism that combines K-Shape coarse-grained clustering with fine-grained semantic similarity matching to construct robust, representative adaptation sets; and (ii) a dual self-supervised learning objective—masked time-series reconstruction and retrospective sequence prediction—to enhance modeling of sparse, heterogeneous IOH patterns. The framework seamlessly integrates state-of-the-art foundation models (e.g., TimesFM, UniTS). Results: On VitalDB under zero-shot evaluation, recall and F1 improve by 7.46% and 5.07%, respectively; under fine-tuning, gains reach 1.33% and 1.13%. The method significantly improves model robustness and generalizability across diverse surgical cohorts.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Intraoperative hypotension (IOH) poses significant surgical risks, but accurate prediction remains challenging due to patient-specific variability. While test-time adaptation (TTA) offers a promising approach for personalized prediction, the rarity of IOH events often leads to unreliable test-time training. To address this, we propose CSA-TTA, a novel Cross-Sample Augmented Test-Time Adaptation framework that enhances training by incorporating hypotension events from other individuals. Specifically, we first construct a cross-sample bank by segmenting historical data into hypotensive and non-hypotensive samples. Then, we introduce a coarse-to-fine retrieval strategy for building test-time training data: we initially apply K-Shape clustering to identify representative cluster centers and subsequently retrieve the top-K semantically similar samples based on the current patient signal. Additionally, we integrate both self-supervised masked reconstruction and retrospective sequence forecasting signals during training to enhance model adaptability to rapid and subtle intraoperative dynamics. We evaluate the proposed CSA-TTA on both the VitalDB dataset and a real-world in-hospital dataset by integrating it with state-of-the-art time series forecasting models, including TimesFM and UniTS. CSA-TTA consistently enhances performance across settings-for instance, on VitalDB, it improves Recall and F1 scores by +1.33% and +1.13%, respectively, under fine-tuning, and by +7.46% and +5.07% in zero-shot scenarios-demonstrating strong robustness and generalization.

Problem

• Predicts intraoperative hypotension with personalized patient adaptation
• Addresses unreliable training due to rare hypotension event occurrences
• Enhances model adaptability to rapid intraoperative physiological dynamics

Innovation

• Cross-sample augmentation for test-time training data
• Coarse-to-fine retrieval using clustering and semantic similarity
• Self-supervised masked reconstruction and sequence forecasting signals