# Towards trustworthy seizure onset detection using workflow notes

arXiv: https://arxiv.org/abs/2306.08728 · HTML: https://mlsanity.com/paper/2306.08728/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 49 references that are in the map.
- First posted on arXiv 14 Jun 2023.

## Builds on (most cited first, 19 of 19)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [SGDR: Stochastic Gradient Descent with Warm Restarts](https://mlsanity.com/paper/1608.03983/) (2016)
- [Efficiently Modeling Long Sequences with Structured State Spaces](https://mlsanity.com/paper/2111.00396/) (2021)
- [Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers](https://mlsanity.com/paper/2110.13985/) (2021)
- [Confounding variables can degrade generalization performance of radiological deep learning models](https://mlsanity.com/paper/1807.00431/) (2018)
- [Data Programming: Creating Large Training Sets, Quickly](https://mlsanity.com/paper/1605.07723/) (2016)
- [Hidden Stratification Causes Clinically Meaningful Failures in Machine Learning for Medical Imaging](https://mlsanity.com/paper/1909.12475/) (2019)
- [The Temple University Hospital Seizure Detection Corpus](https://mlsanity.com/paper/1801.08085/) (2018)
- [Self-Supervised Graph Neural Networks for Improved Electroencephalographic Seizure Analysis](https://mlsanity.com/paper/2104.08336/) (2021)
- [Effectively Modeling Time Series with Simple Discrete State Spaces](https://mlsanity.com/paper/2303.09489/) (2023)
- [Deep Learning Predicts Hip Fracture using Confounding Patient and Healthcare Variables](https://mlsanity.com/paper/1811.03695/) (2018)
- [Learning Representations from EEG with Deep Recurrent-Convolutional Neural Networks](https://mlsanity.com/paper/1511.06448/) (2015)
- [Machine Learning for Predicting Epileptic Seizures Using EEG Signals: A Review](https://mlsanity.com/paper/2002.01925/) (2020)
- [Cross-Modal Data Programming Enables Rapid Medical Machine Learning](https://mlsanity.com/paper/1903.11101/) (2019)
- [Neural Memory Networks for Seizure Type Classification](https://mlsanity.com/paper/1912.04968/) (2019)
- [Neonatal seizure detection from raw multi-channel EEG using a fully convolutional architecture](https://mlsanity.com/paper/2105.13854/) (2021)
- [Automatic Analysis of EEGs Using Big Data and Hybrid Deep Learning Architectures](https://mlsanity.com/paper/1712.09771/) (2017)
- [Convolutional neural network for detection and classification of seizures in clinical data](https://mlsanity.com/paper/1903.08864/) (2019)
- [Gated Recurrent Networks for Seizure Detection](https://mlsanity.com/paper/1801.02471/) (2018)

## Built on by (most cited first, 4 of 4)

- [Generalizable deep learning for photoplethysmography-based blood pressure estimation -- A Benchmarking Study](https://mlsanity.com/paper/2502.19167/) (2025)
- [An Automated Classifier of Harmful Brain Activities for Clinical Usage Based on a Vision-Inspired Pre-trained Framework](https://mlsanity.com/paper/2507.08874/) (2025)
- [Revisiting Performance Claims for Chest X-Ray Models Using Clinical Context](https://mlsanity.com/paper/2509.19671/) (2025)
- [A Neurosymbolic Framework for Interpretable Skeleton-Based Seizure Detection via Concept-Driven Logical Reasoning](https://mlsanity.com/paper/2606.21252/) (2026)

## Abstract (arXiv)

A major barrier to deploying healthcare AI models is their trustworthiness. One form of trustworthiness is a model's robustness across different subgroups: while existing models may exhibit expert-level performance on aggregate metrics, they often rely on non-causal features, leading to errors in hidden subgroups. To take a step closer towards trustworthy seizure onset detection from EEG, we propose to leverage annotations that are produced by healthcare personnel in routine clinical workflows -- which we refer to as workflow notes -- that include multiple event descriptions beyond seizures. Using workflow notes, we first show that by scaling training data to an unprecedented level of 68,920 EEG hours, seizure onset detection performance significantly improves (+12.3 AUROC points) compared to relying on smaller training sets with expensive manual gold-standard labels. Second, we reveal that our binary seizure onset detection model underperforms on clinically relevant subgroups (e.g., up to a margin of 6.5 AUROC points between pediatrics and adults), while having significantly higher false positives on EEG clips showing non-epileptiform abnormalities compared to any EEG clip (+19 FPR points). To improve model robustness to hidden subgroups, we train a multilabel model that classifies 26 attributes other than seizures, such as spikes, slowing, and movement artifacts. We find that our multilabel model significantly improves overall seizure onset detection performance (+5.9 AUROC points) while greatly improving performance among subgroups (up to +8.3 AUROC points), and decreases false positives on non-epileptiform abnormalities by 8 FPR points. Finally, we propose a clinical utility metric based on false positives per 24 EEG hours and find that our multilabel model improves this clinical utility metric by a factor of 2x across different clinical settings.