ERIA: An Instruction-Optimized Agent for Clinical EEG Reporting
Organizations: Department of Computer Science & Engineering, University of Minnesota Twin Cities · Department of Bioengineering, University of Illinois Urbana-Champaign · Department of Genetics, Cell Biology & Development, University of Minnesota Twin Cities
Abstract
Clinical EEG interpretation requires integrating technical quality, background and behavioral states, interictal and ictal abnormalities, and overall clinical context across long recordings. We introduce ERIA (EEG Reporting and Instruction-Optimized Agent), a locally deployable framework in which a single 27B language-model controller analyzes whole-recording EEGs by orchestrating specialized parametric and signal-processing tools. ERIA combines learned tools for background abnormality, sleep-wake state, seizure-like activity, and interictal epileptiform discharges with deterministic spectral, amplitude, symmetry, montage, and signal-quality analyses. Rather than fine-tuning the controller or using multi-agent orchestration, ERIA keeps the controller weights, specialist tools, and safety constraints fixed and optimizes only its natural-language analysis policy with GEPA (Genetic-Pareto), using feedback from expert-written EEG reports. GEPA improves physician-aligned whole-recording report quality over the raw policy, and the optimized policy transfers to external EEG benchmarks. On CerebraGloss-Bench and BrainBench Foundational Analysis, ERIA remains competitive with the models evaluated in the original benchmarks. These results show that specialist tools and language-level policy optimization can extend the clinical and general EEG capabilities of a single locally executable model while preserving local inference and traceable tool use.
Figures & tables
| Tool | Method / source | Scale | Returned evidence |
| Deterministic / inspection tools | |||
| recording_summary | Deterministic metadata inspection | recording; all ch. | metadata, duration, sampling rate, channels |
| assess_signal_quality | Deterministic quality analysis | interval; selected/all ch. | invalid, flat, or clipped channels; quality summary |
| inspect_preprocessing | Deterministic preprocessing inspection | recording; all ch. | preprocessing and transformed-signal statistics |
| inspect_montage | Deterministic montage derivation | recording; 22-ch TCP | montage feasibility; derived channels |
| amplitude_statistics | Deterministic amplitude features | 10–60 s; 1 ch. | mean abs. amplitude, RMS, peak-to-peak, extrema |
| Measure | Raw | GEPA |
|---|---|---|
| A. Clinical report quality | ||
| Validation aggregate score | 0.5099 | 0.6277 |
| Protected test aggregate score | 0.5052 | 0.6024 |
| B. Validation-set execution behavior | ||
| sleep_wake_staging calls | 3.42 | 7.12 |
| Cases sampling early, middle, and late recording portions | 40% | 98% |
| Benchmark | Metric | Raw | GEPA | Reference |
|---|---|---|---|---|
| CerebraGloss-Bench | MCQ accuracy (%) | 71.11 | 71.11 | 80.00 |
| CerebraGloss-Bench | Description ROUGE-1 | 30.70 | 30.17 | 44.19 |
| CerebraGloss-Bench | Conversational QA | 3.60 | 3.87 | 4.13 |
| BrainBench FA | Weighted score | 85.86 | 86.89 | 84.97 |
Electroencephalography
Electroencephalography (EEG) studies brain-generated electrical activity to understand cognition and support diagnosis, monitoring, and brain-computer interfaces.
See all 200 papers