Datasets:
| { | |
| "meta_info": { | |
| "case_id": "Physionet2018_02.edf", | |
| "bench_subset": "NeuroBench-Sleep", | |
| "difficult": 2, | |
| "original_dataset": "Physionet2018" | |
| }, | |
| "agent_input": { | |
| "data_path": "data/sleep/Physionet2018_02.edf", | |
| "label_path": "data/sleep/Physionet2018_02.npy", | |
| "instruction": "Please use the provided sleep EEG/PSG file and paired sleep-stage label file to calculate the whole-night Respiratory Disturbance Index (RDI). Count all apnea events, hypopnea events, and respiratory-effort-related arousal (RERA) events across the full recording from the sleep signals. Compute total sleep time (TST) from the label file as the number of non-Wake epochs (N1, N2, N3, or R/REM) multiplied by 30 seconds and converted to hours. RDI is defined as (apnea_count + hypopnea_count + RERA_count) divided by TST in hours, reported as events per hour. Respiratory channels for Physionet2018: ABD = abdominal respiratory effort; CHEST = thoracic respiratory effort; AIRFLOW = airflow; SaO2 = oximetry. Please clearly report the final RDI value." | |
| }, | |
| "eval_config": { | |
| "parser_prompt": "### ROLE\nYou are a precise JSON Extraction Engine for sleep medicine reports.\n\n### TASK\nExtract the whole-night Respiratory Disturbance Index (RDI).\n\n### STRICT CONSTRAINTS (MANDATORY)\n1. Output ONLY a valid JSON object.\n2. Do not include Markdown, explanations, units, or extra fields.\n3. The key must be EXACTLY \"rdi\".\n4. The value must be a FLOAT or null.\n5. If the value is missing or unclear, return {\"rdi\": null}.\n\n### OUTPUT TEMPLATE\n{\"rdi\": <float|null>}", | |
| "metrics": [ | |
| { | |
| "metric_id": "respiratory_disturbance_index_accuracy", | |
| "type": "numeric_check", | |
| "target_key": "rdi", | |
| "weight": 100, | |
| "params": { | |
| "gt_value": 51.391304347826086, | |
| "tolerance": 10.278260869565218 | |
| } | |
| } | |
| ] | |
| } | |
| } |