Manual validation and correction workflow for OCR-extracted ventilation data during cardiac arrest (JupyterHub interface)

Video showing the data extraction and manual validation/correction workflow using a custom-built interface based on a JupyterHub notebook, here applied to artificial ventilation data during cardiac arrest. Because the device used in this example did not provide direct data export, a conventional webcam was used to record the device display and optical character recognition (OCR) was applied to extract values frame-by-frame. The interface supports plausibility checks and interactive review by displaying extracted values alongside the corresponding video frames and short rolling time windows (e.g., the preceding and subsequent values) with dynamic plots. Users can validate and correct flagged values through simple button-based interactions, with all edits and decisions logged to support transparency and reproducibility. The video is played back at an increased speed.