This technology comprises a unique dataset of 661 transcribed audio recordings collected from ICU rounds. These recordings were processed using automated transcription via Amazon Medical Transcribe and subsequently underwent thorough de-identification to protect patient privacy. The dataset captures real-world medical interactions, ensuring clinical details are accurately transcribed and securely anonymized to comply with privacy standards.
Description
This dataset is differentiated by its rigorous processing methodology and the collaborative input from experts at UPMC and the University of Pittsburgh. Its unique combination of automated transcription technology and systematic de-identification provides a rare resource that offers both depth of clinical insight and strict data protection.
Applications
- Medical NLP model training
- Clinical decision support system
- ICU outcome predictive analytics
- Medical transcription quality control
Advantages
- Enables research using a unique, de-identified ICU dataset while ensuring patient privacy.
- Demonstrates the effective application of automated transcription technology in a medical setting.
- Provides a systematic approach to data collection and de-identification in real-world clinical scenarios.
IP Status
Copyright
