Feature demo: web application for YASA

[raphaelvallat]

Discussed in #185

^{Originally posted by bartromb December 14, 2024}
Hi everyone,

I’ve developed a web application that integrates the excellent yasa library into a user-friendly web platform. The application, named YASA Flaskified, is built with Flask, Gunicorn, and Nginx. I currently use it on internal servers as a quality control step after manual scoring. (I am a pulmonologist and somnologist.)

You can find it on my GitHub repository: YASA Flaskified.

I have used only Somnomedics EDF files so far, but the script attempts to automatically identify EEG channels, so I believe it is usable for other EDF files as well. It generates both a hypnogram PDF and a CSV output for further analysis. You can see a showcase on sleepai.be and sleepai.eu, though due to bandwidth constraints, only admin users can register new accounts. I can provide a test login and password upon request.

I want to thank Raphael for the fantastic yasa library—it has been incredibly useful for this project!
Contributions and feedback are more than welcome, and I’m happy to assist with any issues you encounter during deployment on a server.

Looking forward to hearing your thoughts!

Best regards,
Bart

[raphaelvallat]

Hi @bartromb,

Thanks so much for your work on this. Following up on our email conversation, I tried this on an example EDF file (not from a Somnomedics device), and it did work well in outputting the YASA-scored hypnogram and plot (see attached).

yasa_example_night_young.edf_hypnogram.pdf

yasa_example_night_young.edf.csv

I do have a few questions for you:

• Intended use of this web app: can you confirm that the sleepai.be and sleepai.eu links are only provided for demo — the real intended use is that people would set up their own domain-based or local server, correct?

• It would be great to have the ability to choose the EEG, EOG and EMG channels that are used for sleep scoring, as well as specify the metadata (demographics). There could be an automatic detection (as implemented now) and a manual option.

• Do you see this expanding to other functions in YASA beyond sleep staging? e.g. spindles, slow-waves detection, etc.

Thanks again,
Raphael

[bartromb]

Hi Raphael,

Thank you for trying out the web app and for sharing your feedback. I’m glad to hear it worked well with your example EDF file and successfully generated the hypnogram and CSV outputs.

To address your questions:

Channel Selection and Metadata Specification
The web app has been updated to include the ability to select specific EEG, EOG, and EMG channels for sleep scoring. This feature complements the existing automatic detection and allows for greater flexibility. While metadata such as patient demographics is not yet directly modifiable through the interface, I can adapt the app further to accommodate this if needed.
I have updated the github repository.

Intended Use
Yes, you are correct that the links to sleepai.be and sleepai.eu are intended as showcases for demonstration purposes. The recommended use is for users to deploy the application locally, either on a real or virtual server, to achieve the best performance. The deploy.sh install the server automatically.

Expanding to Other YASA Functions
Expanding to additional YASA functionalities, such as spindle and slow-wave detection, is a great idea and something I can implement if there’s demand. The modular design of the app makes it feasible to incorporate these features in future updates.

Please let me know if you have any other suggestions or requirements. I’d be happy to adapt the app further to align with your vision for YASA's broader use cases.

Best regards,
Bartromb

[raphaelvallat]

Thanks @bartromb — great that you added the channel selection!

I think we can add the link to YASA Flaskified in both the README of YASA and the FAQ. If that's okay with you, feel free to submit a PR.

Thanks,
Raphael

[bartromb]

Hi Raphaël, Thank you! I’ve drafted this description—feel free to adjust it if needed: **YASA Flaskified:** A web-based application leveraging the YASA library for EEG-based sleep analysis. It provides: - **User Authentication:** Secure login for authorized users with password encryption. - **File Management:** An intuitive upload system for `.edf` EEG files, with support for multiple uploads. - **Background Processing:** Efficient processing of EEG data using Redis-backed workers. - **Results Management:** Automatic generation of hypnograms (PDF) and sleep stage data (CSV), viewable and downloadable through the app. - **Monitoring:** Real-time status tracking of uploaded jobs. **Easily Deployable:** YASA Flaskified includes a deployment script, making it easy to install on local or domain-based servers, whether real or virtual. Deployment takes just a few minutes, requiring minimal setup. **Showcases:** Live deployments are available at: - [sleepai.be](http://sleepai.be) - [sleepai.eu](http://sleepai.eu) _User accounts for these showcases can be created on request._ ### What technologies power YASA Flaskified? YASA Flaskified is built on a robust and modern software stack to ensure reliability and performance: - **YASA Library:** At the core of the application, YASA (Yet Another Spindle Algorithm) enables advanced EEG-based sleep staging and hypnogram generation. Kudos to Raphaël Vallat, the developer of YASA, for his significant contributions to sleep science. - **Flask:** A lightweight Python web framework that powers the web application, providing routing, templating, and session management. - **Gunicorn:** A WSGI HTTP server used to serve the Flask application in production environments with support for multiple workers and robust request handling. - **Redis:** A fast, in-memory data store used for managing background task queues and caching. - **RQ (Redis Queue):** A Python library that handles long-running background tasks, such as processing EEG files and generating results. - **SQLAlchemy:** A database toolkit and ORM (Object-Relational Mapper) for managing user accounts and application data. - **Matplotlib & Pandas:** Used for creating hypnograms and processing sleep stage data from EEG files. These technologies work together to provide a scalable, efficient, and user-friendly application for EEG sleep analysis. Explore the app and source code: [YASA Flaskified GitHub Repository]( https://github.com/bartromb/YASAFlaskified).

…

On Sun, Dec 22, 2024 at 3:19 PM Raphael Vallat ***@***.***> wrote: Thanks @bartromb <https://github.com/bartromb> — great that you added the channel selection! I think we can add the link to YASA Flaskified <https://github.com/bartromb> in both the README <https://github.com/raphaelvallat/yasa/blob/master/README.rst#how-do-i-get-started-with-yasa> of YASA and the FAQ <https://github.com/raphaelvallat/yasa/blob/master/docs/faq.rst#sleep-staging>. If that's okay with you, feel free to submit a PR. Thanks, Raphael — Reply to this email directly, view it on GitHub <#186 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BNWMDUKAHMUIYI6KBIG6HRT2G3C6BAVCNFSM6AAAAABT645A5WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNJYGQ3TCOJTGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>