- Project start, Jason calls - Sept - Nov 2024
- Stage 1: Initial research - Nov 2024 - January 2025
- Stage 2: Deeper research, large, 12-signal datasets & papers - February - May 2025
- Stage 2B: Develop ResNet + SE model, 12-signal: May 2025
- compare w plain ResNet, 12-signals treat as 12-dim channel inputs.
- compare w simple CNN, but with longer epochs, 2 and 6 classes, then 24-27 classes.
- compare w Transformer-based 12-signal models.
- try 6 and 12 classes for higher accuracy.
- compare datasets from different hospitals, systematic bias?
- Stage 2C: data augmentation - June 2025
(from public datasets to simulate noisy device data)- Gemimi AI deep research mode - develop synthetic augmented data for v1-v6 node location variance, body movement and shirt noise variance, powerline interference.
- Jason, collect abnormal heart data from Pt Reyes? When? Delayed to August or later.
- Alondra 30 recordings 10 sec long, device output - May 2025
-> Fail, data is garbage. Wasted 30 hours to convert machine codes data. - Jason to send me new device ECG recordings - Delayed. When?
- Jason, to send me 1 ECG device for me to test. Delayed indefinitely.
- Stage 3: SciPy poster: June 2025, due June 30, 2025 Monday.
- Stage 4A: Develop Demo WebApp of AI model on hospital dataset: August - October 2025
- Stage 4B: Areteus device testing and AI model refinements: August - October 2025
- Stage 5: Deploy Areteus AI app, live device signals & continuous AI classifications: Nov 2025
- SciPy talk proposal submit - March 5-7, 2025
- Accepted for virtual poster - April 26, 2025
- Build & test ResNet+SE model - May 2025 (used for class project, Stanford Tech 07)
- Submit virtual poster to conference organizers - June 30, 2025 done.
- image file virtual poster, small thumbnail image
- Create a short video, Run more augmented data tests - to July 8, 2025
- Present: July 9, 2025 6-7pm PT, 9-10pm ET, Gather, virtual meeting space.