Identifying Breaths in Baby Ventilator Data

Part two of a talk at PyDataLondon.

Transcript

1. Identifying Breaths in Baby Ventilator Data (with Dr. Gusztav Belteki) PyDataLondon 2017 License: CC By Attribution Ian Ozsvald @IanOzsvald ModelInsight.io Giles Weaver @GilesWeaver
2. Ian.Ozsvald@ModelInsight.io PyDataLondon Meetup 2017 @gilesweaver @IanOzsvald Ventilator Traces • 20 seconds, 22 breaths • Strong spontaneous breathing by patient • Spontaneous breaths - peak in flow only (no assistance from ventilator) • Mandatory, synchronised breaths - peak in flow and pressure (limited assistance from ventilator)
3. Ian.Ozsvald@ModelInsight.io PyDataLondon Meetup 2017 @gilesweaver @IanOzsvald Gold Standard Annotation • Why? We need accurate manual annotations to evaluate our automated annotation methods against • How? • Manually in spreadsheet. No thanks! • Bokeh + ipywidgets based tool. Rough and ready, but effective. • ipywidgets developers have been helpful
4. • Zoom • Slide timestamp • Capture timestamp • Pan • Repeat
5. Ian.Ozsvald@ModelInsight.io PyDataLondon Meetup 2017 @gilesweaver @IanOzsvald Zoomed Annotation View
6. Ian.Ozsvald@ModelInsight.io PyDataLondon Meetup 2017 @gilesweaver @IanOzsvald Completed Annotation
7. Ian.Ozsvald@ModelInsight.io PyDataLondon Meetup 2017 @gilesweaver @IanOzsvald Start of Breath Algorithm Simple algorithm using scipy.signal.find_peaks_cwt •Detect peaks in flow •Detect troughs in absolute flow • abs(df['flow']) * -1 •Find trough preceding peak
8. Ian.Ozsvald@ModelInsight.io PyDataLondon Meetup 2017 @gilesweaver @IanOzsvald Who Are We? • Ian - “Industrial Data Scientist” for 15 yrs • Giles - Bioinformatician turned Data Sci.