Baseline

[BobAubouin]

Implement the baseline for the task, see this example:

• [x] Use mean arterial pressure value to predict IOH
• [x] Compute the dedicated metrics

[BobAubouin]

The baseline should be this paper: Jacquet-Lagrèze, Matthias; Larue, Antoine; Guilherme, Enrique; Schweizer, Rémi; Portran, Philippe; Ruste, Martin; Gazon, Mathieu; Aubrun, Frédéric; Fellahi, Jean-Luc. Prediction of intraoperative hypotension from the linear extrapolation of mean arterial pressure. European Journal of Anaesthesiology 39(7):p 574-581, July 2022. | DOI: 10.1097/EJA.0000000000001693

[BobAubouin]

The metric to evaluate the model should be something like: When the predictor gives an alarm, how often does hypotension actually occur, e.g., 3 to 10 min later (i.e., positive predictive value); and when hypotension occurs, was there a predicted alarm 3 to 10 min earlier (i.e., sensitivity)? The range of time might be discuss but seems faire for a 5 minutes ahead prediction

[BobAubouin]

Testing on data, it appears that diastolic pressure is a better predictor of hypotension than mean arterial pressure. SO it should be the baseline.

[BobAubouin]

It might be possible that a simpler question to answer for the metric is: After an alarm rise, how often does hypotension actually occurs in the next 10 minutes. It will be easier to explain in the paper.

[BobAubouin]

The metric to evaluate the model should be something like: When the predictor gives an alarm, how often does hypotension actually occur, e.g., 3 to 10 min later (i.e., positive predictive value); and when hypotension occurs, was there a predicted alarm 3 to 10 min earlier (i.e., sensitivity)? The range of time might be discuss but seems faire for a 5 minutes ahead prediction

This have not been implemented (might be a good thing to check)