Skip to content

Quickstart

MedpipeClassifier simple example

Use MedpipeClassifier to create a pipeline for a classification task. The pipeline is defined by a TOML configuration file, which is validated when the pipeline is created.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
from pydantic import ValidationError

from medpipe import MedpipeClassifier

try:
    pipe = MedpipeClassifier("default_classifier_config.toml")
    pipe.run()
except ValidationError as err:
    print(err)
    exit()

The configuration file used above is examples/default_classifier_config.toml. It shows everything that can be written in a classification configuration, and each block is described in the configuration guide. Run the example from the examples/ folder, since the paths in the configuration are relative to where the script is run.

MedpipeRegressor simple example

Use MedpipeRegressor for a continuous outcome. The pipeline is defined in the same way, by a TOML configuration file.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
from pydantic import ValidationError

from medpipe import MedpipeRegressor

try:
    pipe = MedpipeRegressor("default_regressor_config.toml")
    pipe.run()
except ValidationError as err:
    print(err)
    exit()

The configuration file is examples/default_regressor_config.toml. It predicts the length of stay and the blood loss with OrdBoostRegressor, which predicts a full distribution of each outcome. The regression page of the guide lists how a regression configuration differs from a classification one.

Results

Each run is saved in a new numbered folder, artifacts/v1, artifacts/v2, and so on. It contains the fitted models, the evaluation results with confidence intervals, the figures, and the configuration and environment of the run. The run folder page describes each file.