src.algorithms.identify package

Submodules

src.algorithms.identify.trained_model module

algorithms.identify.trained_model

An API for a trained identification model to make predictions for where the centroids of nodules are in the DICOM image.

src.algorithms.identify.trained_model.predict(dicom_path)

Predicts centroids of nodules in a DICOM image.

Given an iterator of DICOM objects, this method will:

1. load the identification model from its serialized state
2. pre-process the dicom into whatever format the identification model expects
3. return centroids with a probability that each centroid is a nodule (as opposed to not a nodule)

Note

This model doesn’t detect whether or not a nodule is cancerous, that is done in the classify model.

Parameters:dicom_path (str) – a path to a DICOM image Returns:

a list of centroids in the form::

{‘x’: int,

‘y’: int, ‘z’: int, ‘p_nodule’: float}

Return type:list(dict)

Module contents