API Reference

This is the API reference for the AUCMEDI framework.

Build your state-of-the-art medical image classification pipeline with the 3 AUCMEDI pillars:

Pillars of AUCMEDI

Pillar	Description
#1: input_interface()	Obtaining general information from the dataset.
#2: NeuralNetwork	Building the deep learning model.
#3: DataGenerator	Powerful interface for loading any images/volumes into the model.

A typical AUCMEDI pipeline

# AUCMEDI library
from aucmedi import *

# Pillar #1: Initialize input data reader
ds = input_interface(interface="csv",
                     path_imagedir="dataset/images/",
                     path_data="dataset/classes.csv",
                     ohe=False, col_sample="ID", col_class="diagnosis")
(index_list, class_ohe, nclasses, class_names, image_format) = ds

# Pillar #2: Initialize a DenseNet121 model with ImageNet weights
model = NeuralNetwork(n_labels=nclasses, channels=3,
                       architecture="2D.DenseNet121",
                       pretrained_weights=True)

# Pillar #3: Initialize training Data Generator for first 1000 samples
train_gen = DataGenerator(samples=index_list[:1000],
                          path_imagedir="dataset/images/",
                          labels=class_ohe[:1000],
                          image_format=image_format,
                          resize=model.meta_input,
                          standardize_mode=model.meta_standardize)
# Run model training with Transfer Learning
model.train(train_gen, epochs=20, transfer_learning=True)

# Pillar #3: Initialize testing Data Generator for 500 samples
test_gen = DataGenerator(samples=index_list[1000:1500],
                         path_imagedir="dataset/images/",
                         labels=None,
                         image_format=image_format,
                         resize=model.meta_input,
                         standardize_mode=model.meta_standardize)
# Run model inference for unknown samples
preds = model.predict(test_gen)