train_flexconn: Training for Fast Lesion Extraction using Convolutional...

In muschellij2/flexconnr: 'FLEXCONN' Model Wrapped in R

Description

Training for Fast Lesion Extraction using Convolutional Neural Networks (FLEXCONN)

Usage

train_flexconn(atlas_dir, use_t2 = FALSE, patch_size = c(35, 35),
  outdir = NULL, gpu = "gpu", normalize = TRUE, verbose = TRUE)

Arguments

atlas_dir	Atlas directory containing atlasXX_T1.nii, atlasXX_FL.nii.gz, atlasXX_mask.nii.gz, where XX=1,2,3, etc All atlas images must be in axial RAI orientation, or whatever orientation FLAIR has the highest in-plane resolution. Atlas T1 and FLAIR images must be coregistered and have same dimensions.Z
use_t2	should T2 images be used?
patch_size	Patch size, e.g. 35x35 or 31x31 (2D). Patch sizes are separated by x. Note that 2D patches are employed because usually FLAIR images are acquired 2D.
outdir	Output directory where the trained models are written.
gpu	Choice for GPU. Use the integer ID for the GPU. Use "cpu" to use CPU. Can also be NULL.
normalize	Should the images be normalized?
verbose	Print diagnostic messages

Value

A list of filenames

Examples

## Not run: 
library(reticulate)
use_python("python3")

## End(Not run)

muschellij2/flexconnr documentation built on May 14, 2019, 11:13 a.m.