Setup_For_Keras: Setup geoLet derived objects for deep learning analysis
In kbolab/moddicom: Package for handling DICOM RT files

Provide the output needed for starting a deep learning model using keras:

Setup_For_Keras(path, outcome = NULL, ratio = 2/3, ROIname,
  ROINameNorm = NULL, TrainTest = TRUE, per_patient_sampling = TRUE,
  New_Pixel_Spacing = c(), ROIBitMask = TRUE, TargetSize = "auto",
  VirtualBiopsy = FALSE, VirtualBiopsyRay = c(10, 10),
  VirtualBiopsyMaxNum = 25, backgroud = c("zero", "min", "mean",
  "runif", "rnorm"), bckwidth = 0.5, threshold_filter = 150,
  ROImap = NULL)

`path`	The `path` of cases structured as moddicom usually requires (one subfolder with a DICOM study containing images and a RT structure DICOM file)
`outcome`	A `data.frame` containing in one column the same IDs of patients listed in `path`, usually corresponding to the folder names, and a column with the given outcome
`ratio`	a number between 0 and 1 giving the ratio between training set and verification set (default 2/3)
`ROIname`	a `character` vector containing the ROI(s) name(s) used for extracting images from studies
`ROINameNorm`	a `character` vector containing the ROI(s) name(s) used for used for normalizing the values, if `NULL` no normalization is applied
`TrainTest`	`logical` value (default is `TRUE`) that gives the outcome already splitted for training and test.
`per_patient_sampling`	a `logical` value (default is `TRUE`) that sets the sampling for training and testing set according the patients IDs. If `FALSE` the sampling is obtained by images slices.
`New_Pixel_Spacing`	a `numeric` vector, length must be 2 values, that sets the new pixel spacing for all the images in order to homogenize the spatial resolution among all patients.
`ROIBitMask`	a `logical` value, default `TRUE`, return only the voxels within the ROI, if `FALSE` returns the whole VoxelCube.
`TargetSize`	default value is `'auto'` that calculates the size in rows and columns of final array as mean of the single VoxelCubes sizes, otherwise user can set a numerical vector of length = 2, that sets number of rows and columns manually.
`VirtualBiopsy`	`logical` value, default is `FALSE`. If `TRUE` returns for each patient returns a number maximum number of `VirtualBiopsyMaxNum` of small sampling images with size `VirtualBiopsyRay * 2 + 1`.
`VirtualBiopsyRay`	a `numeric` vector of length 2, giving the width os extension of Virtual Biopsy around the centroid. Final dimension of biopsy is `VirtualBiopsyRay * 2 + 1`.
`VirtualBiopsyMaxNum`	a `integer` value setting the maqximum number of Virtual Biopsies to be sampled.
`backgroud`	the value of the pixel surrounding the ROI in the background: `'zero'` is 0; `'min'` is the minimum value in the ROI; `'mean'` is the mean value in the ROI; `'runif'` is a random uniform distribution between `min` and `max` values in the ROI; `'rnorm'` is a random normal distrbution with `mean` is the mean value of ROI, `sd` is the standard distribution of ROI values.
`bckwidth`	a `numeric` value between 0 and 1: the width in the range of ROI values for `runif` background option or the multiplier of `sd` for `rnorm` background option.
`threshold_filter`	The threshold number of pixels per slice `NA` below which images are deleted. Default is 150 pixels.
`ROImap`	If available, the result of a map.ROI function, as an object of class ROImap given by moddicom package.

ROIList: The list of ROIs as achieved by moddicom geoLet$getROIVoxels( ) function.
cases_data.frame: A data.frame containing the patients IDs (pt_name), each ROI_name mapped in each patient and the pt_outcome that reports the patient outcome.
images_array: The final array of selected images structured keras requires.
keras_outcome: A numeric vector as required by keras for the classification of outcome
df_outcome: A data.frame containing the labels of the outcome as provided by user and the corresponding labels (numeric values) used for modeling in keras.
cases_data.frame: A data.frame containing the patients IDs (pt_name), each ROI_name mapped in each patient and the pt_outcome that reports the patient outcome.
images_array_train: The final array of selected images structured as keras requires. This is the training dataset. Available if TrainTest=TRUE.
images_array_test: The final array of selected images structured as keras requires. This is the test dataset. Available if TrainTest=TRUE.
keras_outcome_train: A numeric vector as required by keras for the classification of outcome. This is the outcome for the training dataset. Available if TrainTest=TRUE.
keras_outcome_test: A numeric vector as required by keras for the classification of outcome. This is the outcome for the test dataset. Available if TrainTest=TRUE.
df_outcome: A data.frame containing the labels of the outcome as provided by user and the corresponding labels (numeric values) used for modeling in keras.
total_data.frame_train: The data.frame with all cases and slices corresponding to the images in the images_array_train array. Available if TrainTest=TRUE.
total_data.frame_test: The data.frame with all cases and slices corresponding to the images in the images_array_test array. Available if TrainTest=TRUE.