R/num_desc_classifier.R

In ComputationalIntelligenceGroup/hbp-dendrite-arborization-simulation: basal arborization reconstruction and simulation

#' This function compute the features of each node
#'
#' This function compute the features of each node. Thus, for each node defining a neurite a set of features is computed according to node_feature_extractor function
#' @param path is a string indicating the folder where the files representing neurons are saved
#' @param eps is an integer value that represents the level of simplification applied over the neuron
#'
#' @return a dataset with the features of each node. List if features can be found in c_node_feature_extractor
#' @example
#' path<-"/home/universidad/Documents/neuron_data/datos/All"
#' get_features_files(path,60)
#'
#'
#'   #"/home/pablo/Desktop/Simulacion-arborizaciones-basales/Data/HUMANOS_IIIyV/TodosDAT"
get_features_files<-function(path,eps)
{
  factor_var<-c("node_num_descendant","node_order","subtree_max_order","subtree_min_order","subtree_no_bif","subtree_terminals")

  path2files<-file.path(path,list.files(path))
  data<-data.table()
  count<-1
  for(file_path in path2files)
  {
    print(file_path)
    neuron<-neuro_converter(file_path,eps=eps)
    features<-rbindlist(node_feature_extractor(neuron$plain))
    data<-rbind(data,features)
    count<-count+1

  }
  setDF(data)

  data$node_order[as.numeric(as.character(data$node_order))>4]<-4
  for(i in factor_var)
  {
    data[,i]<-as.factor(data[,i])
  }

  #Remove segments of 0 length
  idx_zero<-which(as.numeric(as.character(data$node_num_descendant))>0 & data$desc_length<=0)
  if(length(idx_zero)>0)
  {
    data<-data[-idx_zero,]
  }
  data$desc_length<-log(data$desc_length)

  idx_zero<-which(as.numeric(as.character(data$node_num_descendant))==2 & data$desc_length2<=0)
  if(length(idx_zero)>0)
  {
    data<-data[-idx_zero,]

  }
  data$desc_length2<-log(data$desc_length2)
  return(data)
}



#' This function train a model to distinguish between those nodes that should be terminal, bifurcation or continue the neurite (0, 1 or 2 descendants)
#'
#' This function train a model to distinguish between those nodes that should be terminal, bifurcation or continue the neurite (0, 1 or 2 descendants)
#'
#' @param data is a dataframe which contains the features of each node
#'
#' @return model is an augmented Bayesian network classifier
#'
#' @example

train_num_descendant_model<-function(data)
{
  data<-data[as.numeric(as.character(data$node_num_descendant))<3,]#Remove trifurcations
  attributes(data$node_num_descendant)$levels<-c("0","1","2")
  data<-data[as.numeric(as.character(data$node_order))>0,]#Remove nodes with order 0 because they are deterministic, always continue.

  featured_data<-data[,colnames(data)%in%c("azimuth_angle","compartment_length","distance_to_root","elevation_angle","length_to_brach_root","node_num_descendant","node_order","node_to_brach_root_dist","path_to_root,subtree_box_volume","subtree_length","tortuosity")]

  colnames(featured_data)[which(colnames(featured_data)=="node_num_descendant")]<-"class"
  featured_data$class<-factor(featured_data$class,levels=as.numeric(attributes(table(featured_data$class))$dimnames[[1]]))
  featured_data$node_order<-factor(featured_data$node_order,levels=as.numeric(attributes(table(featured_data$node_order))$dimnames[[1]]))


  model<-learn_BN(featured_data)

  return(model)
}