revdat: Design traits of different studies

In ayanglab/How-to-Perform-Technical-Systematic-Review-And-Meta-Analysis-Tutorial: What the Package Does (One Line, Title Case)

Description

a dataframe with the following columns:

GAN.based

logical vector, whether the study uses GAN

PI

logical vector, whether the study uses parallel imaging

acceleration

string, acceleration ratios across all experimental conditions

architecture

string, deep neural network architecture

augmentation

string, data augmentation technique

better

logical, whether the performance exceeds all the comparison methods listed in the study

category

string (factor), whether the study is 'unrolled optimisation' or 'end-to-end'

channel.merging

string, how channel/coil merging is performed if the study uses parallel imaging

channel.number

string, number of channels in the input image

comparison

string, comparison methods

computation.time

string, GPU computation time per reconstructed image in seconds

country

string, the country of the institute where the first author of the study is affliated to

data.consistency

logical, whether data consistency layer as in DC-CNN is used

dataset

string, public dataset in the study

dimension

numeric, the input dimension of the image, can be 2, 3 or 4

ground.truth

string, how the ground truth is obtained

input.domain

string, the input domain of the image

input.size

string, the size of the input

limitation

string, limitation of the study reported by the authors

loss

string, loss function

mask

string, undersampling mask

metric

string, metric used to assess reconstruction performance

modality

string, the imaging sequence, e.g. T1, T2 weighted

name

string, the name of the model proposed

novelty

string, novelty mentioned by the authors

open.source

string, the url to the source code

optimiser

string, the type of optimiser used

package

string, deep learning framework, e.g. tensorflow, pytorch

parameter.number

string, the number of parameters

region

string, anatomical region in the image dataset, e.g. brain, knee

residual

logical, whether residual learning is applied

result

string, descriptions of reconstruction results

result.table

string, whether result table is provided for quantitative analysis

spatial.domain

string, whether spatial domain was used

strength

string, the strength of the model

testing.data.pathology

logical, whether the testing dataset contains pathological features

testing.data.size

string, sample size of the testing dataset

testing.mode

string, mode of testing, retrospective or prospective

testing.number

numeric, number of scans used in testing

training.data.pathology

logical, whether the training dataset contains pathological features

training.data.size

string, sample size of the training dataset

training.number

numeric, number of scans used in training

unet.like

logical, whether unet-like architecture is used

unrolled

logical, whether unrolled optimisation is used

validation.data.size

string, sample size of the validation dataset

zero.filled

logical, whether zero-filling is used as a comparison method

others

string, other information

ID

string, ID for each study used in this review paper

year

string, publication year according to the 'publish or perish' website

journal

string, in which journal the paper is found

supervised

logical, whether supervised learning is used

institute

logical, the institute which first author of the study is affliated to

full_GT

logical, whether fully sampled ground truth is available

adam

logical, whether adam optimiser is used

central_mask

logical, whether the mask samples the center of the image fully

max_acc

numeric, maximal acceleration ratio

MSE_loss

logical, whether mean squared error loss is used

L1_loss

logical, whether L1 loss is used

L2_loss

logical, whether L2 loss is used

complex_inp

logical, whether the input is complex signal

public_code

logical, whether the code is published

public_data

logical, whether the data is published

reproducibility

string, reproducibility of the study

autoencoder

logical, whether autoencoder is used

MLP

logical, whether multi-layer perceptron is used

stack_conv

logical, whetther stacked convolution is used

model_structure

string, architecture of the model

cluster

string, raw cluster output from mclust::Mclust

new_cluster

string, annotation of the raw cluster

NB: If a particular study uses multiple methods for a particular trait, each method is separated by a comma. A bracket may follow each method describing the dataset that the method is applied in the study.

Usage

revdat

Format

An object of class data.frame with 92 rows and 69 columns.

Author(s)

Yutong Chen

ayanglab/How-to-Perform-Technical-Systematic-Review-And-Meta-Analysis-Tutorial documentation built on Feb. 7, 2022, 12:45 a.m.