Hi FSL community,
We have hand-classified 35 participant’s ICA components from MELODIC. The components were derived from a task-fMRI and these participants are healthy adults that span the age range of the 20s to 90s.
Q1. We tried to train the hand labels using the command:
`/usr/local/fix/fix -t Training -l ${base_dir}/001/ICA/Run1/run1_4d_ICA.ica `${base_dir}/002/ICA/Run1/run1_4d_ICA.ica … ${base_dir}/035/ICA/Run1/run1_4d_ICA.ica`
We get the following error:
copy true labels and features into training directory
FIX features not already generated, so creating them now.
No valid Melodic output directory specified
cp: ${base_dir}/035/ICA/Run1/run1_4d_ICA.ica\r/hand_labels_noise.txt: No such file or directory
cp: ${base_dir}/035/ICA/Run1/run1_4d_ICA.ica\r/fix/features.csv
However, when using the subsequent command with a space at the end of the command:
`/usr/local/fix/fix -t Training -l ${base_dir}/001/ICA/Run1/run1_4d_ICA.ica ${base_dir}/002/ICA/Run1/run1_4d_ICA.ica … ${base_dir}/035/ICA/Run1/run1_4d_ICA.ica `
We get the following error:
copy true labels and features into training directory
FIX features not already generated, so creating them now.
No valid Melodic output directory specified
cp: \r/hand_labels_noise.txt: No such file or directory
cp: \r/fix/features.csv: No such file or directory
It appears that adding a space allows the command to find the last subject’s directory (i.e., 035) but it is still trying to find another subject at the end? Or is there something in the way we have written the command that caused this? If not, is there a solution or workaround for this issue?
Q2. Performing either command (i.e., with and without a space at the end) does provide the Training_LOO_results file with 35 rows. However, using the command without a space results in classifying all the results as signal (all TPR results are 100 and TNR are 0) while using the command with a space does provide us with some different values (unique TPR results but all TNR are 0).
Without space at the end of the command:
mean
100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
75.0 75.0 75.0 75.0 75.0 75.0 75.0 75.0
With space at the end of the command:
mean
55.3 50.4 44.1 39.8 35.3 32.1 28.2 24.3
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
41.5 37.8 33.1 29.8 26.4 24.1 21.2 18.2
Given the errors of finding the last subject (without a space) or a blank subject (with a space), We’re not sure if the LOO results are correct even with the space at the end of the command given the low TPR and all zero TNR. Is it including the missing last subject directory into the LOO analysis? However, if it is correct, do you have any directions on improving the TP and TN rates?
Q3. Is there a way to determine which algorithm is explicitly being used in the LOO analysis?
Q4. Separately, we also tried classifying the same ICA components using the Standard trained weight file using the command:
`/usr/local/fix/fix -c ${base_dir}/001/ICA/Run1/run1_4d_ICA.ica /usr/local/fix/training_files/Standard.RData 20`
However, there is no classification output file (i.e., fix4melview .txt file) so that we can subsequently run the `fix -C` command. Looking within the .fix_2b_predict.log file, there is an error using the ctreefit() function:
Error in ctreefit(object = object, controls = controls, weights = weights, :
no slot of name "remove_weights" for this object of class "TreeGrowControl"
Calls: eval -> eval -> <Anonymous> -> ctreefit
Execution halted
Similarly, is there an error in the way we have coded this? If not, is there a solution/workaround for this issue?
# setup
FSL v6.0.1
FSL FIX v1.06
macOS Mojave (10.14.5)
R v3.5.3
kernlab v0.9.27
ROCR v1.0.7
class v7.3.15
party v1.3.3
e1071 v1.7.2
randomForest v4.6.14
MATLAB 2019a
Any direction/help would be greatly appreciated. Please also let me know if any other information is needed to help solve this.
Thank you,
Ekarin
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
