Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add elbow-less minimal decision tree #1131

Closed
wants to merge 1 commit into from
Closed

Add elbow-less minimal decision tree #1131

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
148 changes: 148 additions & 0 deletions tedana/resources/decision_trees/elbowless_minimal.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,148 @@
{
"tree_id": "elbowless_minimal_decision_tree",
"info": "first version of minimal decision tree without elbows",
"report": "The elbow-less minimal decision tree \\citep{tedana_decision_trees} is a simplified version of the MEICA decision tree \\citep{kundu2013integrated,dupre2021te} without many criteria that do not rely on kappa and rho thresholds, including any criteria based on elbows calculated across components. ",
"necessary_metrics": [
"kappa",
"rho",
"countsigFS0",
"countsigFT2",
"dice_FS0",
"dice_FT2",
"signal-noise_t",
"variance explained"
],
"intermediate_classifications": ["provisionalaccept", "provisionalreject"],
"classification_tags": ["Likely BOLD", "Unlikely BOLD", "Low variance"],
"_comment": "More information on the minimial decision tree and how it differs from other options is at https://tedana.readthedocs.io/en/stable/included_decision_trees.html. Descriptions of the metrics used are in desc-tedana_metrics.json, which is ouputted when this tree is run",
"nodes": [
{
"functionname": "manual_classify",
"parameters": {"new_classification": "unclassified", "decide_comps": "all"},
"kwargs": {
"clear_classification_tags": true,
"dont_warn_reclassify": true
},
"_comment": "All components are initially labeled as 'unclassified'."
},
{
"functionname": "dec_left_op_right",
"parameters": {
"if_true": "rejected",
"if_false": "nochange",
"decide_comps": "all",
"op": ">",
"left": "rho",
"right": "kappa"
},
"kwargs": {"tag_if_true": "Unlikely BOLD"},
"_comment": "The first four steps are for rejecting components that very unlikely to have substantial T2* signal. Any components with rho greater than kappa are rejected. Higher rho than kappa means that the component better fits the TE-independence (S0) model than the TE-dependence (T2*) model."
},
{
"functionname": "dec_left_op_right",
"parameters": {
"if_true": "rejected",
"if_false": "nochange",
"decide_comps": "all",
"op": ">",
"left": "countsigFS0",
"right": "countsigFT2"
},
"kwargs": {
"left2": "countsigFT2",
"op2": ">",
"right2": 0,
"tag_if_true": "Unlikely BOLD"
},
"_comment": "Any components with more voxels that are significant based on the S0 model's F-statistics than the T2* model's are rejected, as long as there is at least one significant voxel for the T2 model."
},
{
"functionname": "calc_median",
"parameters": {
"decide_comps": "all",
"metric_name": "variance explained",
"median_label": "varex"
},
"_comment": "The median variance explained is calculated across all components, for use in later steps."
},
{
"functionname": "dec_left_op_right",
"parameters": {
"if_true": "rejected",
"if_false": "nochange",
"decide_comps": "all",
"op": ">",
"left": "dice_FS0",
"right": "dice_FT2"
},
"kwargs": {
"left2": "variance explained",
"op2": ">",
"right2": "median_varex",
"tag_if_true": "Unlikely BOLD"
},
"_comment": "Any components with higher S0 model beta map-F-statistic map Dice similarity index than T2 model beta map-F-statistic map Dice similarity index and greater than median variance explained are rejected. In slightly plainer English, this step rejects any high-variance components where significant voxels in the F-stat map overlap more with highly S0-associated voxels than T2*-associated voxels."
},
{
"functionname": "dec_left_op_right",
"parameters": {
"if_true": "rejected",
"if_false": "nochange",
"decide_comps": "all",
"op": ">",
"left": 0,
"right": "signal-noise_t"
},
"kwargs": {
"left2": "variance explained",
"op2": ">",
"right2": "median_varex",
"tag_if_true": "Unlikely BOLD"
},
"_comment": "Any components with a negative t-statistic comparing the distribution of T2* model F-statistics from voxels in clusters to those of voxels not in clusters and variance explained greater than median are rejected. That is reject any high-variance components exhibiting more 'speckled' T2*-associated voxels than 'clustered' ones."
},
{
"functionname": "dec_left_op_right",
"parameters": {
"if_true": "accepted",
"if_false": "nochange",
"decide_comps": "provisionalaccept",
"op": ">",
"left": "kappa",
"right": "rho"
},
"kwargs": {"right_scale": 2, "tag_if_true": "Likely BOLD"},
"_comment": "Any provisionally accepted components with kappa greater than two times rho are accepted. That is, even if a component has a high rho value, if kappa above threshold and substantially higher, assume it as something work keeping and accept it"
},
{
"functionname": "dec_variance_lessthan_thresholds",
"parameters": {
"if_true": "accepted",
"if_false": "nochange",
"decide_comps": "provisionalreject"
},
"kwargs": {
"var_metric": "variance explained",
"single_comp_threshold": 0.1,
"all_comp_threshold": 1.0,
"tag_if_true": "Low variance"
},
"_comment": "This step flags remaining low-variance components (less than 0.1%) and accepts up to 1% cumulative variance across these components. This is done because these components don't explain enough variance to be worth further reducing the degrees of freedom of the denoised data."
},
{
"functionname": "manual_classify",
"parameters": {"new_classification": "accepted", "decide_comps": "provisionalaccept"},
"kwargs": {"tag": "Likely BOLD"},
"_comment": "All remaining provisionally accepted components are accepted."
},
{
"functionname": "manual_classify",
"parameters": {
"new_classification": "rejected",
"decide_comps": ["provisionalreject", "unclassified"]
},
"kwargs": {"tag": "Unlikely BOLD"},
"_comment": "All remaining unclassified (nothing should be unclassified) or provisionally rejected components are rejected."
}
]
}