When pain becomes self: limbic-default mode network hyperconnectivity predicts microvascular decompression failure in trigeminal neuralgia.

Code ↔ Paper

5 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 5 matches

[1] § Materials and methods › Effect size ranking analysis ↔ 8SVM.m, lines 224 to 303 · score 0.78 · aTHA, pTHA, Subcortical region, HIP, AMY, CAU · Report a mismatch
[2] § Results › The prediction model ↔ 8SVM.m, lines 305 to 445 · score 0.76 · fold cross validation, SVM classification, classification model, connectivity features, trained, AUC · Report a mismatch
[3] § Materials and methods › Model construction ↔ 8SVM.m, lines 305 to 445 · score 0.67 · SVM classifier, cross validation, kernel, splits, train, model · Report a mismatch
[4] § Results › Top 10 connections by effect size and healthy control group position ↔ 8SVM.m, lines 224 to 303 · score 0.64 · pTHA, cortical regions, AMY, CAU, GP, NAc · Report a mismatch
[5] § Materials and methods › Model construction ↔ 8SVM.m, lines 488 to 539 · score 0.51 · confusion matrices, ROC, curve, SVM, AUC, classification · Report a mismatch

Paper

Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

MATLAB · 1,157 lines · 45 KB · CC-BY-4.0 · 5 matches

%% =====================================================================
% Brain Connectivity Group Comparison Analysis + Permutation Test +
% Modeling with Top 10 Effect Size Connections
% =====================================================================
% clear; close all; clc;
%% 1. Parameter Settings
n_permutations = 5000; % Number of permutations (suggested: 1000-10000)
alpha = 0.05; % Significance level
fdr_threshold = 0.05; % FDR correction threshold
%% 2. Load Data
cd('D:\work\data_analysis\TN')
load E_corr_cortex_subcortex_total_r
E_group = corr_cortex_subcortex_r_FDR_all;
load NE_corr_cortex_subcortex_total_r
NE_group = corr_cortex_subcortex_r_FDR_all;
fprintf('=== Brain Connectivity Group Comparison Analysis ===\n');
fprintf('E-group dimensions: %s\n', mat2str(size(E_group)));
fprintf('NE-group dimensions: %s\n', mat2str(size(NE_group)));
%% 3. Data Preprocessing and Validation
% Ensure correct data format
if size(E_group,1) ~= 100 || size(E_group,2) ~= 16 || size(E_group,3) ~= 30
error('Incorrect E_group dimensions, should be 100¡Á16¡Á30');
end
if size(NE_group,1) ~= 100 || size(NE_group,2) ~= 16 || size(NE_group,3) ~= 30
error('Incorrect NE_group dimensions, should be 100¡Á16¡Á30');
end
% Check data validity
if any(isnan(E_group(:))) || any(isinf(E_group(:)))
warning('E_group contains NaN or Inf values');
E_group(isnan(E_group)) = 0;
E_group(isinf(E_group)) = 0;
end
if any(isnan(NE_group(:))) || any(isinf(NE_group(:)))
warning('NE_group contains NaN or Inf values');
NE_group(isnan(NE_group)) = 0;
NE_group(isinf(NE_group)) = 0;
end
%% 4. Extract Connection Strength Vectors and Reshape Data
% Flatten each subject's 100¡Á16 matrix into 1600-dimensional vector
n_cortical = 100; % Number of cortical regions
n_subcortical = 16; % Number of subcortical regions
n_subjects_E = size(E_group, 3);
n_subjects_NE = size(NE_group, 3);
fprintf('Data reshaping:\n');
fprintf('Connections per subject: %d ¡Á %d = %d\n', n_cortical, n_subcortical, n_cortical*n_subcortical);
% Reshape data: each subject as a row, each column as a connection
data_E = zeros(n_subjects_E, n_cortical * n_subcortical);
data_NE = zeros(n_subjects_NE, n_cortical * n_subcortical);
for subj = 1:n_subjects_E
temp = squeeze(E_group(:,:,subj));
data_E(subj, :) = temp(:)';
end
for subj = 1:n_subjects_NE
temp = squeeze(NE_group(:,:,subj));
data_NE(subj, :) = temp(:)';
end
fprintf('E-group data matrix: %s\n', mat2str(size(data_E)));
fprintf('NE-group data matrix: %s\n', mat2str(size(data_NE)));
%% 5. Calculate Group Difference Statistics (Real Data)
fprintf('\n=== Calculating Real Group Differences ===\n');
% Initialize result matrices
real_t_stats = zeros(n_cortical, n_subcortical);
real_p_values = zeros(n_cortical, n_subcortical);
cohens_d = zeros(n_cortical, n_subcortical); % Effect size
% Perform two-sample t-test for each connection
for i = 1:n_cortical
for j = 1:n_subcortical
% Calculate connection index
conn_idx = (i-1)*n_subcortical + j;
% Extract group data for this connection
conn_E = data_E(:, conn_idx);
conn_NE = data_NE(:, conn_idx);
% Two-sample t-test
[~, p, ~, stats] = ttest2(conn_E, conn_NE, 'Vartype', 'unequal');
% Store results
real_t_stats(i, j) = stats.tstat;
real_p_values(i, j) = p;
% Calculate Cohen's d effect size
mean_diff = mean(conn_E) - mean(conn_NE);
pooled_std = sqrt((std(conn_E)^2 + std(conn_NE)^2)/2);
cohens_d(i, j) = mean_diff / pooled_std;
end
end
fprintf('Uncorrected significant connections: %d (p < %.3f)\n', sum(real_p_values(:) < alpha), alpha);
%% 6. Permutation Test
fprintf('\n=== Starting Permutation Test (%d permutations) ===\n', n_permutations);
% Combine all data
all_data = [data_E; data_NE];
n_total = size(all_data, 1);
labels = [ones(n_subjects_E, 1); 2*ones(n_subjects_NE, 1)];
% Initialize permutation distribution
perm_t_stats = zeros(n_cortical, n_subcortical, n_permutations);
perm_max_t = zeros(n_permutations, 1); % Maximum t-value (for FWE correction)
perm_min_t = zeros(n_permutations, 1); % Minimum t-value (two-sided test)
fprintf('Progress: ');
progress_step = round(n_permutations/20);
for perm = 1:n_permutations
% Display progress
if mod(perm, progress_step) == 0
fprintf('%d%% ', round(perm/n_permutations*100));
end
% Randomly permute labels
perm_labels = labels(randperm(n_total));
% Create permuted groups
perm_group1_idx = perm_labels == 1;
perm_group2_idx = perm_labels == 2;
perm_data1 = all_data(perm_group1_idx, :);
perm_data2 = all_data(perm_group2_idx, :);
% Calculate permuted t-statistics
perm_t_matrix = zeros(n_cortical, n_subcortical);
for i = 1:n_cortical
for j = 1:n_subcortical
conn_idx = (i-1)*n_subcortical + j;
conn1 = perm_data1(:, conn_idx);
conn2 = perm_data2(:, conn_idx);
% Calculate t-statistic (simplified version)
mean1 = mean(conn1);
mean2 = mean(conn2);
std1 = std(conn1);
std2 = std(conn2);
n1 = length(conn1);
n2 = length(conn2);
% Two-sample t-statistic
t_stat = (mean1 - mean2) / sqrt(std1^2/n1 + std2^2/n2);
perm_t_matrix(i, j) = t_stat;
end
end
% Store permutation results
perm_t_stats(:,:,perm) = perm_t_matrix;
% Record extreme statistics (for FWE correction)
perm_max_t(perm) = max(abs(perm_t_matrix(:)));
perm_min_t(perm) = min(perm_t_matrix(:));
end
fprintf('Completed!\n');
%% 7. Calculate Permutation Test P-values
fprintf('\n=== Calculating Permutation Test P-values ===\n');
% Initialize results
perm_p_values = zeros(n_cortical, n_subcortical);
fwe_corrected_p = zeros(n_cortical, n_subcortical); % FWE-corrected p-values
% Calculate p-value for each connection
for i = 1:n_cortical
for j = 1:n_subcortical
% Get real t-value
real_t = real_t_stats(i, j);
% Calculate permutation p-value (two-sided test)
if real_t > 0
% For positive t-values: proportion of permutations ¡Ý real t
extreme_count = sum(perm_t_stats(i,j,:) >= real_t);
else
% For negative t-values: proportion of permutations ¡Ü real t
extreme_count = sum(perm_t_stats(i,j,:) <= real_t);
end
perm_p_values(i, j) = (extreme_count + 1) / (n_permutations + 1);
% Calculate FWE-corrected p-value (based on maximum t-statistic)
if real_t > 0
fwe_extreme = sum(perm_max_t >= abs(real_t));
else
fwe_extreme = sum(perm_min_t <= real_t);
end
fwe_corrected_p(i, j) = (fwe_extreme + 1) / (n_permutations + 1);
end
end
%% 8. FDR Correction
fprintf('=== Performing FDR Correction ===\n');
% Flatten p-value matrix for FDR correction
all_perm_p = perm_p_values(:);
all_fdr_p = mafdr(all_perm_p, 'BHFDR', true); % Using Storey's method
% Reshape back to matrix
fdr_corrected_p = reshape(all_fdr_p, n_cortical, n_subcortical);
% Find FDR-corrected significant connections
fdr_significant = fdr_corrected_p < fdr_threshold;
fwe_significant = fwe_corrected_p < alpha;
fprintf('Permutation test significant connections (uncorrected): %d\n', sum(perm_p_values(:) < alpha));
fprintf('FDR-corrected significant connections (q < %.3f): %d\n', fdr_threshold, sum(fdr_significant(:)));
fprintf('FWE-corrected significant connections (p < %.3f): %d\n', alpha, sum(fwe_significant(:)));
%% 9. Find Top 10 Connections by Effect Size
fprintf('\n=== Finding Top 10 Connections by Effect Size ===\n');
% Load cortical region information
fprintf('Loading cortical region information...\n');
roi_data = readtable('Schaefer2018_100Parcels_7Networks_order_FSLMNI152_1mm.Centroid_RAS.csv');
roi_names_raw = roi_data{:, 2};
if isstring(roi_names_raw)
roi_names = cellstr(roi_names_raw);
elseif istable(roi_names_raw) || istimetable(roi_names_raw)
roi_names = table2cell(roi_names_raw);
else
roi_names = roi_names_raw;
end
% Ensure 100x1 cell array
if iscell(roi_names) && size(roi_names, 2) > 1
roi_names = roi_names(:, 1);
end
% Subcortical region names
subcortical_names = {
'HIP-rh', 'AMY-rh', 'pTHA-rh', 'aTHA-rh', 'NAc-rh', 'GP-rh', 'PUT-rh', 'CAU-rh', ...
'HIP-lh', 'AMY-lh', 'pTHA-lh', 'aTHA-lh', 'NAc-lh', 'GP-lh', 'PUT-lh', 'CAU-lh'
};
% Find top 10 connections by absolute effect size
all_cohens_d_abs = abs(cohens_d(:));
[~, sorted_indices] = sort(all_cohens_d_abs, 'descend');
% Take only top 10
n_top_conn = min(10, length(sorted_indices));
top_conn_indices = sorted_indices(1:n_top_conn);
% Get detailed information for top 10 connections
top_conn_info = cell(n_top_conn, 7); % Store: cortical index, subcortical index, cortical name, subcortical name, t-value, p-value, effect size
for k = 1:n_top_conn
idx = top_conn_indices(k);
[i, j] = ind2sub([n_cortical, n_subcortical], idx);
% Get region names
if i <= length(roi_names)
cortical_name = roi_names{i};
else
cortical_name = sprintf('Cortical_%d', i);
end
if j <= length(subcortical_names)
subcortical_name = subcortical_names{j};
else
subcortical_name = sprintf('Subcortical_%d', j);
end
% Simplify names for display
cortical_short = strrep(cortical_name, '7Networks_', '');
cortical_short = strrep(cortical_short, '_', ' ');
if length(cortical_short) > 30
cortical_short = [cortical_short(1:27) '...'];
end
subcortical_short = strrep(subcortical_name, '-', ' ');
top_conn_info{k, 1} = i;
top_conn_info{k, 2} = j;
top_conn_info{k, 3} = cortical_short;
top_conn_info{k, 4} = subcortical_short;
top_conn_info{k, 5} = real_t_stats(i, j);
top_conn_info{k, 6} = perm_p_values(i, j);
top_conn_info{k, 7} = cohens_d(i, j);
end
% Display top 10 connections
fprintf('\nTop 10 Connections by Effect Size:\n');
fprintf('Rank\tCortical Region\t\t\t\t\tSubcortical Region\t\tCohen''s d\t\tp-value\n');
fprintf('%s\n', repmat('-', 80, 1));
for k = 1:n_top_conn
fprintf('%2d\t%-35s\t%-15s\t%10.3f\t\t%.6f\n', ...
k, top_conn_info{k,3}, top_conn_info{k,4}, ...
top_conn_info{k,7}, top_conn_info{k,6});
end
%% 10. Classification Modeling with Top 10 Connections
fprintf('\n=== Classification Modeling with Top 10 Connections ===\n');
% Set classification parameters
n_classification_permutations = 100; % Random repetitions for classification
n_folds = 5; % Cross-validation folds
% Prepare feature data
top_conn_features = zeros(n_subjects_E + n_subjects_NE, n_top_conn);
top_conn_labels = [ones(n_subjects_E, 1); zeros(n_subjects_NE, 1)]; % E-group=1, NE-group=0
% Extract features
for k = 1:n_top_conn
i = top_conn_info{k, 1};
j = top_conn_info{k, 2};
conn_idx = (i-1)*n_subcortical + j;
% Extract group data for this connection
top_conn_features(:, k) = [data_E(:, conn_idx); data_NE(:, conn_idx)];
end
fprintf('Feature matrix dimensions: %s\n', mat2str(size(top_conn_features)));
fprintf('Label dimensions: %s\n', mat2str(size(top_conn_labels)));
% Initialize result storage
classification_results = struct();
classification_results.accuracy_all = zeros(n_classification_permutations, 1);
classification_results.auc_all = zeros(n_classification_permutations, 1);
classification_results.feature_importance_all = zeros(n_classification_permutations, n_top_conn);
classification_results.predictions_all = cell(n_classification_permutations, 1);
classification_results.true_labels_all = cell(n_classification_permutations, 1);
classification_results.scores_all = cell(n_classification_permutations, 1);
% Start classification modeling
fprintf('Starting classification modeling (%d random repetitions)...\n', n_classification_permutations);
for perm = 1:n_classification_permutations
if mod(perm, 10) == 0
fprintf(' Processing iteration %d/%d...\n', perm, n_classification_permutations);
end
% Randomly shuffle data
rng(perm); % Set random seed for reproducibility
n_samples = size(top_conn_features, 1);
shuffle_idx = randperm(n_samples);
X_shuffled = top_conn_features(shuffle_idx, :);
y_shuffled = top_conn_labels(shuffle_idx);
% 5-fold cross-validation
cv_indices = crossvalind('Kfold', y_shuffled, n_folds);
fold_accuracy = zeros(n_folds, 1);
fold_predictions = [];
fold_true_labels = [];
fold_scores = [];
fold_feature_importance = zeros(n_folds, n_top_conn);
for fold = 1:n_folds
% Split training and testing sets
test_idx = (cv_indices == fold);
train_idx = ~test_idx;
X_train = X_shuffled(train_idx, :);
y_train = y_shuffled(train_idx);
X_test = X_shuffled(test_idx, :);
y_test = y_shuffled(test_idx);
% Train SVM classifier
try
% Use linear SVM
svm_model = fitcsvm(X_train, y_train, ...
'KernelFunction', 'linear', ...
'Standardize', true, ...
'BoxConstraint', 1);
% Predict
[y_pred, scores] = predict(svm_model, X_test);
% Calculate accuracy
fold_accuracy(fold) = sum(y_pred == y_test) / length(y_test);
% Store prediction results
fold_predictions = [fold_predictions; y_pred];
fold_true_labels = [fold_true_labels; y_test];
fold_scores = [fold_scores; scores(:, 2)]; % Positive class probability scores
% Calculate feature importance (SVM weights)
if isprop(svm_model, 'Beta')
fold_feature_importance(fold, :) = abs(svm_model.Beta)';
else
% Alternative method if weights cannot be obtained directly
fold_feature_importance(fold, :) = rand(1, n_top_conn);
end
catch ME
fprintf('Warning: Error in fold %d: %s\n', fold, ME.message);
fold_accuracy(fold) = 0.5; % Random guessing accuracy
end
end
% Store this iteration's results
classification_results.accuracy_all(perm) = mean(fold_accuracy);
classification_results.predictions_all{perm} = fold_predictions;
classification_results.true_labels_all{perm} = fold_true_labels;
classification_results.scores_all{perm} = fold_scores;
classification_results.feature_importance_all(perm, :) = mean(fold_feature_importance, 1);
% Calculate AUC
try
[~, ~, ~, auc] = perfcurve(fold_true_labels, fold_scores, 1);
classification_results.auc_all(perm) = auc;
catch
classification_results.auc_all(perm) = 0.5;
end
end
% Calculate average performance
mean_accuracy = mean(classification_results.accuracy_all);
std_accuracy = std(classification_results.accuracy_all);
mean_auc = mean(classification_results.auc_all);
std_auc = std(classification_results.auc_all);
mean_feature_importance = mean(classification_results.feature_importance_all, 1);
fprintf('\nClassification Results Statistics:\n');
fprintf(' Average Accuracy: %.3f ¡À %.3f (range: %.3f - %.3f)\n', ...
mean_accuracy, std_accuracy, min(classification_results.accuracy_all), max(classification_results.accuracy_all));
fprintf(' Average AUC: %.3f ¡À %.3f (range: %.3f - %.3f)\n', ...
mean_auc, std_auc, min(classification_results.auc_all), max(classification_results.auc_all));
fprintf(' Chance level: 0.500\n');
% Statistical significance (permutation test)
n_above_chance = sum(classification_results.accuracy_all > 0.5);
classification_p_value = (n_above_chance + 1) / (n_classification_permutations + 1);
fprintf(' Permutation test p-value: %.4f\n', classification_p_value);
% Determine significance
if classification_p_value < 0.05
fprintf(' ? Classification performance significantly above chance level (p < 0.05)\n');
else
fprintf(' ? Classification performance not significantly above chance level (p = %.4f)\n', classification_p_value);
end
%% 11. Visualization of Classification Results
fprintf('\n=== Generating Classification Result Visualizations ===\n');
figure('Position', [100, 100, 1400, 900], 'Name', 'Classification Analysis Results', 'Color', 'white');
% Subplot 1: Accuracy Distribution
subplot(2, 3, 1);
histogram(classification_results.accuracy_all, 20, 'FaceColor', [0.2, 0.6, 0.8], 'EdgeColor', 'none');
hold on;
xline(0.5, 'r--', 'LineWidth', 2, 'Color', [0.8, 0.2, 0.2]);
xline(mean_accuracy, 'k-', 'LineWidth', 2);
xline(mean_accuracy + std_accuracy, 'k:', 'LineWidth', 1.5);
xline(mean_accuracy - std_accuracy, 'k:', 'LineWidth', 1.5);
xlabel('Classification Accuracy', 'FontSize', 12, 'FontWeight', 'bold');
ylabel('Frequency', 'FontSize', 12, 'FontWeight', 'bold');
title(sprintf('Accuracy Distribution (n=%d iterations)', n_classification_permutations), ...
'FontSize', 14, 'FontWeight', 'bold');
legend('Distribution', 'Chance level', 'Mean', '¡À1 SD', 'Location', 'best', 'FontSize', 10);
grid on;
% Add statistics text
stats_text = sprintf('Mean = %.3f ¡À %.3f\nRange = [%.3f, %.3f]\np = %.4f', ...
mean_accuracy, std_accuracy, min(classification_results.accuracy_all), ...
max(classification_results.accuracy_all), classification_p_value);
text(0.05, 0.95, stats_text, 'Units', 'normalized', ...
'HorizontalAlignment', 'left', 'VerticalAlignment', 'top', ...
'FontSize', 11, 'FontWeight', 'bold', ...
'BackgroundColor', [1, 1, 0.9], 'EdgeColor', 'k', 'Margin', 2);
% Subplot 2: ROC Curve
subplot(2, 3, 2);
% Combine all iteration predictions for ROC
all_true_labels = [];
all_scores = [];
for perm = 1:n_classification_permutations
all_true_labels = [all_true_labels; classification_results.true_labels_all{perm}];
all_scores = [all_scores; classification_results.scores_all{perm}];
end
% Calculate ROC curve
[X_roc, Y_roc, T_roc, AUC_roc] = perfcurve(all_true_labels, all_scores, 1);
% Plot ROC curve
plot(X_roc, Y_roc, 'b-', 'LineWidth', 2);
hold on;
plot([0, 1], [0, 1], 'r--', 'LineWidth', 1.5); % Random line
xlabel('False Positive Rate (FPR)', 'FontSize', 12, 'FontWeight', 'bold');
ylabel('True Positive Rate (TPR)', 'FontSize', 12, 'FontWeight', 'bold');
title(sprintf('Average ROC Curve (AUC = %.3f)', mean_auc), ...
'FontSize', 14, 'FontWeight', 'bold');
legend(sprintf('SVM (AUC = %.3f)', mean_auc), 'Random Classifier', ...
'Location', 'southeast', 'FontSize', 11);
grid on;
axis equal;
xlim([0, 1]);
ylim([0, 1]);
% Subplot 3: Confusion Matrix
subplot(2, 3, 3);
% Calculate average confusion matrix
all_predictions = [];
all_true = [];
for perm = 1:n_classification_permutations
all_predictions = [all_predictions; classification_results.predictions_all{perm}];
all_true = [all_true; classification_results.true_labels_all{perm}];
end
conf_mat = confusionmat(all_true, all_predictions);
conf_mat_normalized = conf_mat ./ sum(conf_mat, 2);
% Plot heatmap
imagesc(conf_mat_normalized);
colorbar;
caxis([0, 1]);
colormap(jet);
% Add numerical labels
for i = 1:2
for j = 1:2
text(j, i, sprintf('%.3f', conf_mat_normalized(i,j)), ...
'HorizontalAlignment', 'center', 'VerticalAlignment', 'middle', ...
'FontSize', 14, 'FontWeight', 'bold', 'Color', 'w');
end
end
set(gca, 'XTick', 1:2, 'XTickLabel', {'Predicted E', 'Predicted NE'}, ...
'YTick', 1:2, 'YTickLabel', {'True E', 'True NE'}, ...
'FontSize', 12, 'FontWeight', 'bold');
title('Normalized Confusion Matrix (Average)', 'FontSize', 14, 'FontWeight', 'bold');
% Subplot 4: Feature Importance vs Effect Size
subplot(2, 3, 4);
% Sort by feature importance
[sorted_importance, sort_idx] = sort(mean_feature_importance, 'descend');
sorted_names = cell(n_top_conn, 1);
sorted_cohens_d = zeros(n_top_conn, 1);
for k = 1:n_top_conn
idx = sort_idx(k);
sorted_names{k} = sprintf('%s-%s', top_conn_info{idx,3}(1:min(10, length(top_conn_info{idx,3}))), ...
top_conn_info{idx,4}(1:min(5, length(top_conn_info{idx,4}))));
sorted_cohens_d(k) = top_conn_info{idx,7};
end
% Plot double bar chart
x_pos = 1:n_top_conn;
bar_width = 0.35;
bar(x_pos - bar_width/2, sorted_importance, bar_width, ...
'FaceColor', [0.2, 0.6, 0.8], 'DisplayName', 'Feature Importance');
hold on;
bar(x_pos + bar_width/2, sorted_cohens_d, bar_width, ...
'FaceColor', [0.8, 0.4, 0.2], 'DisplayName', 'Cohen''s d');
xlabel('Feature Rank', 'FontSize', 12, 'FontWeight', 'bold');
ylabel('Importance / Effect Size', 'FontSize', 12, 'FontWeight', 'bold');
title('Feature Importance vs Effect Size', 'FontSize', 14, 'FontWeight', 'bold');
set(gca, 'XTick', 1:n_top_conn, 'XTickLabel', sorted_names, 'FontSize', 9, 'XTickLabelRotation', 45);
legend('Location', 'best', 'FontSize', 10);
grid on;
% Subplot 5: Group Comparison for Top 3 Features
subplot(2, 3, 5);
% Select top 3 most important features
n_show_features = min(3, n_top_conn);
for k = 1:n_show_features
idx = sort_idx(k);
i = top_conn_info{idx, 1};
j = top_conn_info{idx, 2};
conn_idx = (i-1)*n_subcortical + j;
% Extract data
feature_data_E = data_E(:, conn_idx);
feature_data_NE = data_NE(:, conn_idx);
% Create subplot
subplot(2, 3, 5);
subplot_pos = 0.1 + (k-1)*0.25;
axes('Position', [subplot_pos, 0.15, 0.2, 0.3]);
% Plot boxplot
boxplot([feature_data_E; feature_data_NE], ...
[ones(n_subjects_E,1); 2*ones(n_subjects_NE,1)], ...
'Labels', {'E Group', 'NE Group'});
hold on;
% Add scatter points
scatter(ones(n_subjects_E,1) + (rand(n_subjects_E,1)-0.5)*0.1, feature_data_E, ...
30, 'filled', 'MarkerFaceColor', [0.2, 0.4, 0.8], 'MarkerFaceAlpha', 0.6);
scatter(2*ones(n_subjects_NE,1) + (rand(n_subjects_NE,1)-0.5)*0.1, feature_data_NE, ...
30, 'filled', 'MarkerFaceColor', [0.8, 0.4, 0.2], 'MarkerFaceAlpha', 0.6);
title(sprintf('Feature #%d: %s', k, sorted_names{k}(1:min(15, length(sorted_names{k})))), ...
'FontSize', 11);
ylabel('Connection Strength', 'FontSize', 10);
grid on;
% Add effect size information
text(1.5, max([feature_data_E; feature_data_NE])*1.05, ...
sprintf('d=%.2f', sorted_cohens_d(k)), ...
'HorizontalAlignment', 'center', 'FontSize', 10, 'FontWeight', 'bold');
end
% Set main title
axes('Position', [0, 0, 1, 1], 'Visible', 'off');
text(0.5, 0.95, 'Group Comparison for Top 3 Features', ...
'HorizontalAlignment', 'center', 'FontSize', 14, 'FontWeight', 'bold');
% Subplot 6: Feature Correlation Matrix
subplot(2, 3, 6);
% Calculate feature correlations
feature_corr = corr(top_conn_features);
imagesc(feature_corr);
colorbar;
caxis([-1, 1]);
colormap(jet);
% Add correlation values
for i = 1:n_top_conn
for j = 1:n_top_conn
if i ~= j
text(j, i, sprintf('%.2f', feature_corr(i,j)), ...
'HorizontalAlignment', 'center', 'VerticalAlignment', 'middle', ...
'FontSize', 8, 'FontWeight', 'bold', 'Color', 'w');
else
text(j, i, '1.00', ...
'HorizontalAlignment', 'center', 'VerticalAlignment', 'middle', ...
'FontSize', 8, 'FontWeight', 'bold', 'Color', 'w');
end
end
end
set(gca, 'XTick', 1:n_top_conn, 'XTickLabel', 1:n_top_conn, ...
'YTick', 1:n_top_conn, 'YTickLabel', 1:n_top_conn, ...
'FontSize', 9, 'FontWeight', 'bold');
xlabel('Feature Index', 'FontSize', 12, 'FontWeight', 'bold');
ylabel('Feature Index', 'FontSize', 12, 'FontWeight', 'bold');
title('Feature Correlation Matrix', 'FontSize', 14, 'FontWeight', 'bold');
%% 12. Detailed Results Output and Saving
fprintf('\n=== Saving Detailed Results ===\n');
% Create detailed results structure
detailed_results = struct();
detailed_results.perm_test_results = struct();
detailed_results.perm_test_results.real_t_stats = real_t_stats;
detailed_results.perm_test_results.real_p_values = real_p_values;
detailed_results.perm_test_results.perm_p_values = perm_p_values;
detailed_results.perm_test_results.fdr_corrected_p = fdr_corrected_p;
detailed_results.perm_test_results.fwe_corrected_p = fwe_corrected_p;
detailed_results.perm_test_results.cohens_d = cohens_d;
detailed_results.perm_test_results.significant_fdr = fdr_significant;
detailed_results.perm_test_results.significant_fwe = fwe_significant;
detailed_results.top_connections = struct();
detailed_results.top_connections.n_top = n_top_conn;
detailed_results.top_connections.indices = top_conn_indices;
detailed_results.top_connections.info = top_conn_info;
detailed_results.top_connections.features = top_conn_features;
detailed_results.top_connections.labels = top_conn_labels;
detailed_results.classification_results = classification_results;
detailed_results.classification_results.mean_accuracy = mean_accuracy;
detailed_results.classification_results.std_accuracy = std_accuracy;
detailed_results.classification_results.mean_auc = mean_auc;
detailed_results.classification_results.std_auc = std_auc;
detailed_results.classification_results.mean_feature_importance = mean_feature_importance;
detailed_results.classification_results.p_value = classification_p_value;
% Save results
save('complete_analysis_results.mat', 'detailed_results');
fprintf('Detailed results saved to complete_analysis_results.mat\n');
% Save top 10 connections table
top_conn_table = cell2table(top_conn_info, ...
'VariableNames', {'Cortical_Index', 'Subcortical_Index', 'Cortical_Name', ...
'Subcortical_Name', 't_stat', 'p_value', 'Cohens_d'});
writetable(top_conn_table, 'top_10_connections.csv');
fprintf('Top 10 connections information saved to top_10_connections.csv\n');
% Save classification results
classification_summary = table();
classification_summary.Metric = {'Mean_Accuracy'; 'Accuracy_SD'; 'Mean_AUC'; 'AUC_SD'; 'Permutation_p_value'};
classification_summary.Value = [mean_accuracy; std_accuracy; mean_auc; std_auc; classification_p_value];
writetable(classification_summary, 'classification_summary.csv');
fprintf('Classification summary saved to classification_summary.csv\n');
%% 13. Final Summary Report
fprintf('\n========== FINAL ANALYSIS SUMMARY ==========\n');
fprintf('Dataset Information:\n');
fprintf(' E-group subjects: %d\n', n_subjects_E);
fprintf(' NE-group subjects: %d\n', n_subjects_NE);
fprintf(' Total connections: %d ¡Á %d = %d\n', n_cortical, n_subcortical, n_cortical*n_subcortical);
fprintf('\nPermutation Test Results:\n');
fprintf(' Permutations: %d\n', n_permutations);
fprintf(' FDR-significant connections: %d\n', sum(fdr_significant(:)));
fprintf(' Maximum effect size (|Cohen''s d|): %.3f\n', max(abs(cohens_d(:))));
fprintf('\nTop 10 Connections by Effect Size:\n');
for k = 1:min(5, n_top_conn)
fprintf(' %d. %s - %s (d=%.3f, p=%.4f)\n', ...
k, top_conn_info{k,3}, top_conn_info{k,4}, ...
top_conn_info{k,7}, top_conn_info{k,6});
end
fprintf('\nClassification Modeling Results:\n');
fprintf(' Random repetitions: %d\n', n_classification_permutations);
fprintf(' Cross-validation folds: %d\n', n_folds);
fprintf(' Mean accuracy: %.3f ¡À %.3f\n', mean_accuracy, std_accuracy);
fprintf(' Mean AUC: %.3f ¡À %.3f\n', mean_auc, std_auc);
fprintf(' Permutation test p-value: %.4f\n', classification_p_value);
if classification_p_value < 0.05
fprintf(' ¡ú Classification performance is SIGNIFICANTLY above chance level!\n');
else
fprintf(' ¡ú Classification performance is NOT SIGNIFICANT\n');
end
fprintf('\n========== ANALYSIS COMPLETED! ==========\n');
fprintf
%% =====================================================================
%% 13. Final Summary Report
fprintf('\n========== FINAL ANALYSIS SUMMARY ==========\n');
fprintf('Dataset Information:\n');
fprintf(' E-group subjects: %d\n', n_subjects_E);
fprintf(' NE-group subjects: %d\n', n_subjects_NE);
fprintf(' Total connections: %d ¡Á %d = %d\n', n_cortical, n_subcortical, n_cortical*n_subcortical);
fprintf('\nPermutation Test Results:\n');
fprintf(' Permutations: %d\n', n_permutations);
fprintf(' FDR-significant connections: %d\n', sum(fdr_significant(:)));
fprintf(' Maximum effect size (|Cohen''s d|): %.3f\n', max(abs(cohens_d(:))));
fprintf('\nTop 10 Connections by Effect Size:\n');
for k = 1:min(5, n_top_conn)
fprintf(' %d. %s - %s (d=%.3f, p=%.4f)\n', ...
k, top_conn_info{k,3}, top_conn_info{k,4}, ...
top_conn_info{k,7}, top_conn_info{k,6});
end
fprintf('\nClassification Modeling Results:\n');
fprintf(' Random repetitions: %d\n', n_classification_permutations);
fprintf(' Cross-validation folds: %d\n', n_folds);
fprintf(' Mean accuracy: %.3f ¡À %.3f\n', mean_accuracy, std_accuracy);
fprintf(' Mean AUC: %.3f ¡À %.3f\n', mean_auc, std_auc);
fprintf(' Permutation test p-value: %.4f\n', classification_p_value);
if classification_p_value < 0.05
fprintf(' ¡ú Classification performance is SIGNIFICANTLY above chance level!\n');
else
fprintf(' ¡ú Classification performance is NOT SIGNIFICANT\n');
end
fprintf('\n========== ANALYSIS COMPLETED! ==========\n');
% É¾³ý»ò×¢ÊÍµôÏÂÃæÕâÐÐ
% fprintf
%% =====================================================================
% ÐÂÔö²¿·Ö£º±£´æ·ÖÀà·ÖÎöÍ¼ÏñÎª¸ßÖÊÁ¿Ê¸Á¿Í¼£¨±£´æµ½Vector_FiguresÎÄ¼þ¼Ð£©
%% 13. Final Summary Report
fprintf('\n========== FINAL ANALYSIS SUMMARY ==========\n');
fprintf('Dataset Information:\n');
fprintf(' E-group subjects: %d\n', n_subjects_E);
fprintf(' NE-group subjects: %d\n', n_subjects_NE);
fprintf(' Total connections: %d ¡Á %d = %d\n', n_cortical, n_subcortical, n_cortical*n_subcortical);
fprintf('\nPermutation Test Results:\n');
fprintf(' Permutations: %d\n', n_permutations);
fprintf(' FDR-significant connections: %d\n', sum(fdr_significant(:)));
fprintf(' Maximum effect size (|Cohen''s d|): %.3f\n', max(abs(cohens_d(:))));
fprintf('\nTop 10 Connections by Effect Size:\n');
for k = 1:min(5, n_top_conn)
fprintf(' %d. %s - %s (d=%.3f, p=%.4f)\n', ...
k, top_conn_info{k,3}, top_conn_info{k,4}, ...
top_conn_info{k,7}, top_conn_info{k,6});
end
fprintf('\nClassification Modeling Results:\n');
fprintf(' Random repetitions: %d\n', n_classification_permutations);
fprintf(' Cross-validation folds: %d\n', n_folds);
fprintf(' Mean accuracy: %.3f ¡À %.3f\n', mean_accuracy, std_accuracy);
fprintf(' Mean AUC: %.3f ¡À %.3f\n', mean_auc, std_auc);
fprintf(' Permutation test p-value: %.4f\n', classification_p_value);
if classification_p_value < 0.05
fprintf(' ¡ú Classification performance is SIGNIFICANTLY above chance level!\n');
else
fprintf(' ¡ú Classification performance is NOT SIGNIFICANT\n');
end
fprintf('\n========== ANALYSIS COMPLETED! ==========\n');
% É¾³ý»ò×¢ÊÍµôÏÂÃæÕâÐÐ
% fprintf
%% =====================================================================
% ±£´æÍ¼Ïñ²¿·Ö´ÓÕâÀï¿ªÊ¼
% =====================================================================
fprintf('\n========================================\n');
fprintf('Saving classification analysis figures as vector graphics...\n');
fprintf('========================================\n');
% ´´½¨±£´æÍ¼ÏñµÄÎÄ¼þ¼Ð
output_folder = 'Vector_Figures';
if ~exist(output_folder, 'dir')
mkdir(output_folder);
end
% ÔÚVector_FiguresÏÂ´´½¨·ÖÀà½á¹ûµÄ×ÓÎÄ¼þ¼Ð
classification_subfolder = fullfile(output_folder, 'Classification_Results');
if ~exist(classification_subfolder, 'dir')
mkdir(classification_subfolder);
end
% »ñÈ¡µ±Ç°Ê±¼ä´Á
timestamp = datestr(now, 'yyyymmdd_HHMMSS');
% ¶¨ÒåÒª±£´æµÄfigure
figure_info = {
1, 'Classification_Analysis_Results';
};
% ÉèÖÃÍ¼ÐÎÊôÐÔ
set(0, 'DefaultAxesFontName', 'Arial');
set(0, 'DefaultTextFontName', 'Arial');
% ±£´æÖ÷Í¼
for i = 1:size(figure_info, 1)
fig_num = figure_info{i, 1};
fig_name = figure_info{i, 2};
if ishandle(fig_num)
figure(fig_num);
set(gcf, 'PaperPositionMode', 'auto');
set(gcf, 'InvertHardcopy', 'off');
set(gcf, 'Renderer', 'painters');
fprintf('\nSaving Figure %d: %s\n', fig_num, fig_name);
% ±£´æÎªPDF
pdf_filename = fullfile(classification_subfolder, sprintf('%s_%s.pdf', fig_name, timestamp));
try
exportgraphics(gcf, pdf_filename, 'ContentType', 'vector');
fprintf(' ? Saved PDF: %s\n', pdf_filename);
catch
print(gcf, pdf_filename, '-dpdf', '-r600');
fprintf(' ? Saved PDF (print): %s\n', pdf_filename);
end
% ±£´æÎªEPS
eps_filename = fullfile(classification_subfolder, sprintf('%s_%s.eps', fig_name, timestamp));
print(gcf, eps_filename, '-depsc', '-r600');
fprintf(' ? Saved EPS: %s\n', eps_filename);
% ±£´æÎªPNG
png_filename = fullfile(classification_subfolder, sprintf('%s_%s.png', fig_name, timestamp));
print(gcf, png_filename, '-dpng', '-r300');
fprintf(' ? Saved PNG: %s\n', png_filename);
end
end
fprintf('\n========================================\n');
fprintf('All figures saved to: %s\n', classification_subfolder);
fprintf('========================================\n');
%% 13. Final Summary Report
fprintf('\n========== FINAL ANALYSIS SUMMARY ==========\n');
fprintf('Dataset Information:\n');
fprintf(' E-group subjects: %d\n', n_subjects_E);
fprintf(' NE-group subjects: %d\n', n_subjects_NE);
fprintf(' Total connections: %d ¡Á %d = %d\n', n_cortical, n_subcortical, n_cortical*n_subcortical);
fprintf('\nPermutation Test Results:\n');
fprintf(' Permutations: %d\n', n_permutations);
fprintf(' FDR-significant connections: %d\n', sum(fdr_significant(:)));
fprintf(' Maximum effect size (|Cohen''s d|): %.3f\n', max(abs(cohens_d(:))));
fprintf('\nTop 10 Connections by Effect Size:\n');
for k = 1:min(5, n_top_conn)
fprintf(' %d. %s - %s (d=%.3f, p=%.4f)\n', ...
k, top_conn_info{k,3}, top_conn_info{k,4}, ...
top_conn_info{k,7}, top_conn_info{k,6});
end
fprintf('\nClassification Modeling Results:\n');
fprintf(' Random repetitions: %d\n', n_classification_permutations);
fprintf(' Cross-validation folds: %d\n', n_folds);
fprintf(' Mean accuracy: %.3f ¡À %.3f\n', mean_accuracy, std_accuracy);
fprintf(' Mean AUC: %.3f ¡À %.3f\n', mean_auc, std_auc);
fprintf(' Permutation test p-value: %.4f\n', classification_p_value);
if classification_p_value < 0.05
fprintf(' ¡ú Classification performance is SIGNIFICANTLY above chance level!\n');
else
fprintf(' ¡ú Classification performance is NOT SIGNIFICANT\n');
end
fprintf('\n========== ANALYSIS COMPLETED! ==========\n');
% É¾³ý»ò×¢ÊÍµôÏÂÃæÕâÐÐ
% fprintf
%% =====================================================================
% ±£´æÍ¼Ïñ²¿·Ö - ±£´æµ½ vector_figures ÎÄ¼þ¼Ð
% =====================================================================
fprintf('\n========================================\n');
fprintf('Saving classification analysis figures to vector_figures folder...\n');
fprintf('========================================\n');
% ´´½¨±£´æÍ¼ÏñµÄÎÄ¼þ¼Ð£¨Ê¹ÓÃÐ¡Ð´ vector_figures£©
output_folder = 'vector_figures';
if ~exist(output_folder, 'dir')
mkdir(output_folder);
fprintf('Created folder: %s\n', output_folder);
end
% ÔÚvector_figuresÏÂ´´½¨·ÖÀà½á¹ûµÄ×ÓÎÄ¼þ¼Ð
classification_subfolder = fullfile(output_folder, 'Classification_Results');
if ~exist(classification_subfolder, 'dir')
mkdir(classification_subfolder);
fprintf('Created subfolder: %s\n', classification_subfolder);
end
% »ñÈ¡µ±Ç°Ê±¼ä´Á
timestamp = datestr(now, 'yyyymmdd_HHMMSS');
% ¶¨ÒåÒª±£´æµÄfigure
figure_info = {
1, 'Classification_Analysis_Results';
};
% ÉèÖÃÍ¼ÐÎÊôÐÔ
set(0, 'DefaultAxesFontName', 'Arial');
set(0, 'DefaultTextFontName', 'Arial');
% ±£´æÖ÷Í¼
for i = 1:size(figure_info, 1)
fig_num = figure_info{i, 1};
fig_name = figure_info{i, 2};
if ishandle(fig_num)
figure(fig_num);
set(gcf, 'PaperPositionMode', 'auto');
set(gcf, 'InvertHardcopy', 'off');
set(gcf, 'Renderer', 'painters');
fprintf('\nSaving Figure %d: %s\n', fig_num, fig_name);
% 1. ±£´æÎªPDF£¨Ê¸Á¿¸ñÊ½£¬×îÊÊºÏ²åÈëÂÛÎÄ£©
pdf_filename = fullfile(classification_subfolder, sprintf('%s_%s.pdf', fig_name, timestamp));
try
exportgraphics(gcf, pdf_filename, 'ContentType', 'vector');
fprintf(' ? Saved PDF (vector): %s\n', pdf_filename);
catch
print(gcf, pdf_filename, '-dpdf', '-r600');
fprintf(' ? Saved PDF (print): %s\n', pdf_filename);
end
% 2. ±£´æÎªEPS£¨Ê¸Á¿¸ñÊ½£¬ÊÊºÏLaTeX£©
eps_filename = fullfile(classification_subfolder, sprintf('%s_%s.eps', fig_name, timestamp));
print(gcf, eps_filename, '-depsc', '-r600');
fprintf(' ? Saved EPS (vector): %s\n', eps_filename);
% 3. ±£´æÎªTIFF£¨¸ß·Ö±æÂÊ£¬ÊÊºÏÆÚ¿¯Í¶¸å£©
tiff_filename = fullfile(classification_subfolder, sprintf('%s_%s_600dpi.tiff', fig_name, timestamp));
print(gcf, tiff_filename, '-dtiff', '-r600');
fprintf(' ? Saved TIFF (600 DPI): %s\n', tiff_filename);
% 4. ±£´æÎªPNG£¨¸ß·Ö±æÂÊ£¬ÊÊºÏ¿ìËÙÔ¤ÀÀ£©
png_filename = fullfile(classification_subfolder, sprintf('%s_%s_600dpi.png', fig_name, timestamp));
print(gcf, png_filename, '-dpng', '-r600');
fprintf(' ? Saved PNG (600 DPI): %s\n', png_filename);
% 5. ±£´æÎªJPEG£¨Ñ¹Ëõ¸ñÊ½£¬ÊÊºÏÍøÒ³£©
jpg_filename = fullfile(classification_subfolder, sprintf('%s_%s_300dpi.jpg', fig_name, timestamp));
print(gcf, jpg_filename, '-djpeg', '-r300');
fprintf(' ? Saved JPEG (300 DPI): %s\n', jpg_filename);
end
end
% ¶îÍâ±£´æ£º½«Figure 1µÄÃ¿¸ö×ÓÍ¼µ¥¶À±£´æ
if ishandle(1)
figure(1);
% »ñÈ¡µ±Ç°figureµÄËùÓÐaxes
all_axes = findall(gcf, 'Type', 'axes');
subplot_counter = 0;
subplot_names = {'01_Accuracy_Distribution', '02_ROC_Curve', '03_Confusion_Matrix', ...
'04_Feature_Importance', '05_Group_Comparison', '06_Correlation_Matrix'};
for ax_idx = 1:length(all_axes)
% Ìø¹ýcolorbarºÍlegend
ax_tag = get(all_axes(ax_idx), 'Tag');
if ~isempty(ax_tag) && (strcmp(ax_tag, 'colorbar') || strcmp(ax_tag, 'legend') || strcmp(ax_tag, 'annotation'))
continue;
end
% »ñÈ¡axesµÄÎ»ÖÃ
ax_pos = get(all_axes(ax_idx), 'Position');
% Ö»±£´æÖ÷ÒªµÄ×ÓÍ¼
if ax_pos(1) > 0.02 && ax_pos(1) < 0.98 && ax_pos(2) > 0.02 && ax_pos(2) < 0.98
subplot_counter = subplot_counter + 1;
% ´´½¨ÐÂµÄfigure
single_fig = figure('Visible', 'off', 'Position', [100, 100, 500, 400], ...
'Color', 'white', 'Renderer', 'painters');
% ¸´ÖÆaxesÄÚÈÝ
new_ax = copyobj(all_axes(ax_idx), single_fig);
set(new_ax, 'Position', [0.13, 0.11, 0.775, 0.815]);
% ÉèÖÃ×ÓÍ¼Ãû³Æ
if subplot_counter <= length(subplot_names)
subplot_name = subplot_names{subplot_counter};
else
subplot_name = sprintf('Subplot_%02d', subplot_counter);
end
% ±£´æµ¥¸ö×ÓÍ¼ÎªPDF
subplot_pdf = fullfile(classification_subfolder, sprintf('%s_%s.pdf', subplot_name, timestamp));
try
exportgraphics(single_fig, subplot_pdf, 'ContentType', 'vector');
fprintf(' ? Saved subplot %d as PDF: %s\n', subplot_counter, subplot_pdf);
catch
print(single_fig, subplot_pdf, '-dpdf', '-r600');
end
% ±£´æµ¥¸ö×ÓÍ¼ÎªTIFF
subplot_tiff = fullfile(classification_subfolder, sprintf('%s_%s_600dpi.tiff', subplot_name, timestamp));
print(single_fig, subplot_tiff, '-dtiff', '-r600');
fprintf(' ? Saved subplot %d as TIFF: %s\n', subplot_counter, subplot_tiff);
% ±£´æµ¥¸ö×ÓÍ¼ÎªPNG
subplot_png = fullfile(classification_subfolder, sprintf('%s_%s.png', subplot_name, timestamp));
print(single_fig, subplot_png, '-dpng', '-r300');
close(single_fig);
end
end
fprintf('\n ? %d individual subplots saved in multiple formats\n', subplot_counter);
end
% ±£´æÉèÖÃÐÅÏ¢
settings_file = fullfile(classification_subfolder, sprintf('Figure_Settings_%s.txt', timestamp));
fid = fopen(settings_file, 'w');
fprintf(fid, 'Classification Analysis Figure Export Settings\n');
fprintf(fid, '==============================================\n');
fprintf(fid, 'Date: %s\n', datestr(now));
fprintf(fid, 'MATLAB Version: %s\n', version);
fprintf(fid, 'Output Folder: %s\n', classification_subfolder);
fprintf(fid, 'Renderer: painters\n');
fprintf(fid, 'Resolutions: PDF/vector, EPS/vector, TIFF 600 DPI, PNG 600 DPI, JPEG 300 DPI\n');
fprintf(fid, 'Font: Arial\n');
fprintf(fid, '\nAnalysis Summary:\n');
fprintf(fid, 'Classification Results:\n');
fprintf(fid, ' Mean Accuracy: %.3f ¡À %.3f\n', mean_accuracy, std_accuracy);
fprintf(fid, ' Mean AUC: %.3f ¡À %.3f\n', mean_auc, std_auc);
fprintf(fid, ' Permutation p-value: %.4f\n', classification_p_value);
fprintf(fid, ' Significant: %s\n', string(classification_p_value < 0.05));
fclose(fid);
fprintf('\n========================================\n');
fprintf('All figures saved to: %s\n', classification_subfolder);
fprintf('Settings file saved: %s\n', settings_file);
fprintf('========================================\n');
% ÏÔÊ¾±£´æµÄÎÄ¼þÁÐ±í
fprintf('\nSaved files in %s:\n', classification_subfolder);
files = dir(fullfile(classification_subfolder, '*'));
file_count = 0;
tiff_count = 0;
pdf_count = 0;
eps_count = 0;
png_count = 0;
jpg_count = 0;
for i = 1:length(files)
if ~files(i).isdir
file_count = file_count + 1;
file_size = files(i).bytes;
if file_size < 1024
size_str = sprintf('%d B', file_size);
elseif file_size < 1024^2
size_str = sprintf('%.2f KB', file_size/1024);
else
size_str = sprintf('%.2f MB', file_size/(1024^2));
end
fprintf(' - %s (%s)\n', files(i).name, size_str);
% Í³¼Æ²»Í¬¸ñÊ½ÎÄ¼þÊýÁ¿
[~, ~, ext] = fileparts(files(i).name);
switch lower(ext)
case '.tiff'
tiff_count = tiff_count + 1;
case '.pdf'
pdf_count = pdf_count + 1;
case '.eps'
eps_count = eps_count + 1;
case '.png'
png_count = png_count + 1;
case '.jpg'
jpg_count = jpg_count + 1;
end
end
end
fprintf('\nFile format summary:\n');
fprintf(' PDF files: %d (vector)\n', pdf_count);
fprintf(' EPS files: %d (vector)\n', eps_count);
fprintf(' TIFF files: %d (600 DPI)\n', tiff_count);
fprintf(' PNG files: %d (600/300 DPI)\n', png_count);
fprintf(' JPEG files: %d (300 DPI)\n', jpg_count);
fprintf(' Total: %d files\n', file_count);
fprintf('\n');
% ´ò¿ª±£´æµÄÎÄ¼þ¼Ð
try
winopen(classification_subfolder);
fprintf('Opened folder: %s\n', classification_subfolder);
catch
fprintf('Files saved to: %s\n', fullfile(pwd, classification_subfolder));
end

8SVM.m, under CC-BY-4.0 · at the source

Overview

Authors: Ying Wang^1,2,3, Chenglong Cao^1,4, Hao Chen⁵, Min Wu¹, Xiaosong He⁶, Xiaofeng Jiang¹

ORCID iDs: Xiaosong He

Department of Neurosurgery, The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, Anhui Province 230001, P. R. China
Anhui Provincial Stereotactic Neurosurgical Institute, Hefei, Anhui Province 230001, P. R. China
Anhui Province Key Laboratory of Brain Function and Brain Disease, Hefei, Anhui Province 230001, P. R. China
MRC Brain Networks Dynamics Unit, University of Oxford, Oxford, UK
Department of Radiology, The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, Anhui 230001, China
Department of Psychology, University of Science and Technology of China, Hefei, Anhui Province 230026, China

Institutions: University of Science and Technology of China (China); University of Oxford (United Kingdom); Brain Network Dynamics Unit (United Kingdom)

Journal: Brain communications, volume 8, issue 3, article fcag220

Dates: received 26 February 2026; accepted 9 June 2026; published online 11 June 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.1093/braincomms/fcag220 · PMID 42327366 · PMCID PMC13281923 · OpenAlex W7164311646

Open access: gold, a free copy (OpenAlex)

Status: code verified (how the code is found and verified)

Categories: structural MRI / diffusion (modality), fMRI (modality), human (organism), pain (population), systems (subfield)

Methods: Statistics, Machine learning, Connectivity, fMRI & imaging

Keywords: trigeminal neuralgia, microvascular decompression, pain embedding, default mode network, functional connectivity

Topic: Trigeminal Neuralgia and Treatments (Clinical Neurology, Medicine), according to OpenAlex

Funding: University of Science and Technology of China; Double First-Class Initiative (YD9110002096); Anhui Provincial Health and Wellness Research Projects (AHWJ2024Aa30440)

Citations: not cited yet (Europe PMC); 35 references in the paper

Abstract

Microvascular decompression is the standard surgical intervention for classical trigeminal neuralgia attributed to NVC. However, up to 30% of patients experience poor outcomes despite technically successful surgery, suggesting non-peripheral mechanisms. This study investigates whether central neural reorganization—manifested as altered cortico-subcortical connectivity and white matter microstructure—underlies poor surgical outcomes in patients with ‘incidental’ low-pressure vascular contact. In this prospective nested case-control study, we enrolled 60 patients with medication-refractory classical trigeminal neuralgia who underwent microvascular decompression and 30 matched healthy controls. Based on the 3-month surgical outcome, patients were retrospectively stratified into effective and ineffective groups. All participants underwent preoperative resting-state functional MRI and diffusion tensor imaging. Whole-brain functional connectivity between 100 cortical (Schaefer atlas) and 16 subcortical (Tian atlas) regions was analysed. Connections were ranked by between-group effect size (Cohen’s d); the top 10 were used to train a SVM classifier. Correlational tractography assessed the relationship between fractional anisotropy and disease duration. Despite comparable preoperative pain, the ineffective group had significantly lower intraoperative vascular pressure (P = 0.02). Compared to the effective group, the ineffective group exhibited significantly enhanced functional connectivity between default mode, somatomotor and control networks and subcortical limbic structures (amygdala, nucleus accumbens, hippocampus) (P < 0.05, permutation test). A SVM classifier trained on these features achieved 86.0% accuracy (AUC = 0.931) in distinguishing outcome groups. Diffusion tensor imaging analysis revealed that longer disease duration was associated with decreased fractional anisotropy in sensorimotor white matter tracts, indicating progressive microstructural decline. Poor surgical outcome is not merely ineffective decompression, but a signature of maladaptive central reorganization—specifically, heightened default mode network–limbic coupling. This hyperconnectivity indicates that pain has become cognitively and affectively ‘self-embedded’, transitioning from a sensory event to a persistent self-referential narrative that sustains pain even after peripheral trigger removal. These results reframe a subset of trigeminal neuralgia as a central network disorder of pain self-representation and identify default mode network–amygdala hyperconnectivity as a candidate target for circuit-based neuromodulation—such as repetitive transcranial magnetic stimulation over medial prefrontal hubs—to ‘unlearn’ the embedded pain trace, either adjunctively or as an alternative to microvascular decompression.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 5 matches between paragraphs and lines of code.

Zenodo 19106465

License: CC-BY-4.0

State: the link answers, verified on 27 September 2026

Evidence: files inventoried

Languages: MATLAB (4), Shell (4)

Size: 8 files, 8 scripts

Software Heritage: not checked

Found in: “Data availability”

Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: Statistics and Machine Learning Toolbox (2 files), AFNI (1 file), Dcm2Bids (1 file), fMRIPrep (1 file), FSL (1 file)

Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)

27 September 2026: the link answers (HTTP 200)

8 files

1.dcm2bids.sh, Shell, 24 lines
2.fmriprep2.sh, Shell, 43 lines
3.xcp.sh, Shell, 25 lines
4.cortex_ts.sh, Shell, 79 lines
6permutation_r_compare.m, MATLAB, 433 lines
7effectsize_HC_top10_leave one out.m, MATLAB, 251 lines
8SVM.m, MATLAB, 1,157 lines, 5 matches
age_duration.m, MATLAB, 26 lines

At the source: doi.org/10.5281/zenodo.19106465

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
8 scripts, each with its path and the digest of its content;
5 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. All custom MATLAB scripts used for data analysis have been deposited in Zenodo (DOI:10.5281/zenodo.19106465 (https://doi.org/10.5281/zenodo.19106465)) and are publicly available at https://zenodo.org/records/19106465.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 6 authors, 5 keywords, 3 funders, 35 references.

Cite

This paper

Wang, Y., Cao, C., Chen, H., Wu, M., He, X., & Jiang, X. (2026). When pain becomes self: limbic-default mode network hyperconnectivity predicts microvascular decompression failure in trigeminal neuralgia. Brain communications, 8(3), fcag220. https://doi.org/10.1093/braincomms/fcag220

BibTeX

@article{wang2026when, author = {Wang, Ying and Cao, Chenglong and Chen, Hao and Wu, Min and He, Xiaosong and Jiang, Xiaofeng}, title = {{When pain becomes self: limbic-default mode network hyperconnectivity predicts microvascular decompression failure in trigeminal neuralgia}}, journal = {Brain communications}, year = {2026}, month = jun, volume = {8}, number = {3}, pages = {fcag220}, publisher = {Oxford University Press}, issn = {2632-1297}, doi = {10.1093/braincomms/fcag220}, url = {https://doi.org/10.1093/braincomms/fcag220}, pmid = {42327366}, pmcid = {PMC13281923} }

RIS

TY - JOUR AU - Wang, Ying AU - Cao, Chenglong AU - Chen, Hao AU - Wu, Min AU - He, Xiaosong AU - Jiang, Xiaofeng TI - When pain becomes self: limbic-default mode network hyperconnectivity predicts microvascular decompression failure in trigeminal neuralgia T2 - Brain communications J2 - Brain Commun PY - 2026 DA - 2026/06/11 VL - 8 IS - 3 SP - fcag220 SN - 2632-1297 PB - Oxford University Press DO - 10.1093/braincomms/fcag220 UR - https://doi.org/10.1093/braincomms/fcag220 LA - en ER -

CSL-JSON

{ "id": "10.1093/braincomms/fcag220", "type": "article-journal", "title": "When pain becomes self: limbic-default mode network hyperconnectivity predicts microvascular decompression failure in trigeminal neuralgia", "container-title": "Brain communications", "author": [ { "family": "Wang", "given": "Ying" }, { "family": "Cao", "given": "Chenglong" }, { "family": "Chen", "given": "Hao" }, { "family": "Wu", "given": "Min" }, { "family": "He", "given": "Xiaosong" }, { "family": "Jiang", "given": "Xiaofeng" } ], "container-title-short": "Brain Commun", "volume": "8", "issue": "3", "page": "fcag220", "DOI": "10.1093/braincomms/fcag220", "PMID": "42327366", "PMCID": "PMC13281923", "ISSN": "2632-1297", "publisher": "Oxford University Press", "URL": "https://doi.org/10.1093/braincomms/fcag220", "language": "en", "issued": { "date-parts": [ [ 2026, 6, 11 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Add the badge to the code's README

If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.

Markdown

<!-- oscr-badge:start -->
[![OSCR](https://openscicode.org/badge/doi/10.1093/braincomms/fcag220.svg)](https://openscicode.org/badge/doi/10.1093/braincomms/fcag220)
<!-- oscr-badge:end -->

reStructuredText

.. oscr-badge:start

.. image:: https://openscicode.org/badge/doi/10.1093/braincomms/fcag220.svg
   :target: https://openscicode.org/badge/doi/10.1093/braincomms/fcag220
   :alt: OSCR

.. oscr-badge:end

HTML

<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1093/braincomms/fcag220"><img src="https://openscicode.org/badge/doi/10.1093/braincomms/fcag220.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->

If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.

Report a mismatch between the code and the paper · Report a code error (signed in)

Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.

Report a reproduction failure (signed in)

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.