Whole Brain AAL Functional Connectivity Calculator
Estimate whole-brain functional connectivity summary values from an AAL parcellation using standard graph and correlation formulas. This calculator helps you quantify the number of possible region-to-region connections, observed network density, mean Fisher z, total connectivity strength, and the minimum scan length needed for more stable correlation estimation.
Interactive Calculator
How to Calculate Whole Brain AAL Functional Connectivity
Whole-brain AAL functional connectivity is usually computed by parcellating the brain into anatomical regions from the Automated Anatomical Labeling atlas, extracting one representative time series for each region, and then correlating every region with every other region. The result is a square connectivity matrix in which each cell contains a statistical association, most commonly Pearson correlation, between two regional fMRI signals. A practical “whole-brain” summary is then derived by averaging all off-diagonal correlations, transforming them with Fisher z, counting suprathreshold edges, or calculating graph measures such as density, strength, clustering, and modularity.
The calculator above focuses on the summary quantities researchers most often report when discussing whole-brain AAL connectivity: the number of possible unique region pairs, observed edge count above a chosen threshold, whole-brain mean correlation, Fisher z transformation, and a density estimate. These are not substitutes for full connectome analysis, but they are extremely useful for study planning, methods reporting, QC, reproducible documentation, and rapid interpretation of network-level findings.
What AAL Means in Functional Connectivity
The AAL atlas is an anatomical parcellation that divides the brain into labeled regions of interest, or ROIs. Two common versions are used in resting-state and task fMRI studies:
- AAL90: often refers to 90 cerebral regions, excluding cerebellar ROIs in some pipelines.
- AAL116: includes 116 regions, commonly including cerebellar and vermis structures.
After preprocessing the fMRI data, you extract the average blood oxygen level dependent time series from each ROI. If you have 116 AAL regions, you will end up with 116 time series, each with T time points. A whole-brain functional connectivity matrix is then produced by correlating each of the 116 signals with every other signal.
Step 1: Count the possible ROI-to-ROI connections
If your matrix is symmetric and you do not count self-connections on the diagonal, the number of unique undirected connections is:
Possible edges = N × (N – 1) / 2
For AAL90: 90 × 89 / 2 = 4,005 unique pairs.
For AAL116: 116 × 115 / 2 = 6,670 unique pairs.
Step 2: Extract regional time series correctly
Your final connectivity estimate is only as valid as your time series extraction strategy. Most pipelines calculate the mean signal across all voxels in each ROI after spatial normalization to a common template and before pairwise correlation. Researchers often apply nuisance regression, motion correction, temporal filtering, and sometimes scrubbing of high-motion volumes. These decisions matter because they change the covariance structure of the final ROI time series.
- Preprocess the fMRI data using a validated workflow.
- Warp the atlas or the subject data into the same space.
- Extract one average time series per AAL ROI.
- Optionally regress nuisance terms such as motion, white matter, and CSF signals.
- Apply the same masking and filtering strategy across participants.
- Compute pairwise correlations using Pearson correlation or another defined measure.
Step 3: Build the connectivity matrix
The standard formula for Pearson correlation between ROI i and ROI j is:
r(i,j) = cov(Xi, Xj) / (sd(Xi) × sd(Xj))
Do this for every pair of ROIs. The resulting matrix is symmetric, with ones on the diagonal. In most whole-brain summaries, the diagonal is discarded because self-correlation does not provide network information.
Once you have the matrix, there are several ways to calculate a “whole-brain AAL functional connectivity” value:
- Mean signed connectivity: average all off-diagonal r values.
- Mean absolute connectivity: average the absolute values of off-diagonal correlations.
- Mean Fisher z: transform each r value to z and average on the z scale.
- Density: proportion of connections exceeding a defined threshold.
- Global strength: sum of all edge weights or mean edge weight across the matrix.
Why Fisher z Transformation Matters
Correlations are bounded between -1 and 1, so their distribution becomes skewed as values move away from zero. For inference, averaging, and group-level statistics, many analysts convert correlations to Fisher z values first:
z = 0.5 × ln((1 + r) / (1 – r))
This stabilizes variance and makes values better suited for parametric analysis. After averaging on the z scale, the result can be converted back to r if needed using the inverse hyperbolic tangent relationship. The calculator above reports the Fisher z of your mean correlation so you can describe both intuitive effect size and analysis-ready transformed values.
Step 4: Calculate network density from thresholded edges
Many studies threshold the matrix so that only connections above a specified magnitude are considered “present.” If you count the number of suprathreshold edges and divide by the total possible edges, you obtain network density:
Density = observed edges / possible edges
For example, if you are using AAL116 and detect 1,600 suprathreshold edges, density is 1,600 / 6,670 = 0.2399, or about 24.0%. This simple summary is useful for comparing subjects, groups, or processing choices, but it depends heavily on your threshold and whether you use signed or absolute correlations.
| AAL configuration | Number of ROIs | Unique undirected connections | Matrix size including diagonal | Interpretation |
|---|---|---|---|---|
| AAL90 | 90 | 4,005 | 90 × 90 = 8,100 cells | Often used when cerebellar regions are excluded from the summary connectome. |
| AAL116 | 116 | 6,670 | 116 × 116 = 13,456 cells | Common full-atlas implementation with cerebellar and vermis regions included. |
| Difference | +26 | +2,665 | +5,356 cells | Adding 26 ROIs creates a much larger increase in pair count because edges scale roughly with N squared. |
How many time points do you need?
The reliability of functional connectivity estimates improves with longer time series, though gains are not perfectly linear and depend on noise, motion, preprocessing, and the exact metric being used. A commonly used approximation for the standard error of the Fisher z transformed correlation is:
SE(z) = 1 / sqrt(T – 3)
This means that as the number of time points increases, the uncertainty around your pairwise connectivity estimate decreases. For example, with 120 time points, SE(z) is around 0.092. With 240 time points, SE(z) is around 0.065. This does not guarantee high test-retest reliability, but it does provide a useful first-order indicator for planning.
| Time points (T) | Approximate SE of Fisher z | TR = 2.0 s scan duration | Practical interpretation |
|---|---|---|---|
| 120 | 0.092 | 4.0 minutes | Often acceptable for exploratory work, but noisier for whole-brain edge estimation. |
| 180 | 0.075 | 6.0 minutes | Common minimum target for more stable resting-state summaries. |
| 240 | 0.065 | 8.0 minutes | Meaningfully better precision for average connectome summaries. |
| 300 | 0.058 | 10.0 minutes | Improved stability for subject-level connectivity analyses and subgroup comparisons. |
Worked Example: Whole-Brain AAL116 Calculation
Suppose you have preprocessed resting-state fMRI data, extracted 116 AAL ROI mean time series, and calculated the full pairwise correlation matrix. Your average off-diagonal correlation is 0.18, you define suprathreshold edges as absolute correlation greater than or equal to 0.20, and you find 1,600 such edges. You have 240 volumes at TR = 2.0 seconds.
- Total possible edges = 116 × 115 / 2 = 6,670.
- Density = 1,600 / 6,670 = 0.2399 = 24.0%.
- Fisher z for mean r = 0.5 × ln((1 + 0.18) / (1 – 0.18)) ≈ 0.1823.
- Approximate total signed strength = 0.18 × 6,670 ≈ 1,200.6 correlation-units across all unique pairs.
- Scan duration = 240 × 2.0 = 480 seconds = 8.0 minutes.
- Approximate SE(z) = 1 / sqrt(237) ≈ 0.0650.
This type of reporting gives a concise but meaningful summary of the connectome: atlas size, edge opportunity space, average strength, transformed effect size, threshold-based density, and temporal sampling precision.
Common mistakes when calculating whole-brain AAL connectivity
- Including the diagonal: self-correlations inflate means and should be excluded.
- Double-counting edges: a symmetric matrix should be summarized using only one triangle.
- Mixing signed and unsigned methods: average signed correlations can be near zero even when strong positive and negative edges exist.
- Ignoring preprocessing choices: motion handling, filtering, and nuisance regression substantially alter correlation structure.
- Comparing densities across different thresholds without disclosure: threshold choice directly changes edge count and graph metrics.
- Failing to transform correlations before group statistics: Fisher z is often the better scale for averaging and inference.
Best practice recommendations
If your goal is a reproducible whole-brain AAL connectivity estimate, explicitly report the atlas version, preprocessing pipeline, denoising strategy, temporal filtering, whether global signal regression was used, the correlation metric, whether values were Fisher z transformed, and the threshold used for any graph-theoretic analysis. Also report the number of time points retained after censoring, because aggressive scrubbing can reduce the effective reliability of the connectome.
For publication-quality methods, it is also wise to state whether the matrix is weighted or binarized, whether negative edges were retained, and whether graph metrics were calculated on a fixed density or fixed correlation threshold. These seemingly small choices often explain why nominally similar studies report different whole-brain summary values.
Authoritative references and data resources
- National Institute of Mental Health for neuroimaging research context and methodology resources.
- University of Wisconsin fMRI Laboratory for educational material on fMRI analysis and signal interpretation.
- USC Laboratory of Neuro Imaging for neuroimaging atlases, connectivity tools, and analysis resources.
Final takeaway
To calculate whole-brain AAL functional connectivity, start with ROI mean time series, compute every pairwise correlation, remove the diagonal, and summarize the off-diagonal values in a transparent way. For AAL90 there are 4,005 unique ROI pairs; for AAL116 there are 6,670. A robust summary usually includes mean correlation, Fisher z, and threshold-based density. The calculator on this page packages those essentials into a fast workflow you can use for planning, reporting, and interpreting your whole-brain connectome results.