OptimalBattery

The OptimalBattery project implements the analyses underlying the paper "Multi-Task Batteries for Precision Functional Mapping" by Arafat, Nettekoven, Xiang & Diedrichsen (2026).

Dependencies

Diedrichsen Lab packages:

• Diedrichsenlab/Functional_Fusion
• Diedrichsenlab/PcmPy

Other dependencies:

• numpy
• pandas
• torch
• scipy
• matplotlib
• seaborn
• nibabel

Core Modules

Module	Description
construct.py	Optimal battery construction using eigenvalue-based criteria (variance, inverse trace, log determinant)
simulate.py	Simulation functions for parcellation and connectivity modeling
evaluate.py	Evaluation metrics for assessing battery performance
estimate.py	Estimation of parcellations (U matrices) and functional profiles (V matrices)
design.py	Experimental design matrix generation for grouped vs. interspersed designs
util.py	Utility functions for matrix operations and data preprocessing
plot.py	Visualization functions for publication-quality figures

Scripts / Code to replicate different parts of the paper

Section 1: Functional Localization (Fig 1)

Single-contrast vs. multi-task localizers

Script	Description
scripts/_1functional_localization/1a.localization_fSNR.py	Estimate fSNR distribution from MDTB dataset (Fig 1d)
scripts/_1functional_localization/1b.localization_sim.py	Simulate single-contrast vs multi-task localization (Fig 1e-g)
scripts/_1functional_localization/1c.localization_real_fSNRxSize.py	fSNR vs ROI size correlation in real data
scripts/_1functional_localization/1d.localization_real.py	Empirical localization of cerebellar language region (Fig 1h-k)

Figures: paper_figures/1.localization.ipynb

---

Section 2: Optimal Battery Selection (Fig 2)

Parcellation and connectivity simulations & real data

Script	Description
scripts/_2battery_selection/2a.parcellation_sim.py	Parcellation simulation across battery sizes (Fig 2a)
scripts/_2battery_selection/2b.parcellation_real_cortical.py	Neocortical parcellation on MDTB (Fig 2c)
scripts/_2battery_selection/2c.parcellation_real_cerebellar.py	Cerebellar parcellation on MDTB (Fig S2b)
scripts/_2battery_selection/2d.connectivity_sim.py	Connectivity modeling simulation (Fig 2b)
scripts/_2battery_selection/2e.connectivity_real.py	Neocortex-cerebellum connectivity on MDTB (Fig 2d)

Figures: paper_figures/2.battery_selection.ipynb

---

Section 3: Experimental Design (Fig 3 & 4)

Grouped vs. interspersed designs, temporal autocorrelation, carryover effects

Script	Description
scripts/_3experimental_design/_1hcp_task_covariance.py	HCP task covariance matrices & baseline noise estimation (Fig 3a-b)
scripts/_3experimental_design/_2sim_blocked_vs_interspersed.py	Simulate grouped vs interspersed design reliability (Fig 3c-d)
scripts/_3experimental_design/_3.temporal_autocorrelation.py	Temporal autocorrelation analysis on MDTB (Fig 4a)
scripts/_3experimental_design/_4.carryover.py	Task carryover effect estimation (Fig 4b)

Figures: paper_figures/3.experimental_design.ipynb

---

Supplementary Materials

Script	Description
scripts/z.supp/1.group_vs_indi_cov.py	Group vs individual covariance matrix comparison (Fig S1)
scripts/z.supp/3.task_library.py	Build the combined task activation library (79 conditions)
scripts/z.supp/4.region_batteries.ipynb	Optimal batteries for Motor, PFC, Cerebellum (Table S1)

---

---

License

MIT License - Diedrichsen Lab (2024)