In the evolving landscape of brain-wide association studies, a recent investigation underscores a pivotal strategy that intriguingly challenges conventional wisdom: extending the duration of functional magnetic resonance imaging (fMRI) scans not only sharpens predictive accuracy but also reduces overall research costs. This paradigm shift confronts the traditional notion that shorter scans paired with larger sample sizes always yield the most efficient outcomes, revealing instead a nuanced equilibrium shaped by the interplay of budgetary constraints, scan time, and participant overhead.
At the heart of this research lies a comprehensive theoretical model that intricately maps the relationship between two core parameters—sample size and scan duration per individual participant—and their collective influence on the accuracy of phenotypic prediction based on fMRI data. Contrary to prevailing assumptions that prioritize maximizing sample sizes, the model reveals that although larger sample sizes do enhance prediction fidelity, increasing scan time per participant exerts a comparably potent impact. This insight is critically important, as it informs researchers on strategically balancing these variables to optimize study design.
What adds complexity to this balance is the fundamental asymmetry rooted in the overhead costs associated with each study participant. Beyond the direct expense of conducting the scan itself, these overheads include expenses linked to participant recruitment, administering neuropsychological assessments, acquisition of additional imaging modalities such as anatomical T1 or diffusion MRI, and other biomarker measurements like positron emission tomography (PET) or blood tests. Frequently, these overhead costs surpass the direct scanning expenses, tipping the cost-benefit analysis in favor of longer scan times with fewer participants rather than many participants with brief scans.
.adsslot_VnoeSNRdyj{width:728px !important;height:90px !important;}
@media(max-width:1199px){ .adsslot_VnoeSNRdyj{width:468px !important;height:60px !important;}
}
@media(max-width:767px){ .adsslot_VnoeSNRdyj{width:320px !important;height:50px !important;}
}
ADVERTISEMENT
The study’s robustness is bolstered by an analysis spanning nine diverse datasets, encompassing six resting-state fMRI cohorts alongside three task-based fMRI datasets from the Adolescent Brain Cognitive Development (ABCD) initiative. With 76 phenotypes examined across these datasets, the theoretical model exhibited exceptional goodness-of-fit, boasting an average coefficient of determination (R²) of 89%. Importantly, these datasets were heterogeneous, representing multiple fMRI acquisition techniques, coordinate systems, and demographic variables including racial backgrounds and clinical conditions, thereby underscoring the generalizability of the model’s conclusions.
Within each phenotype, the predictive accuracy was normalized relative to each phenotype’s maximum theoretical accuracy, thereby yielding a dimensionless fraction reflecting potential gains in prediction precision. When this metric was averaged over all phenotypes under the assumption of a tenfold cross-validation scheme, a strikingly high correlation with empirical results emerged (Pearson’s r = 0.97). This near-perfect alignment underscores the predictive power and practical utility of the model across varied biological and experimental conditions.
One of the most illuminating facets of the analysis interrogated the cost-efficiency landscape by simulating 108 scenarios that encompassed varying accuracy targets, scan costs per hour, and participant overhead costs. Strikingly, in 85% of these scenarios, the cost-optimal scan duration exceeded 20 minutes, defying the prevailing notion that ultra-short scans are universally most economical. This finding refocuses attention on the nuanced role that scanning length plays in balancing costs and prediction outcomes.
Quantifying the financial implications, extending scan time to 30 minutes demonstrated remarkable cost savings relative to 10-minute scans, with up to a 22% reduction in total expenditure for achieving comparable prediction accuracies. This counterintuitive result arises because longer scan durations per participant reduce the need for vastly increasing sample size, thereby mitigating overhead expenses that accumulate with each additional participant.
From a practical standpoint, researchers equipped with fixed fMRI budgets can employ this model to pinpoint the optimal trade-off between sample size and scan time that maximizes phenotypic prediction accuracy within their financial constraints. For instance, under a hypothetical budget of $1 million with scan and overhead costs each pegged at $500, the model advises a scan time of approximately 34.5 minutes per participant. However, if additional modalities such as PET are incorporated, thereby inflating overhead costs dramatically to $5,000 per participant, the optimal scan time surges to nearly 160 minutes, underscoring the sensitivity of scan time optimization to overhead considerations.
Another essential observation stems from the shape of the cost-accuracy curves, which reveal an asymmetrical pattern characterized by a steep initial increase in accuracy with scan duration followed by a more gradual decrease beyond the optimum. This asymmetry suggests that overshooting the optimal scan time—err on the side of slightly longer scans—is preferable to undershooting, as the penalties for too-short scans are considerably more severe than the diminishing returns of overly long scans.
The implications of this work extend well beyond empirical fMRI investigations; they confront systemic inefficiencies entrenched in neuroimaging research infrastructure and funding models. The nuanced understanding of cost structures and prediction accuracy afforded by this study equips investigators with the tools to design more efficient studies that judiciously allocate resources, potentially accelerating discoveries in brain science while curtailing wasteful spending.
Moreover, the broad representativeness of the datasets analyzed—spanning diverse populations, age ranges, clinical statuses, and neuroimaging modalities—strengthens the promise of these findings being adapted universally. Whether in developmental studies of children, analyses of neurological and psychiatric populations, or broad population-based neuroscience efforts, tailoring scan durations according to optimized models promises to enhance reproducibility and robustness of brain-behavior associations.
This research also opens avenues for integrating multimodal imaging data streams systematically into cost-accuracy frameworks. As collecting additional biomarkers significantly influences participant overhead, future work may refine these models further to capture the compounded effects of multimodal acquisition on optimal scan parameters. Such integration will be essential as neuroimaging moves increasingly toward comprehensive, multimodal datasets that promise richer phenotypic characterizations.
Ultimately, the findings urge a shift in prevailing neuroimaging dogma, moving the community away from rigid adherence to maximizing sample size at the expense of scan quality and duration. Instead, a more dynamic and economically mindful approach is proposed—one that recognizes the profound impact of longer imaging sessions on predictive modeling capability and study economy, a strategy that could redefine best practices in brain-wide association research going forward.
In conclusion, the intricate balance between sample size, scan duration, and overhead costs illuminated by this work offers an essential roadmap for future neuroimaging studies. Embracing longer scan durations, despite their initial cost impression, may paradoxically yield more accurate and cost-effective scientific insights. As brain-wide association studies continue to scale and evolve, such nuanced methodological guidance is invaluable, promising a future landscape where efficiency and excellence go hand in hand.
Subject of Research: Brain-wide association studies; Functional MRI scan optimization; Prediction accuracy and cost-efficiency in neuroimaging.
Article Title: Longer scans boost prediction and cut costs in brain-wide association studies.
Article References:
Ooi, L.Q.R., Orban, C., Zhang, S. et al. Longer scans boost prediction and cut costs in brain-wide association studies. Nature (2025). https://doi.org/10.1038/s41586-025-09250-1
Image Credits: AI Generated
Tags: balancing budget and research efficiencybrain-wide association studiescost-effective brain imaging strategiesenhancing phenotypic prediction accuracyfunctional magnetic resonance imaging advancementslonger fMRI scan benefitsneuroscience study design optimizationoptimizing research costs in neuroscienceparticipant overhead expenses in researchpredictive accuracy in brain researchsample size versus scan durationtheoretical models in fMRI studies
