iiinova — where your digital self keeps evolving

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

More and better fMRI data remain ideal. But when scan time is limited, PCM makes uncertainty explicit and allows it to guide more cautious, precise individualized targeting. A huge thanks to our collaborators for making this work possible! #Neuroimaging #fMRI #Neuromodulation

040

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

Why does all of this matter for neuromodulation? If the goal is to stimulate a specific functional network, false-positive assignments can mean stimulating the wrong network. PCM lets us prioritize regions we're most confident belong to the intended network.

Illustration of why network-assignment confidence matters for precision neuromodulation. An example stimulation site targets an individualized functional network. Without confidence thresholding, the target includes more false-positive territory belonging to other networks. PCM thresholding produces a smaller, more precise target with fewer false positives, at the cost of excluding some true-positive network territory.

130

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

Could thresholding improve reliability simply by removing what makes each person unique and leaving a generic network “core”? That’s not what we found. PCM increased agreement with each person’s independent reference while preserving, and even increasing, within-vs-between-subject separation.

Test of whether PCM improves reliability while preserving individual-specific network organization. Action Mode Network maps from four participants are compared with each participant's independent reference and with the references of other participants. PCM thresholding increases within-person agreement while maintaining lower between-person agreement, increasing the separation between within- and between-subject similarity.

120

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

We tested PCM across scan durations and confidence thresholds. Both more data and PCM thresholding increased targeting precision, but PCM provided its largest gains when data were limited. At 5 min, PPV increased from 0.47 → 0.66 with thresholding, vs 0.71 at 70 min.

Quantification of PCM performance across scan durations and confidence thresholds. Positive predictive value increases with both additional fMRI data and more stringent PCM thresholding, with the largest thresholding benefits at short scan durations. Additional panels show reductions in false positives, modest losses in true-positive coverage, and the resulting precision-versus-coverage tradeoff.

120

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

Once uncertainty is quantified, it becomes actionable. Thresholding PCM maps removes unstable assignments. As confidence increases, off-network assignments are progressively removed and the retained network better matches an independent 70-min reference. The benefit is largest with limited data.

Example showing PCM confidence thresholding of the Action Mode Network using 5 versus 70 minutes of fMRI data. Removing progressively lower-confidence network assignments increasingly restricts the map to regions matching an independent 70-minute reference. The improvement is largest with limited data: in the 5-minute example, positive predictive value increases from 0.35 without thresholding to 0.57 at the 99% confidence threshold.

120

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

So how can we know which parts of an individualized network map we can actually trust? PCM bootstraps the available fMRI time series, repeatedly reruns network detection, and asks how consistently each location is assigned to each network. The result: a map of assignment confidence.

Overview of the Precision Confidence Mapping (PCM) workflow. Resting-state fMRI data are divided into temporal partitions and repeatedly bootstrap resampled. Network detection is performed independently on each resampled dataset, producing many network assignment maps. These are aggregated to estimate how consistently each brain location is assigned to a network, generating individualized network confidence maps.

120

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

Functional networks vary substantially across people, making individualized mapping attractive for neuromodulation. But precision comes at a cost: more time in the scanner. With limited data, maps contain more off-network assignments, increasing the risk of targeting the wrong network.

Schematic illustrating the motivation for Precision Confidence Mapping. Group-average functional network maps provide consistent anatomy but miss individual-specific network organization. Individualized mapping captures this variability, but maps generated from only 5 minutes of resting-state fMRI contain substantially more incorrect network assignments than maps generated from 70 minutes. Purple shows correctly assigned Action Mode Network territory, striped regions show misassignments, and yellow outlines indicate the independent reference network.

120

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

This was a huge team effort with @smnelson.bsky.social @drdamienfair.bsky.social, @roberthermosillo.bsky.social @tervoclemmensb.bsky.social @drjuliamoser.bsky.social @katejgodfrey.bsky.social & many others. Very grateful to everyone who helped make PCM possible! So, why was PCM needed? 👇

130

Bene Ramirez @benediktramirez.bsky.social · 23/08/2026

Excited to share our new preprint introducing Precision Confidence Mapping (PCM), a framework for estimating how confident we should be in individualized functional network assignments from resting-state fMRI. Why does this matter for precision neuromodulation? 🧵👇 www.biorxiv.org/content/10.6...

biorxiv.org

13815

Reposted by Bene Ramirez

Julia Moser @drjuliamoser.bsky.social · 12/11/2025

A particular shoutout to Essa Yacoub, @drdamienfair.bsky.social, Alireza Sadeghi-Tarakameh, Jed Elison, @benediktramirez.bsky.social, @smnelson.bsky.social and many more for their important contributions!!

032