Matching parcellations using Optimal Transport: a proof of concept - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2017

Matching parcellations using Optimal Transport: a proof of concept

Abstract

Many techniques have been proposed to divide the brain based on structural connectivity. However, even when produced by the same technique, the resulting parcellations tend to differ in the number, shape, and spatial localization of parcels across subject. Matching parcels across subjects is an open problem. We propose to use Optimal Transport (OT) theory to tackle this issue. OT theory studies the efficient transportation of mass between two probability distributions with respect to a certain cost function. As a proof of concept, we show that OT can match parcels of the Desikan atlas across different subjects, using only the structural connectivity fingerprint of each parcel. We use the Desikan atlas since it is used as a prior parcellation in several connectivity studies. For comparison purposes we also match parcels using two additional methods based on the Euclidean and cosine distance. Our results show that the technique based on OT is the most efficient.
Fichier principal
Vignette du fichier
matching-parcellations-optimal.pdf (145.9 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01713345 , version 1 (20-02-2018)

Identifiers

  • HAL Id : hal-01713345 , version 1

Cite

Guillermo Gallardo, Nathalie Thérèse Hélène Gayraud, Maureen Clerc, Demian Wassermann. Matching parcellations using Optimal Transport: a proof of concept. Computational Brain Connectivity Mapping – Winter School Workshop 2017, Nov 2017, Juan-les-Pins, France. ⟨hal-01713345⟩
157 View
47 Download

Share

Gmail Facebook X LinkedIn More