Automatic segmentation of tractography data into pathways/tracts is a problem traditionally addressed by means of unsupervised techniques, i.e., clustering streamlines. The core of this work is to adopt instead a supervised approach, learning from the segmentation made by an expert neuroanatomist in order to predict tracts in new brains. In this talk a novel set of supervised approaches to the tract segmentation problem will be illustrated. The proposed solutions are based on machine learning topics like “supervised clustering”, “learning with similarity functions” and “transduction”. These solutions allow to exploit both diffusion and functional MRI data, to avoid co-registration between different subjects and to predict tracts in hemispheres different from the training example. Preliminary results support these claims. An intended goal of this talk is to open a discussion on how to map the building blocks of the proposed methods into the PyMVPA framework in order to support tractography data analysis natively and, more in general, to provide novel machine learning approaches to the users.

Supervised Tract Segmentation with Diffusion and Functional MRI Data

Olivetti, Emanuele;Veeramachaneni, Sriharsha;Avesani, Paolo
2009-01-01

Abstract

Automatic segmentation of tractography data into pathways/tracts is a problem traditionally addressed by means of unsupervised techniques, i.e., clustering streamlines. The core of this work is to adopt instead a supervised approach, learning from the segmentation made by an expert neuroanatomist in order to predict tracts in new brains. In this talk a novel set of supervised approaches to the tract segmentation problem will be illustrated. The proposed solutions are based on machine learning topics like “supervised clustering”, “learning with similarity functions” and “transduction”. These solutions allow to exploit both diffusion and functional MRI data, to avoid co-registration between different subjects and to predict tracts in hemispheres different from the training example. Preliminary results support these claims. An intended goal of this talk is to open a discussion on how to map the building blocks of the proposed methods into the PyMVPA framework in order to support tractography data analysis natively and, more in general, to provide novel machine learning approaches to the users.
2009
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11582/44792
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
social impact