Sub-thalamic nucleus (STN) location prediction based on ultra high-field MRI statistical shape relationships

Project Summary

The sub-thalamic nucleus (STN) within the sub-cortical region of the Basal ganglia is a crucial targeting structure for Deep brain stimulation (DBS) surgery, in particular for alleviating Parkinson’s disease (PD) symptoms. Volumetric segmentation of such small and complex structure, which is elusive in clinical MRI protocols, is thereby a pre-requisite process for reliable DBS targeting. While direct visualization and localization of the STN is facilitated with advanced high-field 7T MR imaging, such high fields are not always clinically available. 

Themes and Categories
Year

In this work, we focus on the automatic shape prediction of the STN, exploiting the spatial dependency of the STN on its adjacent structures as predictors, some of which are easy to visualize and localize with standard clinical procedures. Variation modes of the STN and its predictors on five high-quality training sets obtained from 7T MR imaging are first captured using a statistical shape model. We then exploit the partial least squares regression (PLSR) method to induce the spatial relationship between the STN and its predictors. Prediction accuracy is evaluated by measuring the shape similarity and the errors in position, size, and orientation between manually segmented STN and its predicted one. Experimental results demonstrate that the proposed approach enables accurate shape and pose prediction of the STN, critical for Parkinson’s DBS, on 7T MR and on clinical 1.5T MR imaging using the spatial relationship between STNs and its predictors.

Related People

Related Projects

A team of researchers associated with the Applied Machine Learning Lab in Duke’s ECE department will lead a team of students in developing novel machine learning techniques that will be used for improving brain computer interfaces (BCIs) using electroencephalography (EEG) data.  Students will learn how to pre-process EEG data, extract EEG features, and train machine learning algorithms for character selection in a spelling interface that allows “locked in” individuals, like Stephen Hawking, to communicate with the outside world.  In addition to developing machine learning algorithms, students will work to develop a dashboard to visualize EEG signals, trained classifier parameters, classifier outputs, and spelling decisions made by the BCI.

 

Project Lead: Leslie Collins

Project Manager: Evan Stump

Volumetric segmentation of sub-cortical structures such as the basal ganglia and thalamus is necessary for non-invasive diagnosis and neurosurgery planning. This is a challenging problem due in part to limited boundary information between structures, similar intensity profiles across the different structures, and low contrast data.