MICCAI Challenge Workshop on Segmentation: Algorithms, Theory and Applications ("SATA")

Overview

Segmentation is the essential computational step for attaching contextual meaning to imaging data. Voxel-wise segmentations provide a basis for morphological assessment, inter-subject comparisons, and anatomical modeling. Without robust methods to identify local anatomical structure, we cannot interrogate the relationships between local properties and functional, interventional and demographic measures. Manually drawn regions of interest do not provide the throughput to keep pace with modern studies. Given the diversity of imaging sequences and clinically relevant anatomical targets, it is becoming increasing important to quickly develop methods that accurately segment medical images.

Numerous creative approaches have proposed to use atlas (training) information to apply labels to anatomy. However, it is difficult to evaluate the relative advantages and limitations of these methods as they have been applied on very different datasets. This workshop will provide a snapshot of the current progress in the field through extended discussions and provide researchers an opportunity to characterize their methods on three standardized dataset in a grand challenge. The training/testing datasets will be released after the competition of this challenge.

Objectives

The workshop will present a forum to discuss recent advances in segmentation from multiple perspectives. First, we will host a half-day discussion on methods, theory, and applications (similar to the successful 2011/2012 workshops). Traditional MICCAI format papers will be solicited. Second, we will host a multi-part challenge to evaluate the performance of automated segmentation approaches across three different problem domains. The proceedings will be published in a Springer Lecture Notes in Computer Science (LNCS) publication.

Topics of interests include but are not limited to:

• Atlases: design, selection, and exclusion • Active shape and appearance models, manifold learning • Classification • Manifolds theory and applications for voxel, volume, or surface representations • Multi-atlas informed and augmented approaches, including shape modeling • Multi-atlas registration • Segmentation • Shape • Statistical methods for label fusion • Theory and applications with discrete, continuous, or non-traditional label types • Visualization and hypothesis exploration approaches using multi-atlases

Format: Workshop/Tutorial

The workshop will be presented in a full day session divided into novel presentations and challenge discussion. The morning session (9-12:30) will be devoted to discussion of technical contributions while the afternoon (14:00-17:30) will be devoted to the segmentation challenges. Sessions will be organized similarly with 5-6 paper presentations (20 minutes & 10 minutes of discussion), a poster session (1 hour), and one invited review presentation summarizing results from all submissions (10 minutes & 20 minutes of discussion). Both sessions will include a 30 minute coffee break.

Banner.jpg