101 labeled brain images and a consistent human cortical labeling protocol.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Accurate labeling of human brain anatomy is crucial for neuroscience research.
• Existing datasets are limited in size, accessibility, and manual labeling detail.
• Automated brain labeling methods require robust, validated datasets for development and evaluation.

Purpose of the Study:

• Introduce the Mindboggle-101 dataset, the largest publicly available manually labeled human brain MRI dataset.
• Present the Desikan-Killiany-Tourville (DKT) protocol for consistent and accurate cortical area labeling.
• Provide benchmarks for evaluating automated brain registration and labeling algorithms.

Main Methods:

• Manual labeling of macroscopic anatomy in 101 healthy participants' MRI scans.
• Development of the DKT protocol using anatomical landmarks and automated label initialization.
• Comparison of manual labels against automated labels from probabilistic and multi-atlas approaches.

Main Results:

• The Mindboggle-101 dataset comprises 101 manually labeled brain MRIs.
• The DKT protocol enhances ease, consistency, and accuracy in cortical labeling.
• Established benchmarks for automated labeling algorithm evaluation.

Conclusions:

• Mindboggle-101 serves as a valuable resource for brain atlases and normative morphometric data.
• The dataset facilitates the development and validation of automated brain labeling techniques.
• This work advances the field of human brain image analysis and computational neuroanatomy.