Tag Archives: neuroanatomy

Advances in brain imaging

Klein et al 2017The diagnosis of human brain abnormalities depends on knowing the norm and yet defining the range of normal variation is still far from resolved. Understanding what is within the normal human range has been limited by samples and the constraints of producing accurate brain mapping. Access to large brain imaging databases has been possible for a while but producing reliable atlases of key structures including folding patterns (sulci, gyri and fundii), volumes and major shape changes has not had large enough sample sizes to reliably grasp the range of normal brain variation. Current approaches have relied on highly skilled professionals to assess neuroanatomy. While this approach is adequate, it does introduce an inherent level of subjectivity and potential bias with each neuroanatomist dependent on the individual level of experience. To begin reducing this error while increasing sample sizes, new computational technologies allow more automated imaging processes that combine speed and quality.

Mindboggle is a new software platform recently released after development through a long-term research project addressing a need for integrating morphometry (measurements of morphology) to assess the quantitative differences in brain structure. Mindboggle relies on specially developed algorithms to segment brain tissue in MRI images, produce volumetric and structural parallelization of the brain and asses shape variation. Klein and colleagues highlighted issues with similar algorithm-based software that produced errors in segmenting brain from non-brain tissue. Freesurfer was shown to underestimate grey matter while overestimating white matter, while ANTs included more grey matter yet sometimes excluded white matter that extended deep in gyral folds. To resolve this issue, Mindboggle employed a hybrid algorithm that overlays the Freesurfer and ANTs segmentation imaging then combines these to produce a more faithful imaging set negating any errors in volume estimates, folding patterns or shape differences. Further results indicated the geodesic algorithm produced an exaggerated depth for brain regions like the insula, while the time depth algorithm unique to Mindboggle produced more valid results for shallow brain structures than other comparable algorithms. Finally, Mindboggle was shown to be reliable with minimal error estimate showing a consistently greater shape difference between left and right hemispheres than the difference between repeated scans of the same individuals.

Mindboggle also introduced many new and innovative features for extracting and measuring fundii but these algorithms have not yet been thoroughly evaluated. Additionally, the Mindboggle algorithms are developed for human brain anatomy and expansion into non-human neuroanatomy has not yet been fully developed. The potential of Mindboggle and similar platforms lies in the allowance to expand knowledge of normal human brain variation by using much larger samples to more accurately capture the normal range in human neuroanatomy to better inform diagnoses of brain abnormalities.

Alannah Pearson