01 AprReview Article on MRI Segmentation of the Human Brain

A nice review article was recently published, “MRI Segmentation of the Human Brain: Challenges, Methods, and Applications.”  The authors give a good overview of the state of the art!

They mention the main problems that need to be addressed ([Worth 1997] describes additional issues), and the different ways to do neuroanatomical labeling, from manual to the popular automated methods, and they also discuss how results are validated.  Ivana Despotović, Bart Goossens, and Wilfried Philips describe the benefits and problems associated with manual methods:

The manual method is believed to be the most accurate because of the difficulty to accurately and reliably delineate structures in medical images….  This manual segmentation is not only tedious but also particularly prone to errors. … Also, manual segmentation results are often difficult and even impossible to reproduce, because even experienced operators show significant variability with respect to their own previous delineation.

At Neuromorphometrics, we address this problem by embedding the precisely-defined labeling protocol into the labeling software using “SegMentor” scripts.  This not only speeds up the labeling, but also increases the consistency of the results.

The authors cite the IBSR data as, “the most popular repository with real MRI data used for validation of brain MRI segmentation methods.”  This is certainly true because it is publicly available (see, for instance, this recent article).  Neuromorphometrics offers a data set that is even better: more subjects and regions of interest, and done on higher resolution and better quality MRI scans.  While we don’t release it for free, it is very inexpensive compared to the cost of producing such a “gold standard” and is the most consistently labeled data available (cf. this discussion).

We’re working on completing the labeling of 20 subjects that were scanned twice and we labeled both scans.  Let me know if you are interested in getting ahold of that (it may happen faster).

[Worth 1997] Worth, Andrew J., Nikos Makris, Verne S. Caviness Jr, and David N. Kennedy. “Neuroanatomical segmentation in MRI: technological objectives.” International Journal of Pattern Recognition and Artificial Intelligence 11, no. 08 (1997): 1161-1187.