These manually labeled MRI brain scans can be embedded into a medical device or image analysis software where precise neuroanatomical localization is important such as automated brain labeling algorithms, training deep learning, neuroanatomical visualization, and stereotactic radiosurgery, to name a few application areas.  Unlike an atlas of a single subject, our database of scans provides an indication of the variation of the living human brain.  Currently there are 114 scans with subject ages from 5 to 96 years old.

We have the most comprehensively labeled scans and the largest number of scans available anywhere.  We label the entire brain and divide the cortex into regions based on gyral and sulcal landmarks using 1) the “General Segmentation” Protocol defined by the MGH Center for Morphometric Analysis (see here), and 2) the brainCOLOR Cortical Parcellation Protocol (from here).  The consistency of our results is excellent because we embed these protocols into our labeling software as “SegMentor” scripts which not only document the procedures, but also automate as much as possible leaving only steps that require human intervention and oversight.

These are the data sets that we have available now (or will be available soon):

“ADNI30” consists of 30 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI) project and the “20Repeat” set is particularly good for testing algorithms because it is 40 repeat scans (the same 20 subjects scanned twice and labeled twice).  “CANDI13” are 13 young subjects, “OASIS30” are young through older adults, and “Colin27” is the famous averaged scan from the McConnel Brain Imaging Centre of a single subject scanned 27 times. 

We discuss the “20Repeat” scans in our 2015 Neuroscience poster and gave an upper bound on the Dice Similarity Coefficients that are possible with manual labeling of repeat scans.  We made some additional corrections since the poster came out and the updated numbers are at http://www.neuromorphometrics.com/Overlaps_Nov192015/.

Our mission is to build an ever-improving quantitative model of the structure of the living human brain   Our goal is to label every brain on the planet so we can capture and understand the variation in human neuroanatomy.  We use revenue from licensing labeled scans to label additional scans (larger age range, more diverse subjects, etc.) and to improve our existing database by correcting errors, refining anatomical borders, and labeling additional regions.

Neuromorphometrics spun out of the Harvard/MGH Center for Morphometric Analysis with the help of SBIR grants from the NIH and we have been providing brain-labeling services for almost two decades.  It used to take two weeks to label a single brain.  Now, we do it in 2-3 days and we do a far better job with a lot more regions of interest.  As we developed software and anatomical protocols with an eye toward reliability and reproducibility, we decreased the time required for labeling by concentrating on work-flow optimization and adding automation.  One can’t do automated labeling without knowing how to do it manually first.  

We can also label scans that you provide and we are very interested in labeling white matter anatomy as seen in diffusion-weighted MRI scans.  If you want an aggregate version of our data, we can provide it as a probabilistic atlas.

The cost to label a single scan is $2449 (USD).  For academic institutions, two “subscriptions” are available on-line for less than the cost of a single scan.  For commercial entities, we license each neuroanatomically labeled MRI brain scans for $1500 or as an introductory special price, we’ll provide every scan we currently have for $[please ask].  This is a great deal because it has taken many years to generate this data and it requires a lot of expertise to reliably identify anatomy as it appears in MRI, not to mention the fact that it is intensely tedious work.  If you refer another academic customer to us, you and your friend will get a discount.  If you refer a new corporate customer, you get a commission.

For more details, see Neuromorphometrics.com, call +1 617-776-7844 or email Andy@Neuromorphometrics.com.