12 JunRadiology… moves… so… slowly.
We’ve been using a programmable keypad and graphics tablet to annotate MRI brain anatomy for years. It’s nice to hear that Radiology is figuring out user interface issues: Denton, Kevin, Irfanullah Haider, Jacqueline Hill, Suzanne L. Hunt, and Ryan Ash. “Of Mice and Roentgen: Radiologist Satisfaction with a Non-conventional 13-Button Mouse—One Institution’s Experience.” Journal of digital […]
Read More ...
24 AprArtificial Intelligence Hype
Interesting discussions about AI in healthcare at the World Medical Innovation Forum in Boston. There seems to be general agreement on this: “Artificial Intelligence is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it.” Don’t believe […]
Read More ...12 DecRSNA 2017 Summary: AI all the rage, access to annotated data a challenge
There is a nice description of AI in medical imaging from Signify Research that concludes, “Access to radiologist annotated data remains a major challenge for many algorithm developers, who must be prepared to invest significant time and money in data curation. Companies with innovative strategies for obtaining data will have a big advantage.” Hey, that’s what we […]
Read More ...15 NovSfN 2017 Poster
Using a probabilistic atlas to improve manual parcellation of the cerebellum Andrew Worth1, Jason Tourville2 1 Neuromorphometrics, Inc., Somerville, MA 2 Dept. of Speech, Language, and Hearing Sciences, Boston University, Boston, MA Abstract (download poster, 6.1MB) In order to develop automated methods for labeling anatomy in MRI brain scans, regions of interest (ROIs) must first be labeled manually and […]
Read More ...15 JunDeep learning needs [expert] data!
“DeepMind Shows AI Has Trouble Seeing Homer Simpson’s Actions” (link), describes how machine learning (deep learning, neural networks, artificial intelligence) needs annotated data to train on so they went with Amazon’s Mechanical Turk service to create it. That works because we all have visual systems that to a great job. But some deep learning applications need expertise beyond what […]
Read More ...13 NovSfN 2016 Poster
Anatomical evaluation of Colin 27 against a database of labeled brain scans Andrew Worth1, Jason Tourville2 1 Neuromorphometrics, Inc., Somerville, MA 2 Dept. of Speech, Language, and Hearing Sciences, Boston University, Boston, MA Abstract (download poster, 7.8MB) The Colin 27 average brain is a stereotaxic registration model from the Montreal Neurological Institute (see Holmes et al. 1998). This T1 […]
Read More ...15 Oct18 Years Old and Celebrating
Neuromorphometrics turned 18 this month and to celebrate, we’re giving gifts! We have some of the most comprehensively and consistently labeled neuroanatomical regions in MRI brain scans and we’ll be providing our entire manually-labeled database to two students and/or post-docs at no cost. If you have an exciting idea about how you would use our […]
Read More ...03 MayDice Similarity Coefficients (DSCs), How Good is “Good Enough”?
If you have a method for automatic segmentation (labeling anatomy) of the human brain in MRI scans, you can test it using a ground truth segmentation by calculating the Dice Similarity Coefficient (DSC). But what affects DSCs, and how do you know if a DSC value is good or bad? Some neuroanatomical regions such as ventricles are relatively easy to segment because […]
Read More ...08 MarAutomatic labeling of MR brain images: performance approaching human accuracy
C. Ledig and D. Rueckert, in Chapter 14 of this recent book, describe segmentation of MR brain images, and conclude: “The performance of current state-of-the-art techniques is starting to approach that of human observers in terms of accuracy. However, the robustness of current approaches is not yet comparable to human observers. This is especially true […]
Read More ...