Perlis, Gainer, Churchill, Kohane, Weber
Anti-Depressant Pharmacogenetic Consortium led by Perlis with support from i2b2 is going to be launched soon.
NLP review by the psychiatrists on deck. It's hard slogging.
Early end to the day due to snow storm.
Valtchinov, Richter, Keogh, Weber, Kohane, Churchill
Reviewed high performance computing cell that is likely to be released in a few months.
Reviewed Relevance Network desk top application. Good progress but not yet ready for release.
Wei, Plenge, Churchill, Kohane, Gainer, Murphy, Karlson, Tianxi, Szolovits, Churchill
Descriptive statistics of NLP corpus
30K (RA or CCP positive) -> Random 650 for chart review: 128 confirmed to have RA, 58 possible. This is gold standard review. So this means that there are radiologists (for example) coding RA where there is no RA, or rheumatologists coding for "rule out RA:
Tianxi reviewed feature selection via adaptive LASSO. Currently, the lower bound on the 95% confidence interval on the Positive Predictive Value seems to be high (stand by... we had an internal argument how the PPV was calculated)?
Present: Iosifescu, Gainer, Churchill, Kohane, Zeng
Rate limiting step is now not the image processing of the MRI images, but the cycle from NLP identification of cases, submitted the case list to the MGH Radiology Department and then have the images stored on our scratch space server. Then the process of sorting through which images are complete enough, and then using Freesurfer to analyze them is relatively rapid.
Plange, Karlson, Wei, Murphy, Kohane, Szolovits, Kohane, Gainer, Kat
Cases will be completed by January 1.
Controls to follow shortly thereafter.
