Iosifescu, Smoller, Perlis et al, with visitors from Erlangen.
Discussed how to enrich the collection of the patients for MRI's of sufficient quality for segmentation.
Sergei showed his latest version of the NLP annotator and active learning program. He showed the configuration files as well which allow the annotator to be quickly changed for another DBP. Zak suggested including Patrick and Nathan in future conversations.
pharmacovigilance Discussed the interim results for the project.
Discussed how to allocated new resources required for i2b2 collaborating grants (R01s and U01s).
Perlis, Iosifescu, Smoller et al.,
Discussed the low yield of sufficient quality on clinical MRI's for segmentation ( < 10%). Also reviewed sufficiency for DTI.
Almost finished reviewing 140 gold standard cases for MDD. These will then be used to assess the various NLP. Reviewed the per report and the per case classification task.
Reviewed the pharmacovigilance task, the matter of timing and the use of survival rather than epochs as outcome variables.
Sergei will demonstrate the latest version of the classification/learning tool next Friday.
Roy mentioned that we should stratify the accuracy of the classifier based on the confidence of the classification annotation by the clinician.
Yesterday 2,379 blood samples arrived at the Broad Institute from i2b2's RA DBP. DNA will be fully extracted by end of the Summer.
For 384 SNP, there are several big questions to ask;
1. RA risk alleles. Are they reproduced? (45 SNP's)
2. Meta-analysis/
3. Auto-antibodies in RA.
4. Ancestry markers.
Here is Robert in celebratory garb.
