Kavli Affiliate: David Kleinfeld
| Authors: Kui Qian, David Kleinfeld, Beth Friedman and Yoav Freund
| Summary:
One routine and necessary, yet time-consuming task in neuroanatomy is the annotation of labeled cells relative to the background. Currently, staining and imaging techniques enable the marking of specific cell groups with fluorescent dyes. Modern high throughput scanning microscopes allow high resolution multi-channel imaging of the sectioned brain. However, manual identification of labeled cells is prohibitively time consuming. We present a methodology for developing digital assistants that significantly reduce the labor of the anatomist while improving the consistency of the annotation. Machine learning methods are combined with a rigorous way to measure the confidence of the predictions. We compare the error rate of our method to the disagreement rate between human anatomists. This comparison demonstrates that our method can reduce the time to annotate as much as ten-fold without significantly increasing the error rate.