O'Reilly, MT;
Roth, A;
Häusser, M;
Griffin, LD;
(2011)
A circle-based method for detection of neural fibre cross-sections in classically stained 2D electron micrographs.
In:
Sixth International Workshop on Microscopic Image Analysis with Applications in Biology (MIAAB), Heidelberg.
![[thumbnail of OReilly2011-MIAAB.pdf]](https://discovery.ucl.ac.uk/style/images/fileicons/application_pdf.png) Preview |
PDF
OReilly2011-MIAAB.pdf Download (4MB) |
Abstract
Recent developments in electron microscopy now permit the unambiguous reconstruction of even the smallest neural fibres by human experts. However, manual reconstruction of an interesting volume of neural tissue would take thousands of person-years. Techniques to automate such reconstruction are therefore highly desirable and currently under active development. Here we present a novel circle-based technique and assess its performance on classically stained electron micrographs of the molecular layer of mouse cerebellar cortex. We compare its performance to a recently published pixel-based classifier (ilastik), selected because a similar random forest classifier from the same group has shown promising results on images of neural tissue. The performance of our algorithm and that of ilastik are similar, achieving approximately 50% on an overlap-based f-measure.
Archive Staff Only
 |
View Item |
