Adding large EM stack support

Glendon Holst, Stuart Berg, Kalpana Kare, Pierre J. Magistretti, Corrado Cali

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Scopus citations

Abstract

Serial section electron microscopy (SSEM) image stacks generated using high throughput microscopy techniques are an integral tool for investigating brain connectivity and cell morphology. FIB or 3View scanning electron microscopes easily generate gigabytes of data. In order to produce analyzable 3D dataset from the imaged volumes, efficient and reliable image segmentation is crucial. Classical manual approaches to segmentation are time consuming and labour intensive. Semiautomatic seeded watershed segmentation algorithms, such as those implemented by ilastik image processing software, are a very powerful alternative, substantially speeding up segmentation times. We have used ilastik effectively for small EM stacks – on a laptop, no less; however, ilastik was unable to carve the large EM stacks we needed to segment because its memory requirements grew too large – even for the biggest workstations we had available. For this reason, we refactored the carving module of ilastik to scale it up to large EM stacks on large workstations, and tested its efficiency. We modified the carving module, building on existing blockwise processing functionality to process data in manageable chunks that can fit within RAM (main memory). We review this refactoring work, highlighting the software architecture, design choices, modifications, and issues encountered.
Original languageEnglish (US)
Title of host publication2016 4th Saudi International Conference on Information Technology (Big Data Analysis) (KACSTIT)
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
ISBN (Print)9781467389563
DOIs
StatePublished - Dec 1 2016

Fingerprint Dive into the research topics of 'Adding large EM stack support'. Together they form a unique fingerprint.

Cite this