CDeep3M - Plug-and-Play cloud based deep learning for image segmentation of light, electron and X-ray microscopy

Abstract

As biological imaging datasets increase in size, deep neural networks are considered vital tools for efficient image segmentation. While a number of different network architectures have been developed for segmenting even the most challenging biological images, community access is still limited by the difficulty of setting up complex computational environments and processing pipelines, and the availability of compute resources. Here, we address these bottlenecks, providing a ready-to-use image segmentation solution for any lab, with a pre-configured, publicly available, cloud-based deep convolutional neural network on Amazon Web Services (AWS). We provide simple instructions for training and applying CDeep3M for segmentation of large and complex 2D and 3D microscopy datasets of diverse biomedical imaging modalities.

Nanoscale 3D EM reconstructions reveal intrinsic mechanisms of structural diversity of chemical synapses

“…Five $35,000-mm 3 SBEM volumes were collected from the dorsal CA1sr of two WT and three Emx1 IRES-Cre /R26 floxstop-TeNT mice (TeNT) at postnatal day 30 (P30) by using raster images with 5.3nm pixels, 2-msec pixel dwell time, and 60-nm Z steps (Figure 1B). Because manual tracing of dense structures in high-resolution 3D EM stacks is extremely time consuming, we built a pipeline for automatic segmentation of plasma membranes and organelles in a cloud-based convolutional neural network, CDeep3M (Haberl et al, 2018). This deep learning AI platform allowed us to make accurate predictions through retraining manually segmented ground truth labels in the Amazon Web Service (AWS), thereby reducing the effort and time by $90% (Figures 1C and 1D).…”

“…Automatic segmentation of different subcellular structures was performed with CDeep3M, a cloud-based platform utilizing a deep convolutional neural network (Haberl et al, 2018). This recently developed tool enables effective processing of multiple common microscopy modalities, including SBEM.…”

“…Since creating training data from scratch is time consuming, we chose to retrain pre-trained neural networks on specific image sets. This method of domain adaptation reduces effort and time by 90%, while still achieving high segmentation accuracy (Haberl et al, 2018). Pre-trained membrane and mitochondria models were downloaded from the publicly available Cell Image Library and retrained for each image stack using small volumes with manually segmented ground truth labels.…”

MeDUsA: A novel system for automated axon quantification to evaluate neuroaxonal degeneration

“…By using deep learning, it was possible to automatically create a mask that extracts axon terminals from a confocal z-stack image. The image processing capabilities afforded by machine learning are powerful, and recently, many quantitative and segmentation methods using machine learning have been published [87][88][89][90]. MeDUsA is a Python-based method specifically designed to count axons; however, it only quantifies the presence of axons and does not capture pre-degenerative signs such as swelling or fragmentation of axon terminals.…”

Chunkflow: hybrid cloud processing of large 3D images by convolutional nets

“…The images are independent with each other and these services are not applicable for a complete 3D image dataset. Recently, Haberl Matthias et al setup a tool in AWS, called CDeep3M, to segment 2D/3D image stacks including training and inference, but it focus on single instance with small image stacks without distributed computation [10]. To our knowledge, there is a lack of a cloud computing tool to perform terabyte or petabyte scale 3D image dataset ConvNet inference with multiple frameworks utilizing both GPUs and CPUs.…”