Large Scale Image Segmentation with Structured Loss based Deep Learning for Connectome Reconstruction

TL;DR

This paper introduces a scalable deep learning approach combining affinity prediction with region agglomeration for neuron segmentation in electron microscopy, achieving significant accuracy improvements and efficient processing of large datasets.

Contribution

The authors develop a novel structured loss and a quasi-linear gradient computation method for 3D U-NET based neuron segmentation, enhancing accuracy and scalability.

Findings

Achieved 27%, 15%, and 250% improvements on three EM datasets.

Predictions enable simple agglomeration that outperforms complex methods.

Method scales linearly, processing large datasets efficiently.

Abstract

We present a method combining affinity prediction with region agglomeration, which improves significantly upon the state of the art of neuron segmentation from electron microscopy (EM) in accuracy and scalability. Our method consists of a 3D U-NET, trained to predict affinities between voxels, followed by iterative region agglomeration. We train using a structured loss based on MALIS, encouraging topologically correct segmentations obtained from affinity thresholding. Our extension consists of two parts: First, we present a quasi-linear method to compute the loss gradient, improving over the original quadratic algorithm. Second, we compute the gradient in two separate passes to avoid spurious gradient contributions in early training stages. Our predictions are accurate enough that simple learning-free percentile-based agglomeration outperforms more involved methods used earlier on inferior predictions. We present results on three diverse EM datasets, achieving relative improvements over previous results of 27%, 15%, and 250%. Our findings suggest that a single method can be applied to both nearly isotropic block-face EM data and anisotropic serial sectioned EM data. The runtime of our method scales linearly with the size of the volume and achieves a throughput of about 2.6 seconds per megavoxel, qualifying our method for the processing of very large datasets.