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

IEEE Transactions on Pattern Analysis and Machine Intelligence 2018 Jan FunkeFabian David TschoppWilliam GrisaitisArlo SheridanChandan SinghStephan SaalfeldSrinivas C. Turaga

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... (read more)

PDF Abstract

Evaluation results from the paper

Task Dataset Model Metric name Metric value Global rank Compare
Brain Image Segmentation CREMI U-NET MALA VOI 0.606 # 1
Brain Image Segmentation CREMI U-NET MALA CREMI Score 0.289 # 1
Brain Image Segmentation FIB-25 Synaptic Sites U-NET MALA VOI 2.151 # 1
Brain Image Segmentation FIB-25 Whole Test U-NET MALA VOI 1.071 # 1
Brain Image Segmentation SegEM U-NET MALA IED 4.839 # 1